>
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.
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
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.
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.
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.
| See | Data Platform · IT Visibility · Technopedia · hardware & software inventory |
|---|---|
| Comply | IT Asset Management · Effective Licence Position · audit response |
| Secure & retire | Software Vulnerability Manager (Secunia) · lifecycle / end-of-life data |
Source: Flexera 2026 State of ITAM Report, 24 June 2026 (n=512). Figures describe the landscape, not Flexera outcomes.
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
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 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.
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.
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
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.
| Liability | Called by | Arrives as | Flexera instrument |
|---|---|---|---|
| Comply | The software vendor | An audit and a true-up bill | IT Asset Management |
| Secure | An attacker | An exploited unpatched flaw | Software Vulnerability Manager |
| Retire | The support calendar | An unsupported, un-patchable system | Technopedia lifecycle data |
Source: Volge analysis, 2026. Flexera product mapping per Flexera materials, 2026.
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.
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.
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.
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.
| Position | What it means | The action |
|---|---|---|
| Under-licensed | Deployed more than entitled | Close before the auditor finds it |
| Correctly licensed | Deployment matches entitlement | Maintain the evidence |
| Over-licensed | Entitled to more than deployed | Reclaim at renewal: money the vendor will not volunteer |
Source: Volge analysis, 2026. Capability mapping per Flexera IT Asset Management materials, 2024–26.
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.
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.
| Publisher | The mechanism | What it prices | Since |
|---|---|---|---|
| Oracle | Java SE Universal Subscription: per-employee metric | Every employee, not every user of Java | 23 January 2023 |
| SAP | Document-based indirect use, later named Digital Access | Documents created in SAP by non-SAP systems | 10 April 2018 |
| IBM | Sub-capacity PVU, conditional on ILMT reporting | The full cluster, wherever ILMT is absent | Client Relationship Agreement terms |
| Microsoft | Tier mix, CSP billing, Azure Hybrid Benefit | The gap between tiers bought and features used | Continuous; 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.
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.
| Week | The work | What must exist at the end of it |
|---|---|---|
| 1 | Read the audit clause; confirm entity and product scope; acknowledge receipt without conceding method | A written statement of scope, sent, with the clause cited |
| 2 | Freeze deployment change in scope; assemble the contract archive for the named products | An entitlement file with a document behind every line |
| 3 | Produce the internal position for the named products, identifying over-licensed as well as under-licensed holdings | A position the organisation would defend in writing |
| 4 | Agree the data collection method and the confidentiality terms; exchange the first data set | A 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
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.
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.
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.
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.
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.
Source: Volge analysis, 2026. Capability per Flexera Software Vulnerability Manager / Secunia Research materials.
A vulnerability programme is bounded by change windows rather than by intelligence. The advisory arrives in minutes.
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.
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.
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.
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.
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.
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.
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
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.
| State | What it means | The liability |
|---|---|---|
| Supported | Patches still shipping | Normal: keep current |
| End of support approaching | Expiry dated and known | Budget the replacement now |
| End of life | No further security patches | Every new flaw is permanent: retire or isolate |
Source: Volge analysis, 2026. Lifecycle data per Flexera Technopedia.
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.
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.
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.
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.
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.
Volge: the companion premise to The technology tax
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
| Field | Why it is on the register |
|---|---|
| Model and version | The thing that was actually invoked, which is often not the thing that was approved. |
| Endpoint and hosting location | Determines the residency position and whether the call left the jurisdiction. |
| Licence basis and output terms | Whether the output may be used commercially, and who owns it. |
| Named owner | An agent with no owner is a shadow deployment with a budget code. |
| Lifecycle status | Models are deprecated and withdrawn on vendor timetables shorter than any software support cycle. |
| Dependency manifest | The 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.
