>
volgeFlexera · 2027
TECHNOLOGY ASSET MANAGEMENT · VOICE OF VOLGE

The installed liability

See the landscape, before it bills you. Installed software accrues liability on three axes: audit, breach, and obsolescence. Seeing the whole landscape is what lets you comply, secure, and retire on your own terms.

Voice of Volge is a series of books, and the firm's thought leadership on the solutions and the domains Volge specialises in. Each edition takes one of them and sets out what the technology makes possible, what it makes mandatory, and what an operator has to put in place before either is true of them.

See the landscape, before it bills you. Installed software accrues liability on three axes: audit, breach, and obsolescence. Seeing the whole landscape is what lets you comply, secure, and retire on your own terms.

Figures carry their basis and their date. Where published estimates disagree, the range is given and the basis named rather than averaged away, and where a widely quoted number does not survive checking it is corrected here rather than repeated. Every chart plots a figure stated in the text, and nothing is projected beyond what a named source projects.

Contents

  1. 02You cannot comply with, secure, or retire what you cannot see.
  2. 02About Flexera
  3. 03You are liable for everything you have ever installed.
  4. 03Comply.
  5. 07Secure.
  6. 08Retire.
  7. 11Controls, sequencing and operations
  8. 14The owned landscape in the Gulf and India
  9. 19The operator's read.
  10. 21The asset and security leadership agenda
  11. 22Appendices
Voice of Volge is a series of books, and the firm’s thought leadership on the solutions and the domains Volge specialises in.
Flexera · 2027
Technology Asset ManagementVoice of Volge01 •

You cannot comply with, secure, or retire what you cannot see.

Everything an or­gan­isa­tion in­stalls and keeps is quietly work­ing against it. Soft­ware does not merely de­pre­ci­ate. It ac­crues li­ab­il­ity on three axes at once, the audit you can­not pass, the vul­ner­ab­il­ity you have not patched, and the ver­sion that has slipped past end-of-sup­port. The asset re­gister calls it an asset. The aud­itor, the at­tacker, and the vendor’s law­yer each call it some­thing else.

The reason this com­pounds is that own­er­ship is dif­fuse and vis­ib­il­ity is poor. No single leader owns the in­stalled land­scape end to end: pro­cure­ment sees the con­tracts, se­cur­ity sees the vul­ner­ab­il­it­ies, IT asset man­age­ment sees the de­ploy­ments, and no one sees the whole. Each blind spot is someone else’s lever­age, a true-up bill, a breach, an ob­sol­es­cence dead­line that ar­rives without warn­ing.

This re­port is for the CIO and CISO, and the asset, pro­cure­ment and se­cur­ity lead­ers who an­swer to them, who carry that li­ab­il­ity in 2026. It is a gov­ernance ar­gu­ment rather than a product tour. The ques­tion is what it takes to see the soft­ware and AI you own, and then to com­ply with it, se­cure it, and re­tire it on your terms rather than an aud­itor’s. The fig­ures are sourced and dated; the worked ex­ample is de­lib­er­ately mundane.

Flex­era sup­plies the con­crete in­stru­ment, and Volge dis­closes its in­terest as a part­ner. The claim is in­de­pend­ent of any one tool. See the land­scape be­fore it bills you.

The Voice of Volge desk

About Flexera

Flex­era is a tech­no­logy-spend man­age­ment soft­ware com­pany, and this book reads it from the side it star­ted on. That side is the owned land­scape. Be­fore it gov­erned ren­ted cloud and SaaS, Flex­era gov­erned in­stalled soft­ware and the li­cence en­ti­tle­ments and audit ex­pos­ure they carry. It has since added vul­ner­ab­il­ity in­tel­li­gence and li­fe­cycle data, and sells the whole of it as Flex­era One, so that the land­scape can be coun­ted, se­cured and re­tired.

  • 2008. Foun­ded: spun out of Mac­ro­vi­sion as Acresso, re­named Flex­era Soft­ware in 2009; headquartered in Itasca, Illinois
  • 2002. Se­cunia Re­search: vul­ner­ab­il­ity in­tel­li­gence, val­id­at­ing dis­clos­ures since 2002
  • 5M+. Tech­no­pe­dia cata­logue: the nor­m­al­ised hard­ware & soft­ware product ref­er­ence, with li­fe­cycle data

The land­scape now in­cludes AI. Flex­era's 2026 State of ITAM Re­port (24 June 2026) finds only 31 per cent of or­gan­isa­tions with ac­cur­ate vis­ib­il­ity into AI soft­ware, the new­est asset class

to enter the owned land­scape, and the least gov­erned. Flex­era launched AI Cost Man­age­ment against that gap on 10 June 2026.

Exhibit A Counted, then secured, then retired: the owned-landscape build-out

  • 2008. Soft­ware li­cence op­tim­isa­tion
  • Se­cunia. Vul­ner­ab­il­ity in­tel­li­gence (est. 2002)
  • Tech­no­pe­dia. Nor­m­al­ised cata­logue + li­fe­cycle
  • 2024–26 · Snow, Spot, ProsperOps, Chaos Genius. ITAM, SaaS and FinOps depth
  • 2026 · AI Cost Man­age­ment. AI in the land­scape (10 Jun 2026)

2026

Source: Flex­era cor­por­ate and product ma­ter­i­als, 2002–2026. Snow Soft­ware com­pleted Feb­ru­ary 2024; ProsperOps and Chaos Genius an­nounced 6 Janu­ary 2026. Mike Jerich be­came pres­id­ent and chief ex­ec­ut­ive on 16 July 2026, Jim Ryan mov­ing to vice chair­man. Thoma Bravo holds the ma­jor­ity.

Flexera One today: the owned landscape

SeeData Platform · IT Visibility · Technopedia · hardware & software inventory
ComplyIT Asset Management · Effective Licence Position · audit response
Secure & retireSoftware Vulnerability Manager (Secunia) · lifecycle / end-of-life data
  • 48%. Audited in the past year (Flex­era State of ITAM, Jun 2026)
  • 44%. Spent >$1M on soft­ware audits in three years (Flex­era, Jun 2026)
  • 31%. Have ac­cur­ate vis­ib­il­ity into AI soft­ware (Flex­era, Jun 2026)

Source: Flex­era 2026 State of ITAM Re­port, 24 June 2026 (n=512). Fig­ures de­scribe the land­scape, not Flex­era out­comes.

What this means for a buyer in 2026

Flex­era's old­est strength is the owned land­scape, which is count­ing soft­ware and de­fend­ing audits. Its new­est chal­lenge is the same land­scape's new­est res­id­ent, AI, which al­most no one can yet see. The pitch is a single view across both. The test is whether it reaches the as­sets you can­not cur­rently find.

Volge: the work­ing premise of this book

Technology Asset ManagementVoice of Volge02 •

You are liable for everything you have ever installed.

Not just what you use, what you deployed, forgot, and left running. To an auditor, an attacker, and the end of its support life, it is all still yours.

The rented landscape bills you monthly; the owned landscape bills you all at once

The com­pan­ion to this book, The tech­no­logy tax, was about the tech­no­logy you rent, mean­ing the meters that run every second and ar­rive as a monthly bill. This book is about the tech­no­logy you own, and it bills dif­fer­ently. The owned land­scape does not send an in­voice. It sends an aud­itor's let­ter, a breach no­ti­fic­a­tion, or a sup­port-ex­piry no­tice, and each ar­rives without warn­ing and all at once. Rent­ing is a cash-flow prob­lem. Own­ing is a li­ab­il­ity prob­lem.

Everything an or­gan­isa­tion has ever in­stalled and not de­lib­er­ately re­moved is still, in the eyes of the parties that mat­ter, its re­spons­ib­il­ity. The soft­ware vendor counts it and can audit the count. The at­tacker scans it and can ex­ploit the un­patched. The sup­port cal­en­dar ex­pires under it and turns a work­ing sys­tem into an un­sup­por­ted one. None of these de­pends on whether any­one is still using the thing. Li­ab­il­ity at­taches to in­stall­a­tion, not to use.

This is why the owned land­scape is dan­ger­ous in a way the ren­ted one is not. A meter you ig­nore merely over­spends. An asset you ig­nore ac­crues ex­pos­ure si­lently until the day it is called, in the audit true-up you did not budget, the vul­ner­ab­il­ity you did not patch, the end-of-life sys­tem that was quietly run­ning your payroll. The cost is in­vis­ible right up to the mo­ment it is total, and by then the choices that would have pre­ven­ted it are years in the past.

The 2026 fig­ures show the li­ab­il­ity grow­ing, not shrink­ing. Flex­era's State of ITAM Re­port finds 48 per cent of or­gan­isa­tions audited in the past year and 44 per cent spend­ing more than a mil­lion dol­lars on soft­ware audits over three years, with Mi­crosoft audits re­por­ted by 64 per cent of re­spond­ents and Or­acle audits climb­ing to 38 per cent (Flex­era, 24 June 2026). Com­plete IT asset vis­ib­il­ity, mean­while, has fallen to 36 per cent, down from 43 per cent in 2025. The ex­pos­ure is rising as the abil­ity to see it falls.

And a new asset class has entered the land­scape. The AI an or­gan­isa­tion de­ploys is an asset, not merely a bill to be paid. The mod­els and agents run­ning in­side its walls have to be in­vent­or­ied and se­cured, and even­tu­ally re­tired. Yet only 31 per cent of or­gan­isa­tions have ac­cur­ate vis­ib­il­ity into their AI soft­ware, and 59 per cent of re­spond­ents re­port that wasted AI spend has risen year on year (Flex­era, 24 June 2026). The

new­est res­id­ent of the land­scape is the one al­most no one can find.

Flex­era's an­swer, sold as Flex­era One, is to make the owned land­scape vis­ible and then dis­charge its li­ab­il­it­ies in turn. IT Asset Man­age­ment de­fends the audit. Soft­ware Vul­ner­ab­il­ity Man­ager, backed by Se­cunia Re­search, se­cures the un­patched. Tech­no­pe­dia's li­fe­cycle data flags what must be re­tired. Volge is a Flex­era de­liv­ery part­ner and be­ne­fits if a cli­ent buys, and that in­terest is marked wherever it bears on the ar­gu­ment.

Read on for the three li­ab­il­it­ies in turn. Com­ply first, then se­cure and re­tire. Then read the op­er­ator's view of how to carry the owned land­scape as a man­aged risk rather than an un­budgeted sur­prise.

A meter you ignore overspends. An asset you ignore waits, for the auditor, the attacker, or the expiry date.

Comply.

The audit is the owned landscape's first bill, and it arrives because what you are entitled to and what you have deployed are rarely the same number. Software Asset Management is how you know your position before the vendor tells you theirs.

The audit exists because you do not know what you have deployed

A soft­ware audit is a bet the vendor makes that you have lost count, and it is usu­ally a good bet. Between what an or­gan­isa­tion is en­titled to run and what it has ac­tu­ally in­stalled lies a gap that no one is paid to close, serv­ers re-im­aged, li­cences double-as­signed, edi­tions up­graded without a paper trail, vir­tual ma­chines cloned with their soft­ware aboard. The vendor knows this gap ex­ists. The audit is simply the mech­an­ism for pri­cing it, at the mo­ment least con­veni­ent to you and most con­veni­ent to them.

En­ti­tle­ment and de­ploy­ment drift apart for struc­tural reas­ons, not neg­li­gent ones. En­ter­prise li­cens­ing is baff­lingly com­plex. It is priced per core or per user or per server, with vir­tu­al­isa­tion rules that can mul­tiply a single li­cence re­quire­ment across an en­tire cluster. A well-mean­ing en­gin­eer who moves a work­load to a big­ger host can, under some met­rics, quad­ruple the li­cence li­ab­il­ity without touch­ing the soft­ware. The rules are de­signed to be hard, and the dif­fi­culty is not in­cid­ental to the vendor's rev­enue; it is the source of it.

That is the whole game. Who­ever has the bet­ter count sets the terms. An or­gan­isa­tion that can­not state its own ef­fect­ive

Technology Asset ManagementVoice of Volge03 •

li­cence po­s­i­tion ne­go­ti­ates blind and pays the true-up the vendor com­puted. An or­gan­isa­tion that can state its po­s­i­tion, with nor­m­al­ised and de­fens­ible de­ploy­ment data, ne­go­ti­ates from evid­ence and con­tests the in­flated find­ing, and it fre­quently dis­cov­ers it was over-li­censed rather than under. The count is the ad­vant­age, and most or­gan­isa­tions do not have it.

The audit is there­fore best un­der­stood as a sales chan­nel rather than a com­pli­ance ex­er­cise. Vendors under pres­sure to grow rev­enue from a ma­ture product turn to their in­stalled base, and the audit con­verts a cus­tomer's poor re­cord-keep­ing into an un­planned pur­chase, the "true-up", often bundled with a con­veni­ently-timed offer to mi­grate to the cloud edi­tion. The find­ing is real; the li­cence gap usu­ally does exist. But the tim­ing, the fram­ing, and the price are the vendor's to choose un­less the cus­tomer ar­rives with its own num­bers.

Exhibit 1 The three liabilities of the owned landscape

LiabilityCalled byArrives asFlexera instrument
ComplyThe software vendorAn audit and a true-up billIT Asset Management
SecureAn attackerAn exploited unpatched flawSoftware Vulnerability Manager
RetireThe support calendarAn unsupported, un-patchable systemTechnopedia lifecycle data

Source: Volge ana­lysis, 2026. Flex­era product map­ping per Flex­era ma­ter­i­als, 2026.

Nearly half of organisations were audited last year, and the audit bill is not coming down

The soft­ware audit is not a tail risk; it is a near-cer­tainty on a long enough ho­ri­zon. Flex­era's State of ITAM Re­port finds 48 per cent of or­gan­isa­tions audited in the past year alone, and 44 per cent hav­ing spent more than a mil­lion US dol­lars re­spond­ing to soft­ware audits over three years, a share Flex­era notes has stayed flat across its past three re­ports (Flex­era, 24 June 2026). ITAM teams spend the largest share of their time on soft­ware op­tim­isa­tion, at 32 per cent, fol­lowed by audit re­sponse and audit al­loc­a­tion at 22 per cent each, which is two fifths of a scarce func­tion spent on audits.

Exhibit 2 Who is auditing: publisher audits over three years, and any audit in the past year

  • Mi­crosoft. 64%, three years
  • Or­acle. 38%, three years
  • Adobe. 32%, three years
  • Audited by any­one. 48%, past year

Or­acle audits rose to 38% (from 24%) and Adobe to 32% (from 24%), both on the three-year win­dow. The 48% is any audit in the past year. Sep­ar­ately, 44% of re­spond­ents spent over $1m re­spond­ing across three years.

Source: Flex­era 2026 State of ITAM Re­port, 24 June 2026 (n=512).

The con­cen­tra­tion is in­struct­ive. Mi­crosoft, Or­acle, and Adobe, the vendors with the most com­plex met­rics and the largest in­stalled bases, audit most, be­cause that is where the en­ti­tle­ment gap is widest and the true-up largest. These are not ob­scure sup­pli­ers; they are the soft­ware every en­ter­prise already runs, which is pre­cisely why the ex­pos­ure is uni­ver­sal rather than niche. If you run their products, you are on the list; the only ques­tion is tim­ing.

