>
volgeNintex · 2027
PROCESS INTELLIGENCE · VOICE OF VOLGE

Your processes in high definition

Automate the work, not the flowchart. Most organisations automate processes they have never truly seen. Map the designed process, mine the real one, and the gap between them is the finding.

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.

Automate the work, not the flowchart. Most organisations automate processes they have never truly seen. Map the designed process, mine the real one, and the gap between them is the finding.

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

Contents

  1. 02You cannot automate a process you cannot see.
  2. 02About Nintex
  3. 03You cannot automate what you cannot see.
  4. 03The invisible-process problem
  5. 06The map
  6. 08The mine
  7. 12Closing the gap
  8. 14Process intelligence in the Gulf and India
  9. 18The operator's read
  10. 22Appendices
Voice of Volge is a series of books, and the firm’s thought leadership on the solutions and the domains Volge specialises in.
Nintex · 2027
Process IntelligenceVoice of Volge01 •

You cannot automate a process you cannot see.

Every auto­ma­tion pro­gramme be­gins with the same op­tim­ism and ends with the same sur­prise. The op­tim­ism is that soft­ware, and now AI, will take work off people’s hands. The sur­prise is how little the or­gan­isa­tion ac­tu­ally knew about the work it was auto­mat­ing. A pro­cess is not what the policy doc­u­ment de­scribes, nor what the pro­cess owner re­mem­bers; it is what the event logs re­cord, and the three rarely agree.

The cost of that gap is not ab­stract. Auto­mate a broken pro­cess and you in­dus­tri­al­ise the break­age, faster and at scale, with an audit trail that now proves you did it on pur­pose. A large share of di­git­isa­tion budgets goes ex­actly this way, en­cod­ing the cur­rent mess rather than the in­ten­ded design. The least glam­or­ous step in the whole pro­gramme, see­ing the real pro­cess be­fore touch­ing it, is also the one that de­cides whether the rest pays back.

This re­port is not a tools cata­logue. It is an ar­gu­ment about se­quence. Un­der­stand the work, then build on it. We set out why pro­cess vis­ib­il­ity is the pre­con­di­tion for everything that fol­lows, what ‘high defin­i­tion’ means when the evid­ence is event logs rather than opin­ion, and how to read the dis­tance between the pro­cess you de­signed and the one you ac­tu­ally run. The fig­ures are sourced and dated; the ex­amples are the or­din­ary kind, not the demo kind.

Where a spe­cific plat­form earns a men­tion, here, Nin­tex’s pro­cess tool­ing, it is to make the ar­gu­ment con­crete, not be­cause the ar­gu­ment de­pends on it. Volge is a Nin­tex de­liv­ery part­ner, and says so plainly. The claim stands on its own. See the work in high defin­i­tion, then auto­mate.

The Voice of Volge desk

About Nintex

Nin­tex is a pro­cess-and-auto­ma­tion soft­ware com­pany. What began as SharePoint-era work­flow is now a single plat­form that doc­u­ments how work runs and auto­mates it, and, since 2025, or­ches­trates AI agents against it. It was built out over two dec­ades and a hand­ful of poin­ted ac­quis­i­tions.

  • 2006. Foun­ded: Mel­bourne, Aus­tralia; now headquartered in Bel­levue, Wash­ing­ton
  • 7,000+. Or­gan­isa­tions: across 100+ coun­tries (Nin­tex com­pany page, Septem­ber 2026)
  • 213%. For­res­ter TEI ROI: ~9-month pay­back, Nin­tex Work­flow (study con­duc­ted late 2022)

The re­cent chapters mat­ter most to a buyer today: Agen­tic Busi­ness Or­ches­tra­tion (Septem­ber 2025), loc­ally-hos­ted AI in

Auto­ma­tion K2 5.9.1 (April 2026), Agent De­signer and Nin­tex Or­ches­tra­tion now ship­ping in­side the Nin­tex CE Plat­form, and Nin­tex Pro­cesses, launched on 26 Oc­to­ber 2026. Nin­tex Pro­cess Dis­cov­ery was placed end-of-life on 4 May 2026.