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.
| Option | What it buys | What it does not |
|---|---|---|
| Replace | Supported software and a current patch route | Nothing, if the replacement inherits the same unmanaged position |
| Extended support, where offered | A patch route for a stated period at a stated price | A permanent answer; the expiry still arrives |
| Isolate | A reduced blast radius for flaws that can never be closed | Protection against a flaw reachable through the paths you left open |
| Reduce capability | A smaller attack surface on the same software | Anything, unless the withdrawn features were genuinely unused |
| Accept | Time, and a clear record of the choice | Any defence at all, if the acceptance is undated and unowned |
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.
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.
| Stage | What it produces | The exit test |
|---|---|---|
| 01 Discover | A raw inventory of hosts, devices, tenancies and installed software | Reconcile 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 Normalise | One name per product, edition and version, matched to a catalogue | Sample 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 Reconcile | An effective licence position, product by product | Re-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 Remediate | Closed under-licensed exposure and reclaimed shelfware | Each closed position carries a dated evidence pack a third party could follow without asking a question. |
| 05 Govern | A standing position that survives change | Re-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.
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.
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.
| Control | What it is | How you test it |
|---|---|---|
| Discovery coverage | Every 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 normalisation | Raw 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 evidence | Every 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 configuration | Each 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 capture | Host and cluster topology is current for every virtualised workload. | Compare the hypervisor inventory against the asset platform's host records monthly. |
| Vulnerability matching | Validated 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 horizon | Every 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 detection | Software 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 registration | Deployed 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. |
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
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
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.
| Measure | Definition | Direction | Cadence |
|---|---|---|---|
| Discovery coverage | In-scope hosts and tenancies swept successfully in the period, against the authoritative device count | Rising to a stated floor | Monthly |
| Unmatched rate | Inventory records that resolve to no catalogue identity, as a share of records | Falling | Monthly |
| Position age | Days since the effective licence position was last re-derived, per publisher | Flat and low | Monthly |
| Exploited-flaw remediation time | Median days from validated advisory to patched host, for advisories with known exploitation | Falling | Monthly |
| Lifecycle exposure | Installed instances already past end of support, and instances expiring within four quarters | Falling, and scheduled | Quarterly |
| Unregistered AI | Models, agents and AI-enabled features found in the landscape with no owner or licence basis | Falling | Quarterly |
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.
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
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.
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.
The provision is unusual enough to be worth quoting in full in any Bahraini board paper
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 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
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.
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 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.
| Jurisdiction | Flexera One hosting in region | The binding hook on software assets | The supplier gate |
|---|---|---|---|
| Bahrain | None | Ministerial Order 43 of 2022, Article 2(3): licence compliance is a required breach-mitigation measure, criminally backed through the PDPL | No national accreditation scheme; international certifications specified instead |
| Oman | None | MTCIT Summary 20/2025 on software governance in government units, with central Microsoft and Oracle agreements at Summary 14/2025 and 17/2025 | TRA authorisation before contracting, plus classification before migration |
| Saudi Arabia | None | No licence-specific instrument identified; the constraint is the prohibition on exporting government agency data | CST registration at Class A, B or C by data classification |
| United Arab Emirates | None | Federal PDPL in force since 2 January 2022 with no Executive Regulations as at 8 November 2026; health data cannot leave the State | DESC Cloud Service Provider Security Standard for Dubai government and semi-government |
| Kuwait | None | CITRA cloud framework prohibits hosting Level 3 and Level 4 data and government-held personal data outside Kuwait | CITRA cloud licence, requiring a data centre inside Kuwait |
| India | None; nearest region is Australia | SEBI CSCRF GV.SC.S5 mandates an SBOM for new software procurements of core and critical activities | MeitY empanelment and STQC audit status for SEBI regulated entities under the 2023 cloud circular |
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.
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.
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.
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 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
hosted in Europe or Australia is overseas outsourcing on those terms.
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 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.
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.
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.
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.
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.
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.
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.
Source: Volge analysis, 2026. The value is conditional on the pieces sharing one inventory. Verify with your data.