The rising Or­acle and Adobe fig­ures point to a broader pat­tern. As li­cens­ing mod­els shift to­wards the cloud and to­wards con­sump­tion, vendors use the audit to ac­cel­er­ate the mi­gra­tion. A find­ing of non-com­pli­ance on the old model be­comes the lever for a move to the new one. Read­ing the audit as a tech­nical com­pli­ance check, rather than as the open­ing move in a com­mer­cial ne­go­ti­ation, is the mis­take that makes it ex­pens­ive.

  • 48%. audited in the past year. Flex­era, 24 Jun 2026
  • 44%. spent >$1M on audits over three years. Flex­era, 24 Jun 2026
  • 22%. of ITAM team time on audit re­sponse; soft­ware op­tim­isa­tion leads at 32%. Flex­era, 24 Jun 2026

IT Asset Management gives you the count before the vendor sets the price

IT Asset Man­age­ment, Flex­era's product for the dis­cip­line it has prac­tised longest and deepened by the 2024 Snow ac­quis­i­tion, ex­ists to pro­duce one arte­fact the audit turns on, which is the ef­fect­ive li­cence po­s­i­tion, a de­fens­ible re­con­cili­ation of what you are en­titled to against what you have ac­tu­ally de­ployed. Pro­duce it be­fore the vendor does, keep it cur­rent, and the audit changes char­ac­ter from an am­bush into a ne­go­ti­ation between two parties who both have num­bers.

The work starts with dis­cov­ery, be­cause you can­not re­con­cile what you can­not see. Flex­era's in­vent­ory, Data Plat­form and IT Vis­ib­il­ity cap­ab­il­it­ies, nor­m­al­ised against the Tech­no­pe­dia cata­logue of more than five mil­lion hard­ware and soft­ware products, turn the raw and in­con­sist­ent out­put of an en­ter­prise land­scape, the same product named five ways across four tools, into a single, clean count. This nor­m­al­isa­tion is the un­glam­or­ous found­a­tion of everything that fol­lows; without it, the li­cence po­s­i­tion is built on sand.

Technology Asset ManagementVoice of Volge04 •

The re­con­cili­ation then maps that de­ploy­ment against en­ti­tle­ment, mean­ing the con­tracts, the pur­chase re­cords and the li­cens­ing met­rics. That com­par­ison com­putes product by product where the land­scape is over-li­censed, where it is under-li­censed and where the po­s­i­tion is cor­rect. The under-li­censed po­s­i­tions are the ex­pos­ure to close be­fore an aud­itor finds them. The over-li­censed po­s­i­tions, which are com­mon and which no vendor vo­lun­teers, are money to re­claim at the next re­newal. Both are in­vis­ible without the re­con­cili­ation, and both are worth real money.

An audit-ready pos­ture is the stand­ing res­ult. It means a li­cence po­s­i­tion cur­rent enough that a vendor's let­ter prompts a re­view rather than a panic. When the find­ing ar­rives, the or­gan­isa­tion checks it against its own num­bers, con­tests what is in­flated, and settles what is real, often for a frac­tion of the open­ing fig­ure. The sav­ing is not a dis­count. It is the dif­fer­ence between ne­go­ti­at­ing from evid­ence and ne­go­ti­at­ing from fear.

We will state the limit, as we do through­out. SAM pro­duces the po­s­i­tion; it does not ne­go­ti­ate the con­tract. The re­con­cili­ation gives you the evid­ence, but a li­cens­ing ne­go­ti­ation with Or­acle or Mi­crosoft is a spe­cial­ist com­mer­cial ex­er­cise in­volving judge­ment, tim­ing, and nerve that no tool sup­plies. That com­bin­a­tion, the de­fens­ible count and the ne­go­ti­ation around it, is where an un­aided soft­ware pur­chase tends to stall.

Exhibit 3 The effective licence position, three outcomes, product by product

PositionWhat it meansThe action
Under-licensedDeployed more than entitledClose before the auditor finds it
Correctly licensedDeployment matches entitlementMaintain the evidence
Over-licensedEntitled to more than deployedReclaim at renewal: money the vendor will not volunteer

Source: Volge ana­lysis, 2026. Cap­ab­il­ity map­ping per Flex­era IT Asset Man­age­ment ma­ter­i­als, 2024–26.

The publishers who set the audit weather

Audit ex­pos­ure is not evenly dis­trib­uted across the land­scape. Four pub­lish­ers, Mi­crosoft, Or­acle, IBM, and SAP, lead the en­ter­prise audit and true-up risk, and each has built a li­cens­ing mech­an­ism that prices the gap between what you meas­ured and what you run. Know the mech­an­isms and you know where a SAM pro­gramme earns its keep first.

Or­acle moved Java SE to the Uni­ver­sal Sub­scrip­tion on 23 Janu­ary 2023, and with it to an em­ployee-count met­ric. The sub­scrip­tion is priced against the whole work­force, and

Or­acle's price list ex­tends Em­ployee to the staff of agents, con­tract­ors, out­sourcers and con­sult­ants sup­port­ing in­ternal op­er­a­tions. A single JDK in­stall on one server can put that base in scope. Or­acle re­placed the sub-3,000 bands with flat small-busi­ness pri­cing on 8 Oc­to­ber 2026, which lowers the entry cost and leaves the met­ric in­tact. Java dis­cov­ery is now the first ques­tion in an Or­acle con­ver­sa­tion, down to every in­stall, edi­tion and ver­sion. Data­base op­tions, en­abled with a com­mand and li­censed sep­ar­ately, re­main the older trap.

SAP began char­ging for in­dir­ect use by the doc­u­ment on 10 April 2018, launch­ing it as ERP pri­cing for the di­gital age; the name Di­gital Ac­cess, and the Di­gital Ac­cess Ad­op­tion Pro­gram, came af­ter­wards. Or­ders, in­voices and other core doc­u­ments cre­ated in SAP by non-SAP sys­tems carry the li­cence, coun­ted on cre­ation, with ma­ter­ial and fin­an­cial doc­u­ments weighted at 0.2. Every in­teg­ra­tion that writes into SAP, a web shop, a CRM, now an AI agent, is a po­ten­tially metered actor. The ex­pos­ure hides in the ar­chi­tec­ture dia­gram, not the user list.

IBM per­mits the cheaper sub-ca­pa­city count­ing, the vir­tual ma­chine rather than the whole cluster, on one con­di­tion. IBM re­quires its Li­cense Met­ric Tool to be de­ployed, col­lect­ing and re­port­ing, and the only sub­sti­tutes it names are HCL BigFix In­vent­ory and Flex­era One with IBM Ob­serv­ab­il­ity IT Asset Man­age­ment. Ab­sent an ap­proved tool, the de­fault is full ca­pa­city, and on a large vir­tu­al­ised land­scape the dif­fer­ence is a mul­tiple, not a mar­gin. IBM is also fold­ing Pass­port Ad­vant­age into its Cli­ent Re­la­tion­ship Agree­ment, so the terms to read are chan­ging.

Mi­crosoft's mech­an­ism is quieter. It is drift rather than the audit ham­mer. Mi­crosoft an­nounced a price and pack­aging up­date on 4 Decem­ber 2025, ef­fect­ive 1 July 2026, lift­ing Mi­crosoft 365 E3 from $36 to $39 and E5 from $57 to $60, and mov­ing point-tool cap­ab­il­ity into both suites. E3 gains De­fender for Of­fice Plan 1 and three In­tune cap­ab­il­it­ies; E5 adds Se­cur­ity Co­pi­lot, En­d­point Priv­ilege Man­age­ment, Cloud PKI and En­ter­prise Ap­plic­a­tion Man­age­ment. Cap­ab­il­ity bought sep­ar­ately may there­fore already sit in the suite, at either tier. Azure Hy­brid Be­ne­fit claims rest­ing on un­veri­fied li­cence po­s­i­tions are un­changed in kind. The ex­pos­ure ar­rives as a re­newal quote, which makes it easier to miss and no cheaper.

Technology Asset ManagementVoice of Volge05 •

The list is open. Broad­com's VMware ac­quis­i­tion closed on 22 Novem­ber 2023, and on 11 Decem­ber 2023, nine­teen days later, the port­fo­lio went sub­scrip­tion-only and per-core: a be­nign vendor turned audit-weather pub­lisher by a change of own­er­ship. Sales­force raised list prices by an av­er­age of 6 per cent from 1 Au­gust 2025, an­nounced on 17 June 2025, and re­placed its Ein­stein add-ons and Ein­stein 1 edi­tions with Agent­force, which is the audit-rel­ev­ant re­tire­ment. Adobe re­pack­aged Cre­at­ive Cloud All Apps as Cre­at­ive Cloud Pro on 15 May 2025, with gen­er­at­ive cred­its al­loc­ated per li­cence and no roll-over. The de­fence has to be the land­scape, not the pub­lisher.

Exhibit 4 The four mechanisms, and what each one prices

PublisherThe mechanismWhat it pricesSince
OracleJava SE Universal Subscription: per-employee metricEvery employee, not every user of Java23 January 2023
SAPDocument-based indirect use, later named Digital AccessDocuments created in SAP by non-SAP systems10 April 2018
IBMSub-capacity PVU, conditional on ILMT reportingThe full cluster, wherever ILMT is absentClient Relationship Agreement terms
MicrosoftTier mix, CSP billing, Azure Hybrid BenefitThe gap between tiers bought and features usedContinuous; repackaged 1 July 2026

Source: pub­lisher li­cens­ing mod­els as pub­licly dated (Or­acle, 23 Jan 2023; SAP, 10 Apr 2018; IBM sub-ca­pa­city terms, mov­ing to the Cli­ent Re­la­tion­ship Agree­ment; Broad­com-VMware sub­scrip­tion shift, 11 Dec 2023). The four are lead­ing ex­amples, not an ex­haust­ive list; char­ac­ter­isa­tion and se­lec­tion are Volge ana­lysis, 2026.

What to do in the thirty days after the letter arrives

An audit let­ter is a pro­ced­ural doc­u­ment, and the first thirty days de­cide the out­come more than the next six months do. The let­ter will cite a clause, name a scope and pro­pose a method. All three are ne­go­ti­able, and all three are easier to ne­go­ti­ate be­fore any data has been handed over.

Start with the clause. Audit rights are con­trac­tual, and they are nar­rower than the let­ter usu­ally im­plies. Read the ac­tual pro­vi­sion, es­tab­lish which legal en­tit­ies and which products it reaches, and es­tab­lish the no­tice period it re­quires. A let­ter ad­dressed to a group but foun­ded on a sub­si­di­ary's agree­ment reaches the sub­si­di­ary.

Then settle the method in writ­ing. Who runs the scripts, what they col­lect, where the out­put goes and who sees it are all

terms rather than facts. A vendor-run script on a pro­duc­tion land­scape col­lects more than the audit needs, and the sur­plus be­comes the evid­ence for the next find­ing. Sup­ply­ing the or­gan­isa­tion's own nor­m­al­ised ex­tract, in an agreed format, is both de­fens­ible and cheaper.

Only then pro­duce the po­s­i­tion. The order mat­ters, be­cause a po­s­i­tion dis­closed be­fore the scope is fixed sets the scope.

Exhibit The audit response, by week

WeekThe workWhat must exist at the end of it
1Read the audit clause; confirm entity and product scope; acknowledge receipt without conceding methodA written statement of scope, sent, with the clause cited
2Freeze deployment change in scope; assemble the contract archive for the named productsAn entitlement file with a document behind every line
3Produce the internal position for the named products, identifying over-licensed as well as under-licensed holdingsA position the organisation would defend in writing
4Agree the data collection method and the confidentiality terms; exchange the first data setA signed scope and method, and a dated record of what was supplied

Source: Volge ana­lysis, 2026. The four-week shape is a work­ing pat­tern rather than a con­trac­tual timetable; no­tice peri­ods vary by agree­ment.

Two points in that se­quence carry most of the money. The de­ploy­ment freeze in week two pre­vents an en­gin­eer from cre­at­ing new ex­pos­ure dur­ing the audit, which hap­pens more often than any­one ad­mits. And the over-li­censed hold­ings in week three are the asset the vendor will not raise. A find­ing of 400 under-li­censed cores sits very dif­fer­ently along­side 600 cores paid for and never de­ployed.

The limit of this book's claim should be stated. A li­cence po­s­i­tion is evid­ence. A ne­go­ti­ation with a pub­lisher is a com­mer­cial ex­er­cise in­volving tim­ing and lever­age, and no plat­form sup­plies either. What a plat­form sup­plies is the abil­ity to ar­rive with num­bers, which is the dif­fer­ence between con­test­ing a find­ing and ac­cept­ing one.

The scale of the prob­lem is set by two num­bers mov­ing in op­pos­ite dir­ec­tions. The count of dis­closed vul­ner­ab­il­it­ies rises every year into the tens of thou­sands, spread across thou­sands of products. The ca­pa­city of a se­cur­ity team to test, sched­ule, and de­ploy patches does not rise to match. The res­ult is a per­man­ent triage prob­lem: far more known vul­ner­ab­il­it­ies than any team can re­medi­ate, and no ob­vi­ous way to tell the hand­ful that will be ex­ploited from the many that will not.

Re­spond­ing to volume alone makes the prob­lem worse. A team that treats every dis­closed vul­ner­ab­il­ity as equally

Technology Asset ManagementVoice of Volge06 •

ur­gent spreads its fi­nite patch­ing ca­pa­city evenly across a threat land­scape that is any­thing but even, and ex­hausts it­self re­medi­at­ing low-risk flaws while a genu­inely dan­ger­ous one waits in the queue. The met­ric that mat­ters is not how many vul­ner­ab­il­it­ies exist but which of them are present in your land­scape, reach­able, and ac­tu­ally being ex­ploited: a far smal­ler and more ac­tion­able set.

Secure.

The landscape's attack surface is its unpatched software, most of it third-party, most of it invisible to the tools built for Microsoft alone. Software Vulnerability Manager, backed by Secunia Research, tells you what to patch first.

You cannot patch what you cannot see, and most of what you cannot see is not Microsoft

The un­patched soft­ware in a land­scape is its stand­ing in­vit­a­tion to an at­tacker, and the dan­ger­ous part of it is rarely the op­er­at­ing sys­tem. Win­dows patches it­self on a sched­ule most or­gan­isa­tions have auto­mated. The ex­pos­ure lives in the third-party soft­ware layered on top. Browsers and runtimes, PDF read­ers and media lib­rar­ies, and hun­dreds of niche ap­plic­a­tions each carry their own vul­ner­ab­il­it­ies, and none of them is touched by the op­er­at­ing-sys­tem patch cycle. It is a large sur­face, and most or­gan­isa­tions can­not see the whole of it.

That re­fram­ing re­quires two things most or­gan­isa­tions lack: an ac­cur­ate in­vent­ory of what third-party soft­ware is ac­tu­ally in­stalled, and re­li­able in­tel­li­gence about which vul­ner­ab­il­it­ies in that soft­ware are real and ser­i­ous rather than noise. The first is the dis­cov­ery layer from Part I, re­used. The second is a re­search cap­ab­il­ity. A raw feed lists every dis­clos­ure; a val­id­ated, pri­or­it­ised ad­vis­ory tells you where to spend the day's patch­ing ca­pa­city.

The triage arithmetic

If tens of thou­sands of vul­ner­ab­il­it­ies are dis­closed a year and a team can mean­ing­fully patch a frac­tion of them, the en­tire value is in choos­ing the right frac­tion. Patch­ing more is not the goal. Patch­ing the ones that would ac­tu­ally have been ex­ploited, first, is.