Exhibit A Two decades to an agentic platform: 2006 → 2026

  • 2006. Foun­ded; SharePoint-era work­flow
  • 2020 · K2. On-prem work­flow & forms depth
  • 2022 · Kryon. RPA for leg­acy screens
  • 2025 · ABO. De­term­in­istic + AI, human in the loop
  • 2026 · Nin­tex Pro­cesses. CE Plat­form; Pro­cess Dis­cov­ery end-of-life

2025–26

Source: Nin­tex cor­por­ate and product ma­ter­i­als, 2006–2026. Nin­tex is ma­jor­ity-held by TPG Cap­ital (2021), with Thoma Bravo a minor­ity holder. Rat­ings are G2 4.2/5 and TrustRadius 8/10, read Novem­ber 2026.

The platform today: one landscape

This re­port cov­ers the Un­der­stand layer, where Nin­tex doc­u­ments how work is meant to run and sup­plies the evid­ence to test that against how it ac­tu­ally runs.

Understand Focus of this reportProcess Manager · Nintex Processes · Process Capture
Build & runAutomation CE · Automation K2 (on-prem) · RPA · DocGen · Forms · eSign
OrchestrateAgentic Business Orchestration · Agent Designer · Nintex Orchestration
  • 4.2 /5. G2 across 1,326 re­views · TrustRadius 8/10 across 530
  • 7,000+. Or­gan­isa­tions across 100+ coun­tries. Nin­tex com­pany page, Septem­ber 2026
  • Strong Per­former. For­res­ter Wave: Agen­tic Pro­cess Or­ches­tra­tion Plat­forms, Q3 2026

Source: Nin­tex com­pany page, Septem­ber 2026; G2 and TrustRadius, read Novem­ber 2026; For­res­ter Wave, Q3 2026.

What this means for a buyer in 2026

This is not a young product. It is a two-dec­ade plat­form that has just added gov­erned AI and a loc­ally-hos­ted en­gine, ma­ture where it must be, cur­rent where it counts, and now built to or­ches­trate agents at scale.

Every large or­gan­isa­tion runs two ver­sions of every pro­cess, and only one of them is writ­ten down. There is the pro­cess as

Process IntelligenceVoice of Volge02 •

de­signed, the flow­chart in the qual­ity manual, the swim­lane in the on­board­ing deck, the story the head of op­er­a­tions tells the board. And there is the pro­cess as per­formed, with its re­worked ap­provals and its shadow spread­sheet, and the ex­cep­tion that quietly be­came the rule. Auto­ma­tion and AI both act on the first. Value hides in the second.

You cannot automate what you cannot see.

Before the platform, before the agent, before the budget: an honest account of how the work actually runs. Seeing it is the first deliverable, and everything else depends on it.

Every process exists twice.

The pro­cess in the manual and the pro­cess that ac­tu­ally hap­pens are rarely the same doc­u­ment. The dis­tance between them is the sub­ject of this book.

This mat­ters now be­cause the tools have out­run the un­der­stand­ing. An or­gan­isa­tion can buy an AI agent, a ro­botic pro­cess auto­ma­tion li­cence, and a work­flow en­gine in an af­ter­noon. What it can­not buy in an af­ter­noon is an hon­est ac­count of how its own work ac­tu­ally moves, which step waits three days for a sig­na­ture, which in­voice route the fin­ance team aban­doned in 2021, which "stand­ard" pro­ced­ure has el­even live vari­ants across four re­gions.

Point auto­ma­tion at the de­signed map and you auto­mate a fic­tion. The bot handles the happy path and falls over on the ex­cep­tions that make up most of the real volume. Point an AI agent at the same fic­tion and the fail­ure is quieter and more ex­pens­ive. The agent acts con­fid­ently on a pro­cess that no longer ex­ists, and no one no­tices until the num­bers drift.