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.
| Finding | Figure | Source |
|---|---|---|
| Organisations audited in the past year | 48% | Flexera ITAM, 24 Jun 2026 |
| Spent >$1M on software audits over three years | 44% | Flexera ITAM, 24 Jun 2026 |
| Respondents reporting a Microsoft audit in three years | 64% | Flexera ITAM, 24 Jun 2026 |
| Complete IT asset visibility (43% in 2025) | 36% | Flexera ITAM, 24 Jun 2026 |
| Accurate visibility into AI software | 31% | Flexera ITAM, 24 Jun 2026 |
| Respondents reporting wasted AI spend rising | 59% | 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.
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.
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.
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.
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.
Each objection assumes the purchase is the software. It is not. The purchase is a seen, reconciled, continuously governed
landscape, and the discipline to discharge its liabilities before they are called. The software is a means, and the governance is the thing.
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.
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.
| Letter | Un-governed response | One-landscape response |
|---|---|---|
| Oracle audit | Weeks assembling a count; accept the finding | Licence position already current; contest the inflated figure |
| Critical CVE | Is it even installed? Where? Unknown | Matched to inventory in hours; patch the exposed hosts first |
| End-of-life notice | Discovered late; emergency migration | Already on the lifecycle schedule; replacement budgeted |
Source: illustrative composite by Volge, 2026, of patterns commonly seen. Not a specific client benchmark.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
Tear-out version. Six questions from the operator's read (page 17), on a single page you can take into a meeting.
| Ask | If 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.
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 licence | What each one is for |
|---|---|
| Discovery and normalisation | Agent and agentless inventory, resolved against a normalised product catalogue. The foundation of every position downstream. |
| IT asset management | Entitlement capture and the effective licence position, product by product. |
| Vulnerability intelligence | Validated, scored advisories matched against installed inventory rather than against a product list. |
| Lifecycle data | Support and end-of-life dates per product version, so obsolescence is scheduled rather than discovered. |
| AI cost and asset visibility | Models, agents and AI-enabled features brought onto the register. Newest, and the least proven. |
| What else must already be in place | Why it gates the work |
|---|---|
| Scan authority across every segment | Including the segments operational technology owns. Unlisted exclusions become findings. |
| The contract archive | Entitlement that cannot be evidenced is entitlement you do not have. |
| Virtualisation topology | Host, cluster and affinity data, because the per-core publishers price the cluster. |
| A lawful basis for inventory data | Hostnames and usernames are personal data, and their export is regulated. |
| A named owner per publisher | A 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.
Plain definitions of the terms used in this book, in the sense we have used them.
| Term | Definition |
|---|---|
| The installed liability | This book's term for the standing exposure of the owned landscape, audit, breach, and obsolescence, plus the newest asset, deployed AI. |
| The owned landscape | Software and AI an organisation has installed and runs itself, as opposed to the rented cloud and SaaS of the companion book. |
| Flexera One | Flexera's platform brand, covering AI Cost Management, FinOps, SaaS Management and IT Asset Management. |
| ITAM | IT Asset Management: the discipline of inventorying and governing owned hardware and software, and the name of Flexera's product for it. |
| SAM | Software Asset Management: the licence-and-compliance branch of ITAM. A discipline rather than a Flexera SKU; Flexera's oldest strength, deepened by Snow (2024). |
| Entitlement | What a contract permits you to run, by user, core, server, or other metric. |
| Deployment | What 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-up | The bill a vendor issues when an audit finds deployment exceeding entitlement. |
| IT Visibility | Flexera's discovery and normalisation capability, now grouped under IT Asset Management alongside Data Platform: turning inconsistent landscape data into one clean count. |
| Technopedia | Flexera's normalised catalogue of 5m+ hardware and software products, carrying identity and lifecycle data. |
| AI Cost Management | Flexera'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 Research | The team, active since 2002, that validates and scores vulnerability disclosures behind SVM. |
| Third-party software | Non-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 data | The record of support and end-of-life dates per product version, how obsolescence is scheduled, not discovered. |
| The AI you own | Models and agents deployed inside the landscape, to be inventoried, secured, and retired as assets. The subject of this book. |
| The AI you rent | Consumption cost of models and agents, metered by the token. The subject of the companion book. |
Every external claim in this book, with source and date. Listed in order of first appearance.
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.
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.
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.