Secunia Research turns a feed of every flaw into a list of the ones that matter

Soft­ware Vul­ner­ab­il­ity Man­ager ex­ists to solve the triage prob­lem, and its dis­tin­guish­ing asset is Se­cunia Re­search, a team that has val­id­ated and pri­or­it­ised vul­ner­ab­il­ity dis­clos­ures since 2002. SVM dif­fers from a raw vul­ner­ab­il­ity feed as

in­tel­li­gence dif­fers from data. Se­cunia tests dis­clos­ures and re­jects the il­le­git­im­ate ones. It scores the real ones for crit­ic­al­ity and at­tack vec­tor and pairs them with re­medi­ation, so a se­cur­ity team re­ceives a ranked list rather than an un­dif­fer­en­ti­ated stream.

The re­search is de­lib­er­ate rather than auto­mated. Se­cunia's spe­cial­ists con­duct code audits and bin­ary ana­lysis to find is­sues that auto­mated fuzz­ing misses, and, just as im­port­ant, they re­ject il­le­git­im­ate or du­plic­ate re­ports so that teams are not made to chase phantoms. In a field drown­ing in dis­clos­ures, the act of au­thor­it­at­ively say­ing "this one is not real, and this one is" is it­self much of the value. Noise is not a neut­ral cost; it is the thing that hides the sig­nal a team needed to act on.

SVM then joins that in­tel­li­gence to the land­scape. By match­ing val­id­ated, scored ad­visor­ies against what is ac­tu­ally in­stalled, the nor­m­al­ised in­vent­ory from Part I, it pro­duces the only list that mat­ters op­er­a­tion­ally: the vul­ner­ab­il­it­ies that are both ser­i­ous and present in your en­vir­on­ment, ranked so the scarce patch­ing win­dow is spent on the flaws most likely to be ex­ploited. Flex­era pairs this with what it de­scribes as the largest re­pos­it­ory of third-party patch data in the in­dustry, clos­ing the loop from "what is dan­ger­ous" to "here is the patch."

The hon­est limit is the same one every se­cur­ity team knows. Identi­fy­ing and pri­or­it­ising a vul­ner­ab­il­ity is not the same as re­medi­at­ing it; the patch still has to be tested against a live en­vir­on­ment and de­ployed in­side a change win­dow con­trolled by op­er­a­tions rather than by the tool. SVM makes the triage de­cision de­fens­ible and the patch avail­able. It does not ab­sorb the op­er­a­tional risk of de­ploy­ment, and it does not cover the vul­ner­ab­il­it­ies in soft­ware you wrote your­self. It sharpens the judge­ment and shortens the list. Someone still has to do the patch­ing.

Exhibit 5 From every disclosed flaw to the day's patch list

  • All dis­clos­ures. Tens of thou­sands a year, mostly noise
  • Val­id­ated. Se­cunia re­jects the il­le­git­im­ate
  • In your land­scape. Matched to in­stalled in­vent­ory
  • Patch first. Crit­ical, present, ex­ploit­able

Source: Volge ana­lysis, 2026. Cap­ab­il­ity per Flex­era Soft­ware Vul­ner­ab­il­ity Man­ager / Se­cunia Re­search ma­ter­i­als.

The patch window is the constraint, not the advisory

A vul­ner­ab­il­ity pro­gramme is bounded by change win­dows rather than by in­tel­li­gence. The ad­vis­ory ar­rives in minutes.

Technology Asset ManagementVoice of Volge07 •

The patch reaches a pro­duc­tion host when op­er­a­tions per­mits a re­start, and on the sys­tems that mat­ter most that per­mis­sion is scarce. Any triage scheme that ig­nores this is sort­ing a queue nobody can drain.

So the use­ful unit of plan­ning is the win­dow and not the flaw. Count the de­ploy­able win­dows in a month, by sys­tem class, and that num­ber is the pro­gramme's ca­pa­city. Everything else is al­loc­a­tion. A land­scape with four main­ten­ance win­dows a month on its tier-one sys­tems can close four tier-one ex­pos­ures a month, how­ever many ad­visor­ies ar­rive.

Two con­sequences fol­low, and both are un­com­fort­able. First, a back­log is not a fail­ure of di­li­gence; it is arith­metic, and the only hon­est re­sponse is to say which flaws are de­lib­er­ately being car­ried and why. Second, com­pens­at­ing con­trols stop being an ex­cuse and be­come a line item. Net­work isol­a­tion, a vir­tual patch at the gate­way or a with­drawal of the af­fected cap­ab­il­ity are the avail­able moves when the win­dow will not come in time, and each needs an owner and an ex­piry date.

Measure the exploited subset separately

Re­medi­ation stat­ist­ics flat­ter a pro­gramme that patches the easy things. The num­ber that mat­ters is the me­dian time from a val­id­ated ad­vis­ory with known ex­ploit­a­tion to a patched host, re­por­ted apart from everything else. Re­por­ted to­gether, the two meas­ures hide each other, be­cause the volume of low-risk flaws will al­ways dom­in­ate the av­er­age.

A second fig­ure is worth stand­ing be­side it. The share of ad­visor­ies that re­solve to a named host in­side an hour is a meas­ure of the in­vent­ory rather than of the se­cur­ity team. A triage pro­cess that can­not an­swer the ques­tion "where is this in­stalled" is not triaging. It is guess­ing with bet­ter vocab­u­lary.

The triage decision record

Every car­ried vul­ner­ab­il­ity should leave a re­cord, and the re­cord is short. Four fields are enough for an ex­am­iner, an in­surer or a post-in­cid­ent re­view to fol­low the reas­on­ing without an in­ter­view.

  • The ad­vis­ory, with its iden­ti­fier and the date the val­id­ated ad­vis­ory was is­sued.
  • The af­fected hosts, by name, and whether each is in­ter­net-reach­able.
  • The de­cision taken, which is patch now, patch in a named win­dow, com­pensate, or ac­cept.
  • The owner and the re­view date, be­cause an ac­cept­ance with no ex­piry be­comes a per­man­ent po­s­i­tion by de­fault.

Source: Volge ana­lysis, 2026. The four fields are the min­imum, not a tem­plate to be elab­or­ated.

The value of the re­cord is not the pa­per­work. It is that a car­ried risk with a name against it gets re­vis­ited, and a car­ried risk re­cor­ded only in a spread­sheet column does not.

The patch supply chain is itself a dependency

A third-party patch pro­gramme de­pends on pack­ages the or­gan­isa­tion does not build, and that de­pend­ency is rarely re­gistered as a risk. Someone has to re­pack­age each vendor's in­staller, test it, and pub­lish it in a de­ploy­able form. Where that work is bought in, the sup­ply of patches has a vendor, a latency and a cov­er­age bound­ary, and all three be­long in the risk re­gister.

Cov­er­age is the sharpest of the three. A patch cata­logue cov­ers a list of ap­plic­a­tions, and the ap­plic­a­tions out­side that list are the ones a se­cur­ity team must handle by hand. Ask for the list, com­pare it against the nor­m­al­ised in­vent­ory, and the res­ult is a count of ap­plic­a­tions for which there is no sup­por­ted patch route at all. That count is the real back­log.

Latency is the second. The gap between a vendor pub­lish­ing a fix and a re­pack­aged, tested ver­sion be­com­ing avail­able is the period in which the or­gan­isa­tion is ex­posed while be­liev­ing it is covered. It is meas­ur­able. Meas­ure it for the ten ap­plic­a­tions that mat­ter most, rather than for the cata­logue as a whole.

Retire.

The support calendar expires under the landscape whether or not anyone is watching, turning working systems into un-patchable ones. And a new asset class, the AI you deploy, has entered the landscape almost entirely ungoverned.

End-of-life is the quietest liability: nothing breaks, until everything is exposed

Ob­sol­es­cence is the li­ab­il­ity that an­nounces it­self last. When a piece of soft­ware reaches end-of-life, noth­ing vis­ible hap­pens, it keeps run­ning, keeps serving users, keeps look­ing like an asset. What has ac­tu­ally happened is that the vendor has stopped ship­ping se­cur­ity patches for it, so every vul­ner­ab­il­ity dis­closed from that day for­ward is one you can never close. The sys­tem has quietly con­ver­ted from a work­ing asset into a per­man­ent, un-patch­able ex­pos­ure, and it will go on doing its job right up to the mo­ment it be­comes the way in.

The reason ob­sol­es­cence ac­cu­mu­lates is that re­tir­ing soft­ware is nobody's ur­gent pri­or­ity. Re­pla­cing a work­ing sys­tem costs money and ef­fort now to avoid a risk that is dif­fuse and later, which is ex­actly the trade or­gan­isa­tions are worst at mak­ing. So the end-of-life data­base server keeps run­ning the

Technology Asset ManagementVoice of Volge08 •

re­port it has al­ways run, the un­sup­por­ted op­er­at­ing sys­tem stays on the ma­chine in the corner, and the risk com­pounds si­lently on a bal­ance sheet no one reads until an aud­itor or an at­tacker reads it for them.

See­ing ob­sol­es­cence be­fore it bites re­quires know­ing two things the land­scape does not nat­ively re­cord: pre­cisely what ver­sions of what products are in­stalled, and what the sup­port li­fe­cycle of each of those ver­sions is. The first is the dis­cov­ery layer again. The second is li­fe­cycle in­tel­li­gence. That is a main­tained re­cord of when every product ver­sion reaches end of sup­port and end of life. Neither an in­vent­ory tool nor a vendor's web­site gives you both in a form you can act on across a land­scape of thou­sands of products.

This is where a nor­m­al­ised cata­logue with li­fe­cycle data earns its place. Tech­no­pe­dia's five-mil­lion-plus products carry iden­tity and li­fe­cycle status, so the land­scape can be read for­ward in time: what is un­sup­por­ted today, what ex­pires this year, what to budget to re­place be­fore it be­comes an ex­pos­ure. Ob­sol­es­cence then stops being a sur­prise dis­covered dur­ing an in­cid­ent and be­comes a sched­ule man­aged in ad­vance. That is the only af­ford­able way to man­age it.

Exhibit 6 The lifecycle states, and the liability of each

StateWhat it meansThe liability
SupportedPatches still shippingNormal: keep current
End of support approachingExpiry dated and knownBudget the replacement now
End of lifeNo further security patchesEvery new flaw is permanent: retire or isolate

Source: Volge ana­lysis, 2026. Li­fe­cycle data per Flex­era Tech­no­pe­dia.

The AI you deploy is an asset in the landscape, and almost no one can find it

The com­pan­ion book gov­erned the AI you rent, the token bill that ar­rives each month. This book's con­cern is the AI you own. That means the mod­els, the agents and the AI-en­abled fea­tures now in­stalled and run­ning in­side the land­scape, all of them as­sets to be in­vent­or­ied, se­cured and even­tu­ally re­tired ex­actly like any other soft­ware. The dif­fer­ence is that this asset class ar­rived faster than any be­fore it, and the land­scape has no re­cord of most of it. It is the ob­sol­es­cence and vul­ner­ab­il­ity prob­lem of this book, ap­plied to soft­ware that did not exist in the in­vent­ory two years ago.

The fig­ures de­scribe an asset class out­run­ning its gov­ernance. Only 31 per cent of or­gan­isa­tions have ac­cur­ate vis­ib­il­ity into their AI soft­ware, even as nearly half, in Flex­era's

own word­ing, have begun track­ing it as a line of soft­ware spend and 59 per cent of re­spond­ents re­port that wasted AI spend has risen year on year (Flex­era, 24 June 2026). A de­ployed model an or­gan­isa­tion can­not see is a model it can­not se­cure. Nor can it be li­censed cor­rectly or re­tired. Those are the three li­ab­il­it­ies of this en­tire book, re­appear­ing in their new­est and least un­der­stood form.

Exhibit 7 The AI-in-landscape gap: deployed faster than it is governed

  • Ac­cur­ate vis­ib­il­ity into AI soft­ware. 31%
  • Track­ing AI as part of soft­ware spend. Nearly half
  • Re­port­ing rising wasted AI spend. 59%

Only 31% have ac­cur­ate vis­ib­il­ity into AI soft­ware; com­plete IT asset vis­ib­il­ity over­all has fallen to 36%.

Source: Flex­era 2026 State of ITAM Re­port, 24 June 2026 (n=512).

Treat­ing de­ployed AI as an asset rather than only a cost changes the ques­tions. What mod­els are run­ning in our land­scape, and under what li­cence terms? Which of them carry known vul­ner­ab­il­it­ies or were trained on data we must ac­count for? Which are shadow de­ploy­ments a team stood up without re­cord? These are in­vent­ory-and-li­fe­cycle ques­tions, and they are answered with the same dis­cip­line used for any other soft­ware, provided the dis­cov­ery layer can find AI at all, which for most or­gan­isa­tions it can­not yet.

Retiring the landscape, and bringing AI into it, is one discipline, not two

The in­stru­ment for re­tire­ment is the same land­scape view used to com­ply and to se­cure, read for­ward in time. Tech­no­pe­dia's li­fe­cycle data lets an or­gan­isa­tion sched­ule ob­sol­es­cence rather than dis­cover it, and the same dis­cov­ery-and-nor­m­al­ise layer that finds in­stalled soft­ware is what must be ex­ten­ded to find in­stalled AI. The claim Flex­era makes for 2026 is that AI is not a sep­ar­ate gov­ernance prob­lem but the ex­ist­ing land­scape dis­cip­line ap­plied to a new asset class, provided the dis­cov­ery reaches it.

For ob­sol­es­cence, the op­er­at­ing pat­tern is a rolling one. Read the land­scape against li­fe­cycle data, identify what ex­pires in the plan­ning ho­ri­zon, and budget the re­place­ments be­fore end-of-life turns them into per­man­ent ex­pos­ures. This con­verts a re­act­ive scramble, dis­cov­er­ing an un­sup­por­ted sys­tem dur­ing an in­cid­ent, into a line in the tech­no­logy plan. It is un­glam­or­ous asset hy­giene, and it is far cheaper than the breach or the audit it fore­stalls.

Technology Asset ManagementVoice of Volge09 •

For AI, the dis­cip­line is clear and the tool­ing is new rather than ab­sent. Bring­ing de­ployed mod­els and agents into the in­vent­ory, track­ing their li­cence and li­fe­cycle status, and flag­ging shadow de­ploy­ments is the right frame, and Flex­era now sells a product against it. AI Cost Man­age­ment launched on 10 June 2026 and was ex­pan­ded on 20 Oc­to­ber 2026, cov­er­ing agents, mod­els, data and com­pute with token and credit track­ing. Flex­era's own cap­ab­il­ity mat­rix still marks AI policy and gov­ernance as par­tial, so whether that view reaches the full sprawl of AI an or­gan­isa­tion has de­ployed is a ques­tion a buyer should test against their own en­vir­on­ment, not take on faith. We re­com­mend the dis­cip­line con­fid­ently and the ma­tur­ity cau­tiously.

What this part established

End-of-life is the quietest li­ab­il­ity, be­cause noth­ing breaks until everything is ex­posed, and only li­fe­cycle data read for­ward in time pre­vents it. De­ployed AI is the new­est res­id­ent of the land­scape and the least vis­ible, 31 per cent see it ac­cur­ately, and it must be in­vent­or­ied, se­cured, and re­tired like any other asset. Part IV car­ries the three li­ab­il­it­ies into one op­er­ator's read.