The in­tel­li­gent-pro­cess-auto­ma­tion mar­ket is priced as if this prob­lem were already solved. Mar­ket.us es­tim­ates the cat­egory at roughly 16.81 bil­lion US dol­lars in 2024, with a pro­jec­ted 61.23 bil­lion by 2034 at a com­pound an­nual growth rate of 13.8 per cent (Mar­ket.us es­tim­ate, Septem­ber 2025 edi­tion). Firms are spend­ing. The ques­tion this book asks is nar­rower and more use­ful. Are they spend­ing against real­ity, or against the map?

A process you cannot see is a process you cannot automate, cannot govern, and cannot hand to an agent.

Volge is a Nin­tex de­liv­ery part­ner and be­ne­fits if a cli­ent buys, and that com­mer­cial in­terest is marked wherever it ap­pears in this book. The ar­gu­ment it­self does not de­pend on it. You

need the map, you need the mine, and you need the dis­cip­line to work the dif­fer­ence between them.

Nin­tex an­swers half of that ques­tion dir­ectly, and the other half has just changed shape. Nin­tex Pro­cess Man­ager holds the pro­cess as de­signed, the map, and since 26 Oc­to­ber 2026 it sur­faces as Nin­tex Pro­cesses in­side the Nin­tex CE Plat­form. The mine, the pro­cess as per­formed, was Nin­tex Pro­cess Dis­cov­ery, which Nin­tex placed end-of-life on 4 May 2026, with sup­port end­ing in May 2028 and cus­tom­ers mi­grated to Pro­cess Man­ager. The gap between in­ten­tion and real­ity is still the work­list. What has changed is who holds the in­stru­ment that meas­ures it.

Read on for the prob­lem, then the sur­viv­ing in­stru­ment, then the bridge Nin­tex shipped in its Decem­ber 2025 re­lease that turns a tech­nical pro­cess model into a map a busi­ness per­son can read, and the MCP server that now hands that map to an AI as­sist­ant.

Volge: the work­ing premise of this book

The invisible-process problem

Automation and AI fail without process truth, and the money already committed says many firms are about to learn this the expensive way.

Automation applied to a process nobody understands scales the error, not the work

Auto­ma­tion is a mul­ti­plier, and a mul­ti­plier is in­dif­fer­ent to the sign of the num­ber it mul­ti­plies. Ap­plied to a clean pro­cess, doc­u­mented and stable, it re­moves cost and delay. Ap­plied to one that is none of those things, it in­dus­tri­al­ises the con­fu­sion. The tech­no­logy is neut­ral; the un­der­stand­ing is not.

Con­sider the or­din­ary shape of a failed auto­ma­tion pro­gramme. A team picks a pro­cess that is pain­ful and vis­ible, say sup­plier on­board­ing. They doc­u­ment it by ask­ing the people who run it what they do, and those people de­scribe the map, the of­fi­cial steps in the of­fi­cial order. The team auto­mates the map. Then live volume ar­rives, car­ry­ing the forty per cent of cases the map never men­tioned, and the auto­ma­tion either breaks or, worse, pro­cesses them wrongly and si­lently.

The pat­tern re­peats be­cause the dia­gnosis is skipped. In­ter­views cap­ture what people be­lieve they do, which is a poor guide to what they ac­tu­ally do. Memory smooths over the re­runs, the work­arounds, and the in­formal es­cal­a­tions that keep the busi­ness mov­ing. The de­signed pro­cess is a sin­cere ac­count. It is also, routinely, wrong in the spe­cific places that cost the most money.