The AI you rent shows up on a bill. The AI you own shows up in an audit, a breach, or an expiry notice, and most landscapes cannot find it yet.

Volge: the com­pan­ion premise to The tech­no­logy tax

What an AI asset register has to hold

Bring­ing de­ployed AI onto the re­gister is a data-mod­el­ling prob­lem be­fore it is a tool­ing prob­lem, and the fields are not the ones a soft­ware in­vent­ory already car­ries. A model is iden­ti­fied by its weights and its ver­sion, served through an en­d­point that may change be­neath it, under a li­cence whose terms on out­put and on train­ing data dif­fer sharply between sup­pli­ers. Six fields carry the re­gister.

Source: Volge ana­lysis, 2026. The field set is de­signed to be an­swer­able from ex­ist­ing tele­metry rather than from a ques­tion­naire.

The last field is where the bill-of-ma­ter­i­als work meets this book. A soft­ware bill of ma­ter­i­als lists the com­pon­ents of a build. An AI bill of ma­ter­i­als ex­tends the same idea to mod­els, data sets and the tool sur­face an agent can call. CERT-In's Tech­nical Guidelines on SBOM, QBOM, CBOM, AIBOM and HBOM, CISG-2024-02 ver­sion 2.0 of 3 Oc­to­ber 2024, are the only in­stru­ment in the jur­is­dic­tions this book sur­veys that

FieldWhy it is on the register
Model and versionThe thing that was actually invoked, which is often not the thing that was approved.
Endpoint and hosting locationDetermines the residency position and whether the call left the jurisdiction.
Licence basis and output termsWhether the output may be used commercially, and who owns it.
Named ownerAn agent with no owner is a shadow deployment with a budget code.
Lifecycle statusModels are deprecated and withdrawn on vendor timetables shorter than any software support cycle.
Dependency manifestThe libraries, tools and data sources the agent can reach, which is the attack surface.

names one, and they are guid­ance rather than a bind­ing dir­ec­tion.

The prac­tical test is blunt. Name the mod­els run­ning in the land­scape today, with an owner against each. An or­gan­isa­tion that can­not do that in an af­ter­noon is car­ry­ing an asset class it can­not se­cure, li­cense or re­tire, which is the ar­gu­ment of this book ar­riv­ing on new ground.

When replacement is unfunded, isolation is the decision to take

Most end-of-life sys­tems are re­tained be­cause re­place­ment is un­fun­ded, and that is a le­git­im­ate po­s­i­tion badly ex­ecuted. The error is not the re­ten­tion. It is re­tain­ing the sys­tem on the same net­work, with the same reach, as though noth­ing had changed on the day sup­port ended.

Three moves are avail­able short of re­place­ment, and each costs a frac­tion of a mi­gra­tion. The sys­tem can be isol­ated, so that it reaches only the hand­ful of ser­vices it genu­inely needs and noth­ing reaches it ex­cept through a con­trolled path. Its cap­ab­il­ity can be re­duced, by with­draw­ing the in­ter­faces and fea­tures that are not in use and that carry the ex­pos­ure. Or the vendor's ex­ten­ded sup­port can be bought where it ex­ists, which con­verts an un-patch­able sys­tem back into a patched one for a stated period and a stated price.

Each of the three is a de­cision with a cost and a re­view date, which is ex­actly what an un­fun­ded re­ten­tion lacks. An isol­ated end-of-life server with a named owner and an an­nual re­view is a man­aged risk. The same server left on the gen­eral net­work is an un­man­aged one, and the two look identical on an asset re­port.

Source: Volge ana­lysis, 2026. The op­tions are not ex­clus­ive; isol­a­tion and re­duced cap­ab­il­ity are com­monly taken to­gether while a re­place­ment is fun­ded.

Technology Asset ManagementVoice of Volge10 •

Exhibit The end-of-life options, and what each one actually buys

OptionWhat it buysWhat it does not
ReplaceSupported software and a current patch routeNothing, if the replacement inherits the same unmanaged position
Extended support, where offeredA patch route for a stated period at a stated priceA permanent answer; the expiry still arrives
IsolateA reduced blast radius for flaws that can never be closedProtection against a flaw reachable through the paths you left open
Reduce capabilityA smaller attack surface on the same softwareAnything, unless the withdrawn features were genuinely unused
AcceptTime, and a clear record of the choiceAny defence at all, if the acceptance is undated and unowned

Controls, sequencing and operations

A landscape programme is sequenced work rather than a purchase. Discovery comes before normalisation, normalisation before reconciliation, and no position downstream is defensible until the count is clean. Each control below carries a test, and the test is the thing an auditor asks to see.

Land­scape pro­grammes fail on pre­requis­ites far more often than on tool­ing. Three con­di­tions must hold be­fore dis­cov­ery be­gins. Someone must carry the au­thor­ity to run a cre­den­tialled scan or place an agent across every net­work seg­ment in scope, in­clud­ing the seg­ments that op­er­a­tional tech­no­logy owns. The con­tract archive must be as­sembled, be­cause en­ti­tle­ment can­not be re­con­struc­ted from memory. And each pub­lisher re­la­tion­ship needs a named owner, be­cause a re­con­cili­ation with nobody be­hind it pro­duces a re­port rather than a po­s­i­tion.

The third con­di­tion is the one most often skipped, and it is the cheapest to meet. A li­cence po­s­i­tion is an as­ser­tion about a con­tract. As­ser­tions need an au­thor.

What must be in place before discovery starts

  • Scan au­thor­ity. Writ­ten ap­proval to run agents or cre­den­tialled scans across every in-scope seg­ment, with the ex­clu­sions lis­ted and jus­ti­fied. Un­lis­ted ex­clu­sions be­come the gaps an aud­itor finds.
  • The con­tract archive. Pur­chase or­ders, amend­ments, en­ter­prise agree­ments and trans­fer re­cords, held in one place. En­ti­tle­ment that can­not be evid­enced is en­ti­tle­ment you do not have.
  • A pub­lisher owner. One named per­son per audit-prone pub­lisher, hold­ing the po­s­i­tion and the ne­go­ti­ation his­tory.
  • Vir­tu­al­isa­tion to­po­logy. Host, cluster and af­fin­ity data for every vir­tu­al­ised work­load, be­cause the per-core pub­lish­ers price the cluster rather than the guest.
  • A data-pro­tec­tion basis. In­vent­ory data car­ries host­names, user­names and in­stall­a­tion paths, so the law­ful basis for col­lect­ing and ex­port­ing it is settled be­fore the first sweep, not after.
  • A change win­dow. Agent de­ploy­ment touches en­d­points, and an asset pro­gramme that breaks a trad­ing desk loses its man­date in a week.

The sequence, and the exit test at each stage

The order of work is fixed by de­pend­ency rather than by pref­er­ence. Dis­cov­ery pro­duces a raw count. Nor­m­al­isa­tion turns that count into products with names a con­tract re­cog­nises. Re­con­cili­ation sets those products against en­ti­tle­ment. Re­medi­ation closes what is open, and gov­ernance keeps the po­s­i­tion cur­rent as the land­scape moves. Re­verse any two stages and the out­put stops being de­fens­ible, be­cause a re­con­cili­ation against an un­nor­m­al­ised count re­con­ciles noth­ing.

Exhibit The five stages, each with the test that ends it

StageWhat it producesThe exit test
01 DiscoverA raw inventory of hosts, devices, tenancies and installed softwareReconcile device count against the identity directory and the network inventory. A variance above 5 per cent means the sweep is incomplete, not that the directory is wrong.
02 NormaliseOne name per product, edition and version, matched to a catalogueSample 50 records from the three noisiest source tools and confirm each resolves to a single catalogue entry. Unmatched records are counted and owned, never discarded.
03 ReconcileAn effective licence position, product by productRe-derive the position for two publishers by hand from the contracts. If the hand calculation and the tool disagree, the metric has been configured wrongly.
04 RemediateClosed under-licensed exposure and reclaimed shelfwareEach closed position carries a dated evidence pack a third party could follow without asking a question.
05 GovernA standing position that survives changeRe-run the position after a quarter of ordinary change and measure the drift. Drift above a stated threshold triggers a review rather than an alert nobody reads.

Source: Volge ana­lysis, 2026. The vari­ance and sampling thresholds are Volge's own work­ing prac­tice and are offered as a start­ing point, not as an in­dustry stand­ard.

Technology Asset ManagementVoice of Volge11 •

Two of those tests de­serve em­phasis. The hand re-de­riv­a­tion at stage three is the single most use­ful hour in the whole pro­gramme, be­cause a li­cence en­gine con­figured against the wrong met­ric pro­duces a con­fid­ent num­ber that is wrong in the vendor's fa­vour or yours. Neither dir­ec­tion is safe. And the drift meas­ure­ment at stage five is what sep­ar­ates a pro­ject from a prac­tice. A po­s­i­tion that was ac­cur­ate in March and has not been re­freshed is a po­s­i­tion, and it is also a haz­ard, be­cause it will be quoted in a ne­go­ti­ation as though it were cur­rent.

The controls, and how each one is tested

Con­trols on the owned land­scape are or­din­ary and the tests are cheap, which is why the ab­sence of them is hard to de­fend. Each con­trol below an­swers a ques­tion an aud­itor, an ex­am­iner or a post-in­cid­ent re­view will ask, and each leaves evid­ence that can be pro­duced without a re­con­struc­tion ex­er­cise.

The shadow-de­ploy­ment test is the one teams res­ist, and it is the only one that meas­ures the thing that mat­ters. Cov­er­age stat­ist­ics de­scribe the sweep. A planted in­stall de­scribes the land­scape's ac­tual abil­ity to no­tice some­thing new. The gap between the two is usu­ally weeks.

Source: Volge ana­lysis, 2026. Con­trol names are gen­eric rather than Flex­era's, so the set can be ap­plied to whichever plat­form a land­scape already runs.

Exhibit The control set, with the test for each

ControlWhat it isHow you test it
Discovery coverageEvery in-scope host and tenancy is swept on a stated cadence, with exclusions registered.Pick three segments at random and confirm the last successful sweep date. Any segment without one is an exclusion that was never registered.
Catalogue normalisationRaw inventory records resolve to one catalogue identity per product version.Measure the unmatched rate and its trend. A rising rate means new software is entering faster than the catalogue is updated.
Entitlement evidenceEvery entitlement in the position traces to a document.Sample ten entitlements and ask for the document. Any that cannot be produced is removed from the position.
Metric configurationEach publisher's counting rule is configured as the contract states it.Re-derive two positions by hand. Record the configuration decision and its contractual basis in writing.
Virtualisation captureHost and cluster topology is current for every virtualised workload.Compare the hypervisor inventory against the asset platform's host records monthly.
Vulnerability matchingValidated advisories are matched to installed inventory, not to a product list.Take one recent critical advisory and ask which hosts carry the affected version. An answer in hours is the pass.
Lifecycle horizonEvery product version carries a support status and an expiry date.Ask what reaches end of support in the next four quarters. If assembling the answer takes more than a day, obsolescence is still being discovered rather than scheduled.
Shadow-deployment detectionSoftware and AI standing outside the approved catalogue is surfaced and routed to an owner.Introduce a known unapproved install in a controlled segment and measure the days to detection.
AI asset registrationDeployed models, agents and AI-enabled features appear in the inventory with an owner and a licence basis.Name the models running in the landscape today. Most organisations cannot, which is the finding.

Failure modes, and what each looks like in production

Land­scape gov­ernance fails in a small num­ber of re­cog­nis­able ways. None of them an­nounces it­self as a fail­ure. Each presents as a work­ing pro­gramme pro­du­cing num­bers nobody has reason to doubt.

01 The con­fid­ent wrong count

A li­cence en­gine is con­figured against a met­ric the con­tract does not use, and it re­ports com­pli­ance. The fail­ure sur­faces when the pub­lisher's aud­itor ap­plies the con­trac­tual met­ric and ar­rives at a dif­fer­ent num­ber. In pro­duc­tion this looks like a pro­gramme that passed its own tests for two years

Technology Asset ManagementVoice of Volge12 •

and then took a seven-fig­ure find­ing. The con­trol is the hand re-de­riv­a­tion, and the evid­ence is the writ­ten con­fig­ur­a­tion de­cision.

02 Nor­m­al­isa­tion drift

New soft­ware enters the land­scape faster than the cata­logue re­cog­nises it, so the un­matched pile grows quietly. The po­s­i­tion stays ac­cur­ate about what it can see and be­comes less and less rep­res­ent­at­ive. In pro­duc­tion this looks like a steady po­s­i­tion along­side a rising count of re­cords nobody has clas­si­fied. The con­trol is the un­matched-rate trend, re­por­ted as a num­ber rather than as a re­as­sur­ance.

03 The orphaned po­s­i­tion

A re­con­cili­ation is pro­duced, cir­cu­lated and never re­freshed. Six months later it is quoted in a re­newal ne­go­ti­ation as the or­gan­isa­tion's po­s­i­tion. In pro­duc­tion this looks like a con­fid­ent fin­ance team ar­guing from a stale arte­fact. The con­trol is the drift meas­ure­ment, and the rem­edy is a stated re­fresh ca­dence with an owner against it.

04 Triage by volume

A vul­ner­ab­il­ity pro­gramme re­ports the num­ber of flaws closed, so the team closes the easy ones. The ex­ploited flaw sits in a queue be­cause patch­ing it re­quires a change win­dow on a sys­tem that mat­ters. In pro­duc­tion this looks like ex­cel­lent re­medi­ation met­rics and an in­cid­ent. The con­trol is to meas­ure time to re­medi­ate for the ex­ploited sub­set sep­ar­ately, and to re­port the two num­bers apart.

05 The end-of-life sys­tem nobody owns

A sys­tem reaches end of sup­port while still work­ing, so no budget line is cre­ated for it. It keeps run­ning be­cause it is re­li­able, which is pre­cisely why it is never sched­uled. In pro­duc­tion this looks like a four-year-old ex­cep­tion that has been re­newed an­nu­ally without a de­cision. The con­trol is a li­fe­cycle ho­ri­zon re­viewed at the same ca­dence as the cap­ital plan, with an owner named against every ex­cep­tion and an ex­piry on the ex­cep­tion it­self.

06 The asset plat­form as a single point of fail­ure

A land­scape view de­pends on con­nect­ors to the sys­tems it counts, and those con­nect­ors break. Flex­era's own status re­cord for 3 to 6 Novem­ber 2026 doc­u­ments a SaaS Man­age­ment in­cid­ent across all pro­duc­tion re­gions, caused by HTTP 504 er­rors from a Mi­crosoft 365 con­nector, with mail­box usage and ap­plic­a­tion user de­tail re­trieval af­fected (Flex­era sys­tem status, in­cid­ent re­cord, read 8 Novem­ber 2026). In pro­duc­tion this looks like a dash­board that is up and a data set that is stale. The con­trol is a fresh­ness in­dic­ator on every panel, show­ing the age of the new­est re­cord rather than the health of the ser­vice.

Volge: on asset plat­forms as de­pend­ency chains

A dashboard that reports the health of the service, and not the age of the data, will tell you everything is fine on the day it stops knowing anything.

What to measure

Four meas­ures carry a land­scape pro­gramme, and all four are lead­ing rather than lag­ging. Audit find­ings and breach counts are the out­comes, and they ar­rive too late to man­age. The meas­ures below move first.

MeasureDefinitionDirectionCadence
Discovery coverageIn-scope hosts and tenancies swept successfully in the period, against the authoritative device countRising to a stated floorMonthly
Unmatched rateInventory records that resolve to no catalogue identity, as a share of recordsFallingMonthly
Position ageDays since the effective licence position was last re-derived, per publisherFlat and lowMonthly
Exploited-flaw remediation timeMedian days from validated advisory to patched host, for advisories with known exploitationFallingMonthly
Lifecycle exposureInstalled instances already past end of support, and instances expiring within four quartersFalling, and scheduledQuarterly
Unregistered AIModels, agents and AI-enabled features found in the landscape with no owner or licence basisFallingQuarterly

Source: Volge ana­lysis, 2026. Defin­i­tions are offered so that a fig­ure means the same thing in two con­sec­ut­ive quar­ters, which is most of the value of a meas­ure.

Note what is ab­sent. There is no ma­tur­ity score, be­cause a score com­presses the one thing a pro­gramme needs to act on, which is the spe­cific gap. And there is no sav­ing fig­ure, be­cause the sav­ing de­pends en­tirely on how badly gov­erned the start­ing land­scape was. A large head­line sav­ing usu­ally de­scribes the baseline rather than the tool.

What done looks like

A land­scape pro­gramme is fin­ished, in the only sense that mat­ters, when four ques­tions can be answered from stand­ing re­cords on the day they are asked. What are we en­titled to run, by pub­lisher, with the doc­u­ment be­hind each en­ti­tle­ment. What is in­stalled, nor­m­al­ised, with the un­matched pile coun­ted and owned. Which in­stalled ver­sions carry a val­id­ated ad­vis­ory with known ex­ploit­a­tion, and which hosts hold

Technology Asset ManagementVoice of Volge13 •

them. And what reaches end of sup­port in­side the plan­ning ho­ri­zon, with a budget line against each item.

None of those four is a pro­ject de­liv­er­able. Each is a stand­ing cap­ab­il­ity, and the test of it is the clock. An or­gan­isa­tion that needs three weeks to an­swer any of the four does not have a land­scape pro­gramme. It has an archive, and an archive is what an aud­itor's let­ter is ad­dressed to.

The owned landscape in the Gulf and India

Bahrain has made software licence compliance a statutory security measure under its data-protection law, which turns a commercial discipline into a criminally backed one. Oman publishes the central government licensing circulars that bind its own enterprise agreements. India mandates a software bill of materials for securities firms, and the platform service that answers it is hosted in North America only. The regional ground changes who is liable, not just what it costs.

Bahrain is the jur­is­dic­tion that takes soft­ware li­cence com­pli­ance out of the com­mer­cial re­gister and into the crim­inal one. Min­is­terial Order No. 43 of 2022, is­sued on 17 March 2022 under the Per­sonal Data Pro­tec­tion Law, re­quires ef­fect­ive breach-mit­ig­a­tion meas­ures and lists among them "com­ply­ing with soft­ware li­censes" at Art­icle 2(3). The same art­icle sits in­side a re­gime whose pen­al­ties are crim­inal rather than ad­min­is­trat­ive. Art­icle 2(4) adds peri­odic vul­ner­ab­il­ity as­sess­ment and pen­et­ra­tion test­ing. Art­icle 4 sets a 72-hour breach no­ti­fic­a­tion dead­line to the Au­thor­ity from dis­cov­ery, with a jus­ti­fic­a­tion re­quired where no­ti­fic­a­tion is late.

Read that pro­vi­sion slowly, be­cause al­most nobody cites it in this con­text. A Bahraini con­trol­ler that suf­fers a per­sonal data breach must show the meas­ures it had in place to mit­ig­ate it. One of the meas­ures the in­stru­ment names is li­cence com­pli­ance. An un­li­censed in­stall is there­fore not only a true-up ex­pos­ure. It is evid­ence of a short­fall in a stat­utory se­cur­ity ob­lig­a­tion, avail­able to a su­per­visor as­sess­ing a breach that may have had noth­ing to do with li­cens­ing at all.

That re­frames the busi­ness case in a way a fin­ance re­view un­der­stands. Every­where else in this book, the li­cence po­s­i­tion de­fends against a bill. In Bahrain it also de­fends against a find­ing. The two de­fences are built from the same arte­fact, which is a nor­m­al­ised count of what is in­stalled set against what the con­tracts per­mit.

Complying with software licenses, listed among the required breach-mitigation measures. Ministerial Order No. 43 of 2022, Article 2(3).

The pro­vi­sion is un­usual enough to be worth quot­ing in full in any Bahraini board paper

Bahrain's wider position, and why it cuts the other way on residency

Bahrain's pos­ture on data loc­a­tion runs against the rest of the re­gion, and a land­scape pro­gramme be­ne­fits from it. The Cloud First Policy ver­sion 1.0, dated 24 April 2017 on its title page and ap­proved by the Su­preme Coun­cil for In­form­a­tion and Com­mu­nic­a­tion Tech­no­logy, states under Data Sov­er­eignty that "the be­ne­fits of cloud are best real­ized when there are no data res­id­ency re­stric­tions placed on data", and that such re­stric­tions "un­der­mine the eco­nom­ies of scale and se­cur­ity be­ne­fits to be gained from shared com­put­ing in­fra­struc­ture". Agen­cies wor­ried about ex­tra­ter­rit­orial ac­cess are dir­ec­ted to vendor se­cur­ity con­trols rather than to local host­ing. The policy binds gov­ern­ment en­tit­ies and not the private sec­tor.

So an in­vent­ory plat­form hos­ted out­side Bahrain faces no res­id­ency ob­jec­tion from the Bahraini state, while the thing it counts car­ries a stat­utory com­pli­ance duty. That is an un­usu­ally clean po­s­i­tion to work in. The Per­sonal Data Pro­tec­tion Law, Law No. 30 of 2018, still gov­erns the in­vent­ory data it­self, and in­vent­ory data is per­sonal data wherever it car­ries a user­name.

Oman publishes its licensing circulars, which almost no state does

Oman runs cent­ral en­ter­prise agree­ments with named soft­ware vendors on be­half of gov­ern­ment units, and it pub­lishes the cir­cu­lars that do it. The Min­istry of Trans­port, Com­mu­nic­a­tions and In­form­a­tion Tech­no­logy is­sued Sum­mary 14/2025 on 24 April 2025, re­new­ing the Mi­crosoft soft­ware li­cens­ing agree­ment for gov­ern­ment in­sti­tu­tions, and Sum­mary 17/2025 on 8 July 2025, cov­er­ing an un­lim­ited usage agree­ment for Or­acle soft­ware li­cences and the re­newal of tech­nical sup­port for gov­ern­ment units. The gov­ern­ing in­stru­ment is Sum­mary 20/2025 of 9 Oc­to­ber 2025, on the gov­ernance of the de­vel­op­ment, use and pub­lic­a­tion of soft­ware in gov­ern­ment units.

Those are the only in­stru­ments found in any of the seven jur­is­dic­tions in this chapter that name a spe­cific soft­ware vendor's li­cens­ing ar­range­ment in a pub­lished gov­ern­ment cir­cu­lar. For any­one ne­go­ti­at­ing in the re­gion, their value is prac­tical. An Omani gov­ern­ment unit's li­cens­ing po­s­i­tion is set cent­rally, which means the ne­go­ti­ation that mat­ters happened at the min­istry and not at the unit. A sup­plier or a sub­si­di­ary that treats it as a local ne­go­ti­ation is ar­guing in the wrong build­ing.

The pro­cure­ment chain around them is equally ex­pli­cit. The Cloud Com­put­ing First Cir­cu­lar, MTCIT Sum­mary 10/2026 of

Technology Asset ManagementVoice of Volge14 •

28 June 2026, re­quires gov­ern­ment en­tit­ies to clas­sify data be­fore stor­ing or pro­cessing it in a cloud en­vir­on­ment, and re­quires cloud ser­vice pro­viders to hold ap­proved cer­ti­fic­a­tions be­fore con­tract­ing. The Cloud First Policy pub­lished on 1 July 2026 re­quires con­tract­ing only with pro­viders hold­ing a valid au­thor­isa­tion is­sued by the Tele­com­mu­nic­a­tions Reg­u­lat­ory Au­thor­ity, and ap­plies to gov­ern­ment bod­ies ex­clud­ing se­cur­ity and mil­it­ary agen­cies. Clas­si­fic­a­tion is a pre­con­di­tion of mi­gra­tion in Oman, not a par­al­lel work­stream.

Oman's cross-border penalty is the one asymmetry worth designing around

An in­vent­ory sweep ex­ports per­sonal data, and Oman prices that ex­port dif­fer­ently from every other breach of its law. The Per­sonal Data Pro­tec­tion Law, Royal De­cree No. 6/2022, car­ries crim­inal fines im­posed by the com­pet­ent court. Most con­tra­ven­tions sit between 500 and 20,000 Rial Omani. Art­icle 23, un­law­ful cross-bor­der trans­fer, sits at 100,000 to 500,000. The cross-bor­der band is five to a hun­dred times every other band, and it at­taches to nat­ural per­sons as well as to legal ones.

The re­gime has been fully en­force­able since 5 Feb­ru­ary 2026, when the trans­ition period under the Ex­ec­ut­ive Reg­u­la­tions closed. Two fur­ther ob­lig­a­tions bear dir­ectly on an asset plat­form. The Reg­u­la­tions re­quire a re­cord of pro­cessing activ­it­ies from both con­trol­ler and pro­cessor, and that re­cord must cover any move­ment or pro­cessing of per­sonal data out­side Oman. And a data pro­tec­tion of­ficer is man­dat­ory for con­trol­lers, with the of­ficer's name and con­tact de­tails pub­lished, which makes Oman the only one of these jur­is­dic­tions with a pub­lished-of­ficer re­quire­ment.

For a land­scape pro­gramme the con­sequence is nar­row and con­crete. Host­names, user­names and in­stall­a­tion paths leav­ing Oman for a host­ing re­gion else­where is a cross-bor­der trans­fer, and it be­longs in the re­cord of pro­cessing activ­it­ies with its legal basis stated. The gate under the private-sec­tor re­gime is con­sent plus a doc­u­mented trans­fer risk as­sess­ment rather than a prior min­is­terial ap­proval. Gov­ern­ment-held per­sonal data is dif­fer­ent. Cir­cu­lar Sum­mary 6/2024 of 31 March 2024 em­phas­ises pro­cessing within Oman and re­quires prior ap­proval of the Cyber De­fence Centre be­fore gov­ern­ment-held per­sonal data moves out.

India mandates the bill of materials, and the service that answers it sits in North America

India is the sharpest col­li­sion in this chapter, and both sides of it are veri­fi­able. The Se­cur­it­ies and Ex­change Board of India's Cy­ber­se­cur­ity and Cyber Re­si­li­ence Frame­work, cir­cu­lar SEBI/HO/ITD-1/ITD_CSC_EXT/P/CIR/2024/113 of 20 Au­gust

2024, sets stand­ard GV.SC.S5 in these words: "SBOM shall be ob­tained for all new soft­ware pro­cure­ments of core and crit­ical activ­it­ies and kept up­dated with every up­grade or change." Where an SBOM can­not be ob­tained for leg­acy or pro­pri­et­ary sys­tems, the board, part­ners or pro­pri­etor must ap­prove that po­s­i­tion with a stated lim­it­a­tion, ra­tionale and risk man­age­ment ap­proach. Every com­pli­ance date under the frame­work has passed, the last of them on 31 Au­gust 2025.

Flex­era One's SBOM Man­age­ment ser­vice, on Flex­era's own sys­tem status re­gister read on 8 Novem­ber 2026, is lis­ted under North Amer­ica and nowhere else. The re­gister shows Flex­era One op­er­at­ing in three re­gions, North Amer­ica, Europe and APAC, with the API en­d­points api.flex­era.com, api.flex­era.eu and api.flex­era.au. The APAC re­gion is Aus­tralia, which the main­ten­ance no­tices con­firm by name. There is no Middle East re­gion and no India re­gion for any Flex­era One mod­ule.

So a SEBI-reg­u­lated firm car­ries a bind­ing In­dian ob­lig­a­tion to hold a soft­ware bill of ma­ter­i­als, and the plat­form cap­ab­il­ity built for that ob­lig­a­tion is hos­ted on the other side of the world. The ob­lig­a­tion is still met by ob­tain­ing the doc­u­ment from the sup­plier, which is what the stand­ard re­quires. What can­not be done is to meet it in­side an In­dian host­ing bound­ary using this plat­form. A firm whose su­per­vis­ory sub­mis­sions as­sert In­dian host­ing for its com­pli­ance tool­ing should check that as­ser­tion be­fore an ex­am­iner does.

Sources: Flex­era sys­tem status re­gister, read 8 Novem­ber 2026, for host­ing re­gions. Legal in­stru­ments as cited in the text, each with its is­su­ing body and date. Com­pil­a­tion by Volge, 2026. Qatar is omit­ted be­cause the Qatari cloud and clas­si­fic­a­tion in­stru­ments could not be veri­fied from a primary source.

Technology Asset ManagementVoice of Volge15 •

Exhibit The regional position for an asset management programme

JurisdictionFlexera One hosting in regionThe binding hook on software assetsThe supplier gate
BahrainNoneMinisterial Order 43 of 2022, Article 2(3): licence compliance is a required breach-mitigation measure, criminally backed through the PDPLNo national accreditation scheme; international certifications specified instead
OmanNoneMTCIT Summary 20/2025 on software governance in government units, with central Microsoft and Oracle agreements at Summary 14/2025 and 17/2025TRA authorisation before contracting, plus classification before migration
Saudi ArabiaNoneNo licence-specific instrument identified; the constraint is the prohibition on exporting government agency dataCST registration at Class A, B or C by data classification
United Arab EmiratesNoneFederal PDPL in force since 2 January 2022 with no Executive Regulations as at 8 November 2026; health data cannot leave the StateDESC Cloud Service Provider Security Standard for Dubai government and semi-government
KuwaitNoneCITRA cloud framework prohibits hosting Level 3 and Level 4 data and government-held personal data outside KuwaitCITRA cloud licence, requiring a data centre inside Kuwait
IndiaNone; nearest region is AustraliaSEBI CSCRF GV.SC.S5 mandates an SBOM for new software procurements of core and critical activitiesMeitY empanelment and STQC audit status for SEBI regulated entities under the 2023 cloud circular

There is no sovereign edition of this software anywhere

Flex­era ships no gov­ern­ment or sov­er­eign edi­tion of Flex­era One in any jur­is­dic­tion. The only gov­ern­ment entries on its status re­gister are CloudCheckr GOV and CloudCheckr Fed­eral, both Amer­ican and both a dif­fer­ent product. Cloud Com­mit­ment Man­age­ment runs in the United States only.

Set that against what the Gulf states are doing and the asym­metry is the story. Four of them op­er­ate a na­tional or emir­ate-level ac­cred­it­a­tion gate that a cloud sup­plier must pass be­fore it can sell to the pub­lic sec­tor at all. Saudi Ar­a­bia re­quires re­gis­tra­tion with the Com­mu­nic­a­tions, Space and Tech­no­logy Com­mis­sion, at Class A for pub­lic gov­ern­ment data, Class B for con­fid­en­tial and Class C for secret or top secret. Dubai's Elec­tronic Se­cur­ity Cen­ter op­er­ates a Cloud Ser­vice Pro­vider

Se­cur­ity Stand­ard that is man­dat­ory, in its own words, for all pro­viders wish­ing to offer cloud ser­vices to Dubai gov­ern­ment and semi-gov­ern­ment en­tit­ies, with an­nual sur­veil­lance audits and re­cer­ti­fic­a­tion every three years. Kuwait re­quires a CITRA cloud li­cence, and clause 2.1 of the Cloud Com­put­ing Reg­u­lat­ory Frame­work ver­sion 2.4 per­mits only a CITRA-li­censed party with a data centre and in­fra­struc­ture in­side Kuwait to con­tract with the pub­lic sec­tor. Oman re­quires a TRA au­thor­isa­tion.

Gulf states are cent­ral­ising soft­ware li­cens­ing at the level of the state, and the vendor that sells li­cens­ing man­age­ment has no state-level of­fer­ing any­where in the re­gion. A pub­lic-sec­tor buyer in Riy­adh, Kuwait City or Mus­cat is there­fore choos­ing between a manual prac­tice and a plat­form its own pro­cure­ment rules may not admit. That is a sourcing de­cision to take de­lib­er­ately, and to re­cord.

India's prudential rules already treat asset management as supervision

In­dian fin­an­cial reg­u­lat­ors reached soft­ware asset man­age­ment be­fore most In­dian boards did, and they wrote it as pruden­tial rule rather than as good prac­tice. The Re­serve Bank of India's In­form­a­tion Tech­no­logy Gov­ernance Dir­ec­tions of 2023 pro­hib­ited run­ning out­dated and un­sup­por­ted hard­ware or soft­ware, re­quired mon­it­or­ing of end-of-sup­port dates and main­ten­ance con­tracts, and re­quired a tech­no­logy re­fresh plan. They also re­quired source code for all crit­ical ap­plic­a­tions to be ob­tained from the vendor or placed in es­crow, and a de­veloper or vendor cer­ti­fic­ate that the ap­plic­a­tion is free of known vul­ner­ab­il­it­ies, mal­ware "and any cov­ert chan­nels in the code", re­freshed on ma­ter­ial change.

The source-code and cer­ti­fic­ate ob­lig­a­tions are car­ried into the Re­serve Bank of India (In­form­a­tion Tech­no­logy and Cyber Se­cur­ity Gov­ernance) Dir­ec­tions for com­mer­cial banks of 2026, at para­graphs 90 and 92, fol­low­ing the con­sol­id­a­tion of 31 July 2026. The end-of-sup­port and re­fresh-plan pro­vi­sion sat in the re­pealed 2023 Dir­ec­tion, and its sur­vival into the 2026 text was not sep­ar­ately con­firmed in the re­search be­hind this chapter. Check it against the cur­rent in­stru­ment be­fore re­ly­ing on it in a board paper.

Technology Asset ManagementVoice of Volge16 •

One carve-out mat­ters for how a pro­gramme is scoped. Ap­pendix III to the Mas­ter Dir­ec­tion on Out­sourcing of In­form­a­tion Tech­no­logy Ser­vices, RBI/2023-24/102 of 10 April 2023 and ef­fect­ive 1 Oc­to­ber 2023, ex­cludes "li­censed soft­ware ac­quis­i­tion with no or min­imal cus­tom­isa­tion" from the defin­i­tion of out­sourcing. An or­din­ary li­cence pur­chase there­fore does not trig­ger the out­sourcing re­gime. A cus­tom­ised or man­aged de­ploy­ment does, and with it clause 16(7) on stor­age of data only in India where reg­u­lat­ory re­quire­ments apply, clause 17(8) re­quir­ing the reg­u­lated en­tity to re­port a cyber in­cid­ent to the Re­serve Bank within six hours of de­tec­tion by the ser­vice pro­vider, and the Chapter X exit terms under which a pro­vider is pro­hib­ited from eras­ing, pur­ging, re­vok­ing, al­ter­ing or chan­ging any data dur­ing trans­ition.

The AI in the landscape, and who is liable for what it does

India holds the only pro­vi­sion in these jur­is­dic­tions that al­loc­ates li­ab­il­ity for an AI sys­tem's out­put, and it lands on the firm de­ploy­ing it. The Se­cur­it­ies and Ex­change Board of India (In­ter­me­di­ar­ies) (Amend­ment) Reg­u­la­tions, 2025, no­ti­fic­a­tion SEBI/LAD-NRO/GN/2025/226, in­ser­ted reg­u­la­tion 16C on re­spons­ib­il­ity for the use of ar­ti­fi­cial in­tel­li­gence tools. A reg­u­lated per­son using AI or ma­chine learn­ing tools, whether de­signed in-house or pro­cured from a third-party tech­no­logy ser­vice pro­vider, is solely re­spons­ible for the pri­vacy, se­cur­ity and in­teg­rity of in­vestor and stake­holder data, for the out­put arising from the use of such tools, and for com­pli­ance with ap­plic­able law, "ir­re­spect­ive of the scale and scen­ario of ad­op­tion". Sources give 6 Feb­ru­ary 2025 and 10 Feb­ru­ary 2025 for gaz­ette pub­lic­a­tion and the dis­crep­ancy could not be re­solved from a primary page.

Buy­ing the model does not move the li­ab­il­ity. That is the whole con­tent of the pro­vi­sion, and it is the strongest avail­able ar­gu­ment for treat­ing de­ployed AI as an asset on the re­gister rather than as a ser­vice on an in­voice. A model you can­not name is a model whose out­put you are non­ethe­less an­swer­able for.

India also holds the only in­stru­ment in these jur­is­dic­tions that names an AI bill of ma­ter­i­als. CERT-In's Tech­nical Guidelines on SBOM, QBOM, CBOM, AIBOM and HBOM, CISG-2024-02 ver­sion 2.0, were is­sued on 3 Oc­to­ber 2024. They are non-bind­ing guidelines rather than dir­ec­tions under sec­tion 70B(6) of the In­form­a­tion Tech­no­logy Act, 2000, and they should be cited as guid­ance. CERT-In added Ar­ti­fi­cial In­tel­li­gence Se­cur­ity Audit­ing Guidelines, CISG-2026-04, on 4 Novem­ber 2026. Taken to­gether with reg­u­la­tion 16C, the In­dian dir­ec­tion of travel is clear enough to plan against, and an AI asset re­gister is the arte­fact both sides of it re­quire.

Saudi Arabia, where the constraint is export rather than licensing

Saudi Ar­a­bia has no in­stru­ment mak­ing li­cence com­pli­ance a se­cur­ity meas­ure, and it has the re­gion's hard­est rule on mov­ing data out. Art­icle 3-3-5 of the Cloud Com­put­ing Ser­vices Pro­vi­sion­ing Reg­u­la­tions ver­sion 4 pro­hib­its a pro­vider or a sub­scriber from trans­fer­ring con­tent com­pris­ing data of Saudi gov­ern­ment agen­cies out­side the King­dom, for any pur­pose and in any form, per­man­ently or tem­por­ar­ily, in­clud­ing tem­por­ary stor­age and backup, un­less an­other Saudi in­stru­ment ex­pressly per­mits it. The Reg­u­la­tions were is­sued by the Com­mu­nic­a­tions, Space and Tech­no­logy Com­mis­sion under De­cision No. 506/1445 and took ef­fect on 10 Oc­to­ber 2023. The art­icle num­ber and word­ing rest on named law-firm com­ment­ary, Ald­habaan Law Firm writ­ing in the Cham­bers Cloud Com­put­ing 2026 guide, be­cause the Com­mis­sion's own file could not be read.

For an asset pro­gramme serving a Saudi gov­ern­ment cus­tomer, that pro­vi­sion gov­erns the in­vent­ory it­self. An in­vent­ory re­cord of a gov­ern­ment agency's in­stalled soft­ware is con­tent com­pris­ing that agency's data. No Flex­era One host­ing re­gion ex­ists in the King­dom, so the hon­est op­tions are a manual prac­tice in­side the King­dom, or an in­stru­ment ex­pressly per­mit­ting the trans­fer. Treat the second as a legal ques­tion rather than an ar­chi­tec­tural one.

One widely re­peated Saudi claim should be dropped rather than passed on. The Na­tional Data Man­age­ment Of­fice's Data Man­age­ment and Per­sonal Data Pro­tec­tion Stand­ards, ver­sion 1.5 of Janu­ary 2021, con­tain no data res­id­ency, loc­al­isa­tion or in-King­dom host­ing re­quire­ment, which was es­tab­lished by search­ing the whole in­stru­ment. The Na­tional Cy­ber­se­cur­ity Au­thor­ity's Cloud Cy­ber­se­cur­ity Con­trols, CCC-2:2024, de­leted the sub-con­trols of the 2020 edi­tion that had re­quired cloud ser­vices to be provided from in­side the King­dom, re­cord­ing that loc­al­isa­tion con­trols had moved to the Na­tional Data Man­age­ment Of­fice. The re­ceiv­ing in­stru­ment has not been iden­ti­fied. Quote the in­stru­ment being re­lied on, and name it.

For a Saudi fin­an­cial in­sti­tu­tion the bind­ing con­straint is an ap­proval. The Saudi Cent­ral Bank's Cyber Se­cur­ity Frame­work, No. 381000091275 of 24 May 2017 and in force, states at 3.4.3 that in prin­ciple only cloud ser­vices loc­ated in Saudi Ar­a­bia should be used, with ex­pli­cit ap­proval re­quired where they are not, and it re­quires ap­proval be­fore cloud ser­vices are used at all. The Rules on Out­sourcing, No. 41027017 of 15 Decem­ber 2019, add a writ­ten no-ob­jec­tion for over­seas out­sourcing to­gether with a ra­tionale for why the work can­not be done within the King­dom. An asset man­age­ment plat­form

Technology Asset ManagementVoice of Volge17 •

hos­ted in Europe or Aus­tralia is over­seas out­sourcing on those terms.

Kuwait writes the exit terms, which matter more here than the hosting rule

Kuwait's cloud frame­work is the only in­stru­ment in the re­gion that writes port­ab­il­ity and exit into the rules, and an asset man­age­ment plat­form is ex­actly the kind of sys­tem those rules were draf­ted for. Chapter 6 of the Cloud Com­put­ing Reg­u­lat­ory Frame­work places data own­er­ship with the sub­scriber at 6.4.9, lim­its use to ser­vice pro­vi­sion at 6.4.10, per­mits sub-con­tract­ing with ul­ti­mate re­spons­ib­il­ity rest­ing on the pro­vider at 6.4.14, and at 6.4.17 ob­liges the pro­vider to re­turn con­tent in a cus­tom­ary format or mi­grate it dir­ectly to an­other pro­vider.

Then come the two clauses worth quot­ing in a con­tract re­view. Clauses 6.4.17.5 and 6.4.17.6 pro­hibit a pro­vider from with­hold­ing gov­ern­ment data or ob­struct­ing mi­gra­tion on grounds of un­settled pay­ments or an on­go­ing dis­pute, and guar­an­tee ser­vice con­tinu­ity for at least six months from the onset of a dis­pute. Chapter 7 voids un­fair lim­it­a­tion-of-li­ab­il­ity terms. A land­scape plat­form holds the re­cord of everything an or­gan­isa­tion owns, which makes a host­age scen­ario dur­ing a billing dis­pute a live op­er­a­tional risk rather than a the­or­et­ical one.

The host­ing side of the Kuwaiti frame­work is re­strict­ive and simple. Level 3 and Level 4 data may not be hos­ted out­side Kuwait, along with in­di­vidu­als' per­sonal data held by gov­ern­ment bod­ies, private-sec­tor com­pan­ies or ser­vice pro­viders, and Level 3 data must sit in a private or hy­brid cloud under clauses 4.2.1.1 and 3.1.4.3 of the Ar­abic primary text. Clauses 4.3.3 and 4.3.4 for­bid a pro­vider from com­ply­ing with a for­eign jur­is­dic­tional de­mand for sub­scriber con­tent. There is no Kuwaiti host­ing re­gion for any plat­form in this book, so a Kuwaiti pub­lic-sec­tor land­scape pro­gramme is a manual prac­tice until one ex­ists.

The United Arab Emirates runs two regimes that pull apart

The fed­eral Per­sonal Data Pro­tec­tion Law, Fed­eral De­cree-Law No. 45 of 2021, has been in force since 2 Janu­ary 2022 and still had no Ex­ec­ut­ive Reg­u­la­tions as at 8 Novem­ber 2026. It there­fore car­ries no com­pli­ance com­mence­ment date, no stat­utory breach no­ti­fic­a­tion dead­line and no pub­lished pen­alty sched­ule. Art­icle 2(2) carves out gov­ern­ment data, free zones with their own le­gis­la­tion, health data and bank­ing data. DIFC Law No. 5 of 2020 and the Abu Dhabi Global Mar­ket reg­u­la­tions sit sep­ar­ately, and the Abu Dhabi text could not be veri­fied in the re­search be­hind this chapter.

Against that soft­ness sits the coun­try's hard­est loc­al­isa­tion rule, and it is not a data-pro­tec­tion in­stru­ment at all. Art­icle 13 of Fed­eral Law No. 2 of 2019 on the Use of In­form­a­tion and Com­mu­nic­a­tions Tech­no­logy in Health Fields, in force since 14 May 2019, pro­hib­its stor­ing, pro­cessing, gen­er­at­ing or trans­fer­ring health data re­lat­ing to health ser­vices provided in­side the State out­side the State, free zones in­cluded. Art­icle 24 at­taches a fine of not less than AED 500,000 and not more than AED 700,000. An in­vent­ory sweep of a UAE hos­pital's clin­ical sys­tems col­lects data that pro­vi­sion reaches, and the pro­vi­sion is crim­inal rather than ad­min­is­trat­ive.

What to write down, and where

A re­gional po­s­i­tion is only use­ful once it sits on paper in three places, and the same facts serve all three. Keep the word­ing identical across them. A dif­fer­ence between them is what an ex­am­iner finds.

  • In the ar­chi­tec­ture de­cision re­cord. The host­ing re­gion of the plat­form and of every mod­ule in use, the data fields the in­vent­ory col­lects, and the coun­tries from which sup­port staff can read them.
  • In the con­tract. The host­ing re­gion re­lied upon, by name, to­gether with the exit and port­ab­il­ity terms. Where a jur­is­dic­tion writes those terms into law, as Kuwait does, cite the clause rather than the policy.
  • In the su­per­vis­ory sub­mis­sion. The law­ful basis for col­lect­ing and ex­port­ing in­vent­ory data, the ac­cred­it­a­tion the sup­plier holds or does not hold, and the con­trol that would de­tect a breach of the stated po­s­i­tion rather than the policy for­bid­ding it.

The line most often miss­ing from all three is the simplest. State plainly that no host­ing re­gion ex­ists in the jur­is­dic­tion, be­cause that is the fact on which every other po­s­i­tion in this chapter rests, and it does not im­prove by being left out.

What this constrains

The re­gional ground changes the order of work and the iden­tity of the buyer, rather than the ar­gu­ment of this book. Settle the law­ful basis for in­vent­ory data be­fore the first sweep, be­cause that data is per­sonal data in every jur­is­dic­tion here and its ex­port is the most heav­ily pen­al­ised act in the Omani stat­ute. Es­tab­lish where the plat­form is hos­ted and say so in writ­ing, be­cause none of these jur­is­dic­tions has a host­ing re­gion and three of them reg­u­late who may sup­ply the ser­vice at all. In Bahrain, carry the li­cence po­s­i­tion into the se­cur­ity file as well as the fin­ance file, be­cause Order 43 of 2022 puts it there. In Oman, read the cent­ral cir­cu­lars be­fore ne­go­ti­at­ing loc­ally.

Technology Asset ManagementVoice of Volge18 •

And in India, hold the bill of ma­ter­i­als as a doc­u­ment the sup­plier owes you, in­de­pend­ent of whatever plat­form stores it.

One more line be­longs in every re­gional po­s­i­tion and is usu­ally miss­ing. Name what is not in re­gion, in­clud­ing the in­vent­ory plat­form, the vul­ner­ab­il­ity ad­vis­ory feed and any sup­port ac­cess from out­side the geo­graphy. A con­straint dis­closed in ad­vance is a design de­cision. The same fact found by a su­per­visor is a find­ing.

The operator's read.

Three liabilities, one landscape. How to carry the owned technology landscape as a managed risk, the buying questions, the honest objections, a worked example, and the routes to capability.

The three liabilities share one root cause: a landscape no one can fully see

Com­ply, se­cure and re­tire look like three prob­lems owned by three func­tions, namely asset man­age­ment, se­cur­ity and in­fra­struc­ture. They are three symp­toms of one con­di­tion. An or­gan­isa­tion can­not see, ac­cur­ately and in one place, what it has in­stalled. The audit ex­ploits the gap between de­ploy­ment and en­ti­tle­ment. The breach ex­ploits the un­patched soft­ware no one in­vent­or­ied. The ob­sol­es­cence ac­crues on ver­sions no one tracked. Fix the vis­ib­il­ity and you do not fix three prob­lems; you re­move the ground all three grow in.

This is the ar­gu­ment for one land­scape view rather than three point tools. When the same nor­m­al­ised, Tech­no­pe­dia-ref­er­enced in­vent­ory feeds the li­cence po­s­i­tion, the vul­ner­ab­il­ity triage, and the li­fe­cycle sched­ule, the three func­tions are work­ing from one truth in­stead of three par­tial ones. The se­cur­ity team's patch list and the asset team's li­cence po­s­i­tion are drawn from the same count of what is in­stalled, which is the only way the AI a team de­ployed last month shows up in all three views at once rather than in none of them.

Gov­erned sep­ar­ately, the li­ab­il­it­ies hide in each other's blind spots. A se­cur­ity tool built for Mi­crosoft misses the third-party sprawl; an asset tool built for li­cences misses the vul­ner­ab­il­ity; neither is look­ing for the end-of-life date or the shadow model. The single land­scape is what makes an asset gov­ern­able on all three axes at the mo­ment it is dis­covered, rather than on one axis by one team while the other ex­pos­ures wait to be called.

And we keep the dis­closed scep­ti­cism from the com­pan­ion volume. Flex­era as­sembled this breadth over years and by ac­quis­i­tion, and the single-land­scape prom­ise is con­di­tional on those pieces shar­ing one clean in­vent­ory in prac­tice, not just on one slide. A buyer should ask to see the li­cence po­s­i­tion, the vul­ner­ab­il­ity list, and the li­fe­cycle sched­ule drawn from one dis­cov­ery of their own en­vir­on­ment be­fore be­liev­ing the one-land­scape story. We ask on our cli­ents' be­half.

Exhibit 8 One landscape view, three liabilities discharged

  • One in­vent­ory. Nor­m­al­ised, Tech­no­pe­dia-ref­er­enced
  • Com­ply. Li­cence po­s­i­tion
  • Se­cure. Vul­ner­ab­il­ity triage
  • Re­tire. Li­fe­cycle sched­ule

Source: Volge ana­lysis, 2026. The value is con­di­tional on the pieces shar­ing one in­vent­ory. Verify with your data.

What the survey shows, and what we will not pretend it is

A dis­closed-in­terest ana­lysis should be can­did about the qual­ity of its own evid­ence, and here cand­our re­quires a dis­tinc­tion the com­pan­ion book did not. Much of this book's data comes from Flex­era's own 2026 State of ITAM Re­port, a sur­vey of 512 prac­ti­tion­ers that is genu­inely in­de­pend­ent in its re­spond­ents but pro­duced by the vendor selling the rem­edy. We treat those fig­ures as dir­ec­tional rather than de­cis­ive, and we lean the ar­gu­ment on the struc­tural facts, which do not de­pend on any sur­vey.

Evidence The 2026 landscape, in Flexera's survey figures

FindingFigureSource
Organisations audited in the past year48%Flexera ITAM, 24 Jun 2026
Spent >$1M on software audits over three years44%Flexera ITAM, 24 Jun 2026
Respondents reporting a Microsoft audit in three years64%Flexera ITAM, 24 Jun 2026
Complete IT asset visibility (43% in 2025)36%Flexera ITAM, 24 Jun 2026
Accurate visibility into AI software31%Flexera ITAM, 24 Jun 2026
Respondents reporting wasted AI spend rising59%Flexera ITAM, 24 Jun 2026

Source: Flex­era 2026 State of ITAM Re­port, 24 June 2026 (n=512). This is a vendor sur­vey of in­de­pend­ent prac­ti­tion­ers, and it is treated as dir­ec­tional.

Technology Asset ManagementVoice of Volge19 •

The struc­tural ar­gu­ment stands without the sur­vey. That en­ter­prise soft­ware vendors use audits as a rev­enue chan­nel, that li­cens­ing met­rics are com­plex enough to guar­an­tee an en­ti­tle­ment gap, that end-of-life soft­ware stops re­ceiv­ing patches, and that de­ployed AI is soft­ware a land­scape must in­vent­ory: none of these is a sur­vey find­ing. They are how the li­cens­ing, se­cur­ity, and sup­port sys­tems are built, and they would be true if no one had run a sur­vey at all. The fig­ures put a size on a prob­lem whose ex­ist­ence does not de­pend on them.

So we use the num­bers to in­dic­ate scale and dir­ec­tion, not to close the case, and we would rather you dis­count them as vendor re­search than have you feel misled when you no­tice their source. The case for gov­ern­ing the owned land­scape rests on the mech­an­ics of audits, of vul­ner­ab­il­it­ies and of sup­port cal­en­dars. It rests too on your own in­vent­ory, which is the only evid­ence about your land­scape that mat­ters.

Six questions to ask before you spend on landscape governance

The fol­low­ing six ques­tions are the ones Volge puts to a cli­ent be­fore re­com­mend­ing any plat­form, and they in­ter­rog­ate us as read­ily as Flex­era. A pro­gramme that can an­swer them with num­bers and named own­ers is ready to buy. One that an­swers with in­ten­tions should start with a dis­cov­ery, which is small, fast, and the hon­est first step to­ward car­ry­ing the land­scape as a man­aged risk rather than an un­budgeted one.

01 Can we state our own li­cence po­s­i­tion today?

For Mi­crosoft, Or­acle, and the other audit-prone vendors, be­fore a let­ter ar­rives. If not, the next audit is ne­go­ti­ated on the vendor's num­bers, not yours.

02 Do we see third-party soft­ware, or only the OS?

The at­tack sur­face is the third-party layer. A se­cur­ity pro­gramme blind to it is patch­ing the wrong land­scape di­li­gently.

03 What in our land­scape is end-of-life right now?

If the an­swer takes weeks to as­semble, ob­sol­es­cence is being dis­covered dur­ing in­cid­ents rather than sched­uled in ad­vance.

04 Can we find the AI we have de­ployed?

Only 31 per cent can. If you can­not in­vent­ory your mod­els and agents, you can­not se­cure them. You can­not li­cense or re­tire them either.

05 Do our three func­tions share one in­vent­ory?

If asset, se­cur­ity, and in­fra­struc­ture work from three counts, the single-land­scape be­ne­fit is the­or­et­ical. Ask to see one dis­cov­ery feed all three.

06 Who op­er­ates it after the li­cence is signed?

A land­scape view left un-op­er­ated ages into an­other un­sup­por­ted sys­tem. De­cide who runs the loop, or do not buy the loop.

The test

A pro­gramme that an­swers these six with num­bers and names is ready to spend. One that an­swers with hopes should start with a small and fast dis­cov­ery. That is the hon­est first step.

The four objections a sceptical board raises, and the honest reply to each

A well-run board does not buy on an audit stat­istic, and the four ob­jec­tions below are the ones we meet most often, each reas­on­able, none dis­missed with a slo­gan. We set them out be­cause an ana­lysis that ducks the hard ques­tions is not worth read­ing, and be­cause the an­swers are, in most cases, stronger than the ob­jec­tion.

01 “Our se­cur­ity team already scans for vul­ner­ab­il­it­ies.”

Very likely, and the scan­ner pro­duces a list far longer than the team can ac­tion. The gap is not de­tec­tion but pri­or­it­isa­tion, mean­ing which flaws are real, present and ex­ploit­able. Val­id­ated re­search earns its place there. An­other scan­ner does not.

02 “SAM is an old, un­glam­or­ous dis­cip­line.”

It is, and that is the point. The audit is old and un­glam­or­ous too, and it costs 44 per cent of the or­gan­isa­tions in Flex­era's June 2026 sur­vey over a mil­lion dol­lars. Bor­ing dis­cip­lines that pre­vent ex­pens­ive sur­prises are ex­actly the ones worth fund­ing.

03 “The AI-asset tool­ing is im­ma­ture.”

Partly. The dis­cip­line is right: in­vent­ory, then se­cure, then re­tire de­ployed AI. Flex­era shipped AI Cost Man­age­ment on 10 June 2026 across agents, mod­els, data and com­pute, and ex­pan­ded it on 20 Oc­to­ber 2026, so the tool­ing ex­ists. Its own cap­ab­il­ity mat­rix still marks AI policy and gov­ernance as par­tial. Start with soft­ware and audit, where the cap­ab­il­ity is proven, and treat AI gov­ernance as the newer ground it is.

04 “Why not buy the soft­ware dir­ect?”

You can, and the li­cence is the easy part. The value leaks in the work around it, which is the re­con­cili­ation, the audit ne­go­ti­ation, the patch triage and the re­tire­ment sched­ule. None of that is a login. A land­scape plat­form bought and left un­op­er­ated ages into an­other un­sup­por­ted sys­tem on its own re­gister, which is the fail­ure this book opened with.

The pattern in the answers

Each ob­jec­tion as­sumes the pur­chase is the soft­ware. It is not. The pur­chase is a seen, re­con­ciled, con­tinu­ously gov­erned

Technology Asset ManagementVoice of Volge20 •

land­scape, and the dis­cip­line to dis­charge its li­ab­il­it­ies be­fore they are called. The soft­ware is a means, and the gov­ernance is the thing.

One landscape, three letters, an audit, a CVE, and an expiry notice

A single com­pos­ite shows the method more clearly than de­scrip­tion, so here is one land­scape re­ceiv­ing, in a single quarter, the three let­ters this book is about, il­lus­trat­ive, drawn to match the shape gov­ernance com­monly re­veals rather than any one cli­ent. The point is not the amounts but the pat­tern. One land­scape view an­swers all three, and how their ab­sence turns each into a scramble.

The three letters

An Or­acle audit no­tice. A crit­ical vul­ner­ab­il­ity dis­closed in a widely-in­stalled third-party lib­rary. And a vendor's no­tice that a data­base ver­sion run­ning a core re­port reaches end-of-life in ninety days. Three func­tions and three dead­lines on one land­scape, and, in the un-gov­erned case, three sep­ar­ate pan­ics con­duc­ted from three par­tial in­vent­or­ies.

Exhibit 9 Three liabilities called at once: governed versus not (illustrative)

LetterUn-governed responseOne-landscape response
Oracle auditWeeks assembling a count; accept the findingLicence position already current; contest the inflated figure
Critical CVEIs it even installed? Where? UnknownMatched to inventory in hours; patch the exposed hosts first
End-of-life noticeDiscovered late; emergency migrationAlready on the lifecycle schedule; replacement budgeted

Source: il­lus­trat­ive com­pos­ite by Volge, 2026, of pat­terns com­monly seen. Not a spe­cific cli­ent bench­mark.

Reading the teardown

The gov­erned column is not faster be­cause the team is bet­ter; it is faster be­cause all three an­swers come from one in­vent­ory that already ex­is­ted be­fore the let­ters ar­rived. The li­cence po­s­i­tion was cur­rent, so the audit is a ne­go­ti­ation. The vul­ner­ab­il­ity matches a known land­scape, so triage is hours not days. The end-of-life date was on a sched­ule, so the mi­gra­tion is planned not pan­icked. The un-gov­erned column is three scrambles be­cause there was no shared count to an­swer from, which is the whole cost of not hav­ing one.

The honest caveat

A vendor deck would total a sav­ing here. We will not, be­cause it de­pends en­tirely on how un-gov­erned the start­ing

land­scape was. A large head­line num­ber usu­ally means the baseline was badly man­aged rather than the tool being bril­liant. Your own dis­cov­ery pro­duces the only fig­ure that mat­ters.

The asset and security leadership agenda

Six decisions a chief information officer or chief information security officer can take this quarter, each of them cheap and none of them a platform purchase.

The de­cisions that de­term­ine whether the owned land­scape is gov­erned are or­gan­isa­tional rather than tech­nical, and they can be taken be­fore any plat­form is se­lec­ted. Six are set out below. Each has an owner, a cost meas­ured in days rather than quar­ters, and a test show­ing whether it was ac­tu­ally taken.

  • Name an owner for each audit-prone pub­lisher. One per­son hold­ing the po­s­i­tion, the con­tract archive and the ne­go­ti­ation his­tory. The test is whether that per­son can state the cur­rent po­s­i­tion without ask­ing any­one.
  • Settle the law­ful basis for in­vent­ory data, and re­cord where it is hos­ted. Host­names and user­names are per­sonal data in every jur­is­dic­tion in chapter five, and their ex­port is the most heav­ily pen­al­ised act in the Omani stat­ute.
  • Re­quire one in­vent­ory be­hind the li­cence po­s­i­tion, the patch list and the li­fe­cycle sched­ule. Ask to see all three drawn from a single dis­cov­ery of your own land­scape be­fore ac­cept­ing that they are.
  • Re­port ex­ploited-flaw re­medi­ation time sep­ar­ately from everything else. Re­por­ted along­side the volume, it dis­ap­pears.
  • Put the li­fe­cycle ho­ri­zon on the cap­ital plan, with an ex­piry date on every ex­cep­tion. An ex­cep­tion re­newed an­nu­ally without a de­cision is a per­man­ent po­s­i­tion nobody chose.
  • Open an AI asset re­gister this quarter, how­ever in­com­plete. Six fields and an owner per entry is enough to start, and the first draft is where the shadow de­ploy­ments sur­face.

None of the six re­quires a li­cence. Three of them re­quire noth­ing but au­thor­ity, which is why a pro­gramme that has not taken them should be scep­tical of its own tool­ing plans.

One fur­ther de­cision be­longs at board level rather than in the tech­no­logy func­tion. De­cide, in writ­ing, which li­ab­il­it­ies the or­gan­isa­tion is de­lib­er­ately car­ry­ing. Under-li­censed po­s­i­tions left open against a judge­ment about audit prob­ab­il­ity, vul­ner­ab­il­it­ies car­ried for want of a change win­dow, and end-of-life sys­tems re­tained be­cause re­place­ment is un­fun­ded are all de­fens­ible choices. Un­defen­ded, they are find­ings. The dif­fer­ence is a dated de­cision with a name against it.

Technology Asset ManagementVoice of Volge21 •

Appendices

Methodology & editorial standards

This book fol­lows the same rules as its com­pan­ion, de­signed to keep a dis­closed-in­terest ana­lysis hon­est. Every ex­ternal fig­ure is at­trib­uted and dated. Vendor po­s­i­tion­ing is la­belled as po­s­i­tion­ing. Newly-shipped cap­ab­il­ity is la­belled as un­proven where it is un­proven. Il­lus­trat­ive fig­ures are la­belled as il­lus­trat­ive. Our com­mer­cial in­terest is dis­closed wherever it bears on the ar­gu­ment, and, in this book es­pe­cially, the proven­ance of our head­line fig­ures is stated plainly.

On the figures, and their source

Most of this book's num­bers come from Flex­era's own 2026 State of ITAM Re­port (24 June 2026, n=512). We flag this openly on page 16: it is a sur­vey of in­de­pend­ent prac­ti­tion­ers pro­duced by the vendor selling the rem­edy, and we treat it as dir­ec­tional rather than de­cis­ive. Un­like the com­pan­ion volume, we could not an­chor the owned-land­scape fig­ures to a fully in­de­pend­ent data house, and we would rather say so than imply an in­de­pend­ence we do not have.

On the structural argument

The case does not rest on the sur­vey. That vendors use audits as a rev­enue chan­nel, that li­cens­ing met­rics guar­an­tee an en­ti­tle­ment gap, that end-of-life soft­ware stops re­ceiv­ing patches, and that de­ployed AI is soft­ware to in­vent­ory, these are prop­er­ties of how the sys­tems are built, not sur­vey find­ings, and they carry the ar­gu­ment on their own.

On the illustrative exhibits

Ex­hibit 9, the three let­ters, is an il­lus­trat­ive com­pos­ite, not a meas­ure­ment of a named cli­ent. We quote no single head­line sav­ing, be­cause an hon­est one de­pends en­tirely on how un-gov­erned the start­ing land­scape was.

Our conflict, stated plainly

Volge is a Flex­era de­liv­ery part­ner and be­ne­fits if you buy. We have handled that con­flict by dis­clos­ing the proven­ance of our fig­ures, rest­ing the ar­gu­ment on struc­tural facts rather

than vendor claims, la­belling new cap­ab­il­ity as un­proven, and keep­ing the first step, a dis­cov­ery, small.

Corrections

If any fig­ure here has aged or any at­tri­bu­tion is wrong, we would rather know. This is a first edi­tion, and the AI-asset ma­ter­ial in par­tic­u­lar will have moved by the time you read it.

On independence: stated again

The head­line audit and AI fig­ures are Flex­era's own sur­vey re­search, la­belled as such through­out. We have not dressed them as third-party val­id­a­tion. The ar­gu­ment rests on the struc­tural mech­an­ics of audits, of vul­ner­ab­il­it­ies and of sup­port li­fe­cycles, which hold re­gard­less of any sur­vey.

The reader's checklist

Tear-out ver­sion. Six ques­tions from the op­er­ator's read (page 17), on a single page you can take into a meet­ing.

Checklist Before you spend on landscape governance

AskIf the answer is vague
Can we state our own licence position today?The next audit is on the vendor's numbers
Do we see third-party software, or only the OS?You are patching the wrong landscape
What is end-of-life right now?Obsolescence is found in incidents, not scheduled
Can we find the AI we have deployed?You cannot secure or retire what you cannot see
Do our three functions share one inventory?The single-landscape benefit is only theoretical
Who operates it after signing?The landscape view ages into shelfware

Source: Volge ana­lysis, 2026. See page 17.

  • Com­ply. The audit: know your po­s­i­tion be­fore the vendor sets the price
  • Se­cure. The breach: patch what is present and ex­ploit­able
  • Re­tire. The ex­piry: sched­ule ob­sol­es­cence, and find the AI
Technology Asset ManagementVoice of Volge22 •

What you need in place to see the landscape

Land­scape gov­ernance has fewer pre­requis­ites than any other layer in this series, and the gat­ing item is never the li­cence. Two lists fol­low. The first is what must be pro­cured. The second is what must already exist for a dis­cov­ery to pro­duce a de­fens­ible count.

Platform elements to licenceWhat each one is for
Discovery and normalisationAgent and agentless inventory, resolved against a normalised product catalogue. The foundation of every position downstream.
IT asset managementEntitlement capture and the effective licence position, product by product.
Vulnerability intelligenceValidated, scored advisories matched against installed inventory rather than against a product list.
Lifecycle dataSupport and end-of-life dates per product version, so obsolescence is scheduled rather than discovered.
AI cost and asset visibilityModels, agents and AI-enabled features brought onto the register. Newest, and the least proven.
What else must already be in placeWhy it gates the work
Scan authority across every segmentIncluding the segments operational technology owns. Unlisted exclusions become findings.
The contract archiveEntitlement that cannot be evidenced is entitlement you do not have.
Virtualisation topologyHost, cluster and affinity data, because the per-core publishers price the cluster.
A lawful basis for inventory dataHostnames and usernames are personal data, and their export is regulated.
A named owner per publisherA reconciliation with nobody behind it produces a report, not a position.

Source: Volge ana­lysis, 2026. Ele­ment names are gen­eric rather than a Flex­era SKU list; con­firm cur­rent pack­aging with the vendor.

Noth­ing on either list re­quires a data centre. The gat­ing item is ac­cess and au­thor­ity, and the hard­est of them to ob­tain is usu­ally the con­tract archive.

Technology Asset ManagementVoice of Volge23 •
Technology Asset Management

A dashboard that reports the health of the service, and not the age of the data, will tell you everything is fine on the day it stops knowing anything.

Reference

Glossary

Plain defin­i­tions of the terms used in this book, in the sense we have used them.

TermDefinition
The installed liabilityThis book's term for the standing exposure of the owned landscape, audit, breach, and obsolescence, plus the newest asset, deployed AI.
The owned landscapeSoftware and AI an organisation has installed and runs itself, as opposed to the rented cloud and SaaS of the companion book.
Flexera OneFlexera's platform brand, covering AI Cost Management, FinOps, SaaS Management and IT Asset Management.
ITAMIT Asset Management: the discipline of inventorying and governing owned hardware and software, and the name of Flexera's product for it.
SAMSoftware Asset Management: the licence-and-compliance branch of ITAM. A discipline rather than a Flexera SKU; Flexera's oldest strength, deepened by Snow (2024).
EntitlementWhat a contract permits you to run, by user, core, server, or other metric.
DeploymentWhat you have actually installed and are running, regardless of entitlement.
Effective licence position (ELP)The reconciliation of deployment against entitlement, product by product: the artefact an audit turns on.
True-upThe bill a vendor issues when an audit finds deployment exceeding entitlement.
IT VisibilityFlexera's discovery and normalisation capability, now grouped under IT Asset Management alongside Data Platform: turning inconsistent landscape data into one clean count.
TechnopediaFlexera's normalised catalogue of 5m+ hardware and software products, carrying identity and lifecycle data.
AI Cost ManagementFlexera's product for AI spend across agents, models, data and compute, launched 10 June 2026 and expanded 20 October 2026.
Software Vulnerability Manager (SVM)Flexera's platform for identifying, prioritising, and patching vulnerable software in the landscape.
Secunia ResearchThe team, active since 2002, that validates and scores vulnerability disclosures behind SVM.
Third-party softwareNon-OS applications, browsers, runtimes, readers, that carry most of the unpatched attack surface.
End of life (EOL)The point at which a vendor stops shipping security patches; every subsequent flaw is permanent.
Lifecycle dataThe record of support and end-of-life dates per product version, how obsolescence is scheduled, not discovered.
The AI you ownModels and agents deployed inside the landscape, to be inventoried, secured, and retired as assets. The subject of this book.
The AI you rentConsumption cost of models and agents, metered by the token. The subject of the companion book.
Technology Asset ManagementVoice of Volge24 •
Sources

Every ex­ternal claim in this book, with source and date. Lis­ted in order of first ap­pear­ance.

  • Flexera: 2026 State of ITAM Report (24 June 2026, n=512): 48% audited in the past year; 44% spent >$1M on software audits over three years, a share flat across the past three reports; Microsoft audits reported by 64% of respondents, Oracle up to 38% (from 24%), Adobe up to 32% (from 24%), all on a three-year window; audit response 22% of ITAM team time, behind software optimisation at 32%; complete IT asset visibility 36%, down from 43% in 2025.
  • Flexera: 2026 State of ITAM Report (AI): only 31% have accurate visibility into AI software, with no 2025 baseline for comparison; nearly half track AI as part of software spend; 59% of respondents report wasted AI spend rising year on year.
  • Flexera AI Cost Management: launched 10 June 2026 at FinOps X, covering agents, models, data and compute with token and credit tracking, alongside FinOps Assist; expanded 20 October 2026. Flexera's own capability matrix marks AI policy and governance as partial.
  • Flexera Secunia Research: Vulnerability intelligence since 2002; validates and scores disclosures via code audit and binary analysis; rejects illegitimate reports; criticality and attack-vector scoring.
  • Flexera Software Vulnerability Manager: Identifies and prioritises vulnerable software in the landscape; described by Flexera as backed by the largest repository of third-party patch data in the industry.
  • Flexera Technopedia: Normalised catalogue of 5m+ hardware and software products, carrying lifecycle / end-of-life data.
  • Flexera IT Asset Management: Effective licence position and audit response, within the Flexera One platform; depth extended by the Snow Software acquisition (completed February 2024). Software Asset Management is the discipline, not a Flexera SKU.
  • Flexera corporate history: Spun out of Macrovision as Acresso (2008), renamed Flexera Software (2009); Secunia vulnerability research active since 2002; Snow Software completed February 2024; Spot acquired from NetApp in 2025; ProsperOps and Chaos Genius announced 6 January 2026. Mike Jerich became president and chief executive on 16 July 2026, Jim Ryan moving to vice chairman. Majority-held by Thoma Bravo.
  • Publisher licensing mechanics: Oracle Java SE Universal Subscription, 23 January 2023, with flat small-business pricing from 8 October 2026; SAP indirect use priced by document, 10 April 2018, later named Digital Access; IBM sub-capacity conditional on ILMT, with Passport Advantage folding into the Client Relationship Agreement; Microsoft price and packaging update announced 4 December 2025, effective 1 July 2026; Broadcom-VMware close 22 November 2023, subscription-only and per-core from 11 December 2023; Salesforce list increase from 1 August 2025 with Agentforce replacing the Einstein editions; Adobe Creative Cloud Pro, 15 May 2025.
Technology Asset ManagementVoice of Volge25 •

Volge and our expertise on technology asset management

Volge counts what a cli­ent has in­stalled and then de­fends it, and the work is a prac­tice rather than a re­port. The dis­cip­line sits where this book loc­ates the li­ab­il­ity, in the dis­tance between en­ti­tle­ment and de­ploy­ment and in the ver­sions nobody has tracked. De­liv­ery is by senior asset and se­cur­ity prac­ti­tion­ers who run the re­con­cili­ation them­selves, in­side the cli­ent's own ten­ancy and tool­ing, through to a stand­ing li­cence po­s­i­tion that sur­vives the next audit let­ter.

The work has a usual shape. A dis­cov­ery pass pro­duces the raw count and meas­ures its own cov­er­age against the iden­tity dir­ect­ory, be­cause a po­s­i­tion built on an in­com­plete sweep is worse than no po­s­i­tion at all. Nor­m­al­isa­tion fol­lows, with the un­matched re­cords coun­ted rather than dis­carded. Re­con­cili­ation then pro­duces the ef­fect­ive li­cence po­s­i­tion pub­lisher by pub­lisher, with two po­s­i­tions re-de­rived by hand against the con­tracts, which is where con­fig­ur­a­tion er­rors sur­face. Re­medi­ation closes the under-li­censed ex­pos­ure and re­claims the over-li­censed, and gov­ernance sets the re­fresh ca­dence that keeps the po­s­i­tion cur­rent.

Who does the work. A de­ploy­ment strategist holds the com­mer­cial out­come and the pub­lisher re­la­tion­ships, in­clud­ing the ne­go­ti­ation when an audit let­ter ar­rives. A trans­form­a­tion ar­chi­tect owns the po­s­i­tion it­self, the met­ric con­fig­ur­a­tion and the evid­ence pack be­hind every en­ti­tle­ment. Se­cur­ity prac­ti­tion­ers own the vul­ner­ab­il­ity triage and the li­fe­cycle ho­ri­zon, so the patch list and the li­cence po­s­i­tion are drawn from one count rather than two. The same people stay from the first sweep through to the first audit de­fen­ded, which is what makes the con­trol tests in chapter four drills the team has run.

The re­gional po­s­i­tion is often what a Gulf or In­dian cli­ent needs first. Volge is headquartered in the Gulf, and the Bahraini pro­vi­sion set out in chapter five is treated as a se­cur­ity ob­lig­a­tion rather than a pro­cure­ment note, be­cause that is what the in­stru­ment makes it. Where no plat­form host­ing re­gion ex­ists in coun­try, which is every­where in this set, the op­tions are presen­ted with their legal con­sequences at­tached and the choice is re­cor­ded. Where a na­tional ac­cred­it­a­tion gate de­cides whether a plat­form may be sup­plied at all, as in

Saudi Ar­a­bia, Kuwait, Dubai and Oman, that gate is tested be­fore a li­cence is scoped rather than after.

An en­gage­ment starts with a scoped dis­cov­ery, usu­ally three to five weeks. It pro­duces a cov­er­age fig­ure with its ex­clu­sions lis­ted, a nor­m­al­ised count with the un­matched pile sized, a li­cence po­s­i­tion for the two or three pub­lish­ers car­ry­ing the most ex­pos­ure, a li­fe­cycle ho­ri­zon for the next four quar­ters, and a re­gister of the de­ployed AI nobody had re­cor­ded. Volge is a Flex­era de­liv­ery part­ner and be­ne­fits com­mer­cially if a cli­ent buys, which a reader should weigh against the re­com­mend­a­tions in this book. The con­ver­sa­tion be­gins wherever the land­scape stands today.

Technology Asset ManagementVoice of Volge26 •
Version · VVF-TAM-005 · 2026
volgeEnable Powerful
Transformations
About Volge

Volge is an engineering firm providing enterprise technology transformation services to clients across the Middle East, India, APAC, Europe, and North America.

Delivery stands on two pillars: Forward Deployed Engineering, with senior specialists embedded in the client’s own team, and Intelligent Command Centre Operations, run from our Global Delivery Centres, which keeps what they build running. Senior engineers stay with an engagement from the first diagnosis through to production.

Our services span Strategy, Data, Platforms, AI and Automation, Cyber Trust, Technology Operations, and Applied Functional Engineering. Engagements begin with a diagnosis. The firm is built on working autonomy: engineers decide, innovate, co-create, and own their work. Motivated engineers build the systems that sustain our clients’ growth.

Terms of use

This edition may be shared with organisations evaluating a Volge engagement. It may not be altered or republished without Volge’s permission, and figures carry their stated basis and date. Volge, Intelligent Command Centre, and Intelligent Command Centre Operations are marks of Volge; other names are the trademarks of their respective owners.