Process IntelligenceVoice of Volge03 •

AI raises the stakes. A rule-based bot fails loudly, it stops, throws an error, waits for a human. An AI agent is built to keep going. Give it an in­com­plete pic­ture of a pro­cess and it will im­pro­vise a plaus­ible com­ple­tion, act­ing on steps that were re­tired, ap­provers who have left, and thresholds that changed two budgets ago. The fail­ure mode is not a crash. It is con­fid­ent, well-format­ted, and wrong.

This is why pro­cess un­der­stand­ing has be­come the entry point for AI rather than an af­ter­thought to it. Nir­an­jan Vi­jayaragavan, then Nin­tex's chief product and tech­no­logy of­ficer, made the case under that head­ing in For­bes on 3 Feb­ru­ary 2026: “The AI entry point no one is talk­ing about: Pro­cess in­tel­li­gence”. He has since left; Five9 ap­poin­ted him chief tech­no­logy of­ficer ef­fect­ive 29 June 2026. An agent is only as good as its model of the work. Give it the de­signed map and it in­her­its the map's blind spots; meas­ure the work as per­formed and it in­her­its real­ity.

None of this ar­gues against auto­ma­tion. It ar­gues against auto­mat­ing in the dark. The se­quence that works is dull and re­li­able. See the pro­cess as de­signed, meas­ure it as per­formed, re­solve the dif­fer­ence, and only then de­cide what to auto­mate, what to sim­plify, and what to leave alone.

Exhibit 1 The two processes every organisation owns

DimensionProcess as designed: the mapProcess as performed: the mine
SourceInterviews, manuals, workshops, policyEvent logs from ERP, CRM, and line systems
ShowsHow work is meant to runHow work actually ran, case by case
BiasOptimistic, tidy, singularComplete, messy, many-variant
Nintex instrumentProcess Manager, now Nintex ProcessesProcess Capture at the desktop; mining no longer sold
Fails whenYou automate it as if it were trueYou never look at it at all

Source: Volge ana­lysis, 2026. Nin­tex in­stru­ments per Nin­tex product pages and re­lease notes, read 8 Novem­ber 2026.

The money is already committed; the understanding is the part that lags

Spend­ing on in­tel­li­gent pro­cess auto­ma­tion is pro­jec­ted to nearly quad­ruple in a dec­ade, which means the de­mand for tools has already been de­cided and the de­mand for pro­cess truth is the con­straint that re­mains. Mar­ket.us es­tim­ates the cat­egory at 16.81 bil­lion US dol­lars in 2024, with a pro­jec­ted 61.23 bil­lion by 2034, a com­pound an­nual growth rate of 13.8 per cent (Mar­ket.us es­tim­ate, Septem­ber 2025 edi­tion). Treat

the exact fig­ure as a dir­ec­tional es­tim­ate. Treat the dir­ec­tion as settled.

Exhibit 2 Intelligent process automation market, 2024 to 2034 (US dollars, estimate)

  • 2024. $16.81B
  • 2027 (in­ter­pol­ated). ≈ $24.7B
  • 2030 (in­ter­pol­ated). ≈ $36.4B
  • 2034. $61.23B

In­ter­pol­ated years as­sume the stated 13.8% CAGR and are il­lus­trat­ive only.

Source: Mar­ket.us es­tim­ate, Septem­ber 2025 edi­tion. En­d­points re­por­ted; in­ter­me­di­ate years in­ter­pol­ated by Volge at the stated CAGR.

Rising spend on auto­ma­tion does not, by it­self, buy res­ults. It buys ca­pa­city. Whether that ca­pa­city lands on real bot­tle­necks or ima­gin­ary ones de­pends en­tirely on the qual­ity of the pro­cess pic­ture feed­ing it. A firm can be a heavy, soph­ist­ic­ated buyer of auto­ma­tion and still be auto­mat­ing the wrong forty per cent, be­cause the pic­ture it is work­ing from is the tidy map rather than the messy mine.

The cor­rect­ive is cheap re­l­at­ive to the spend it gov­erns. Pro­cess min­ing and pro­cess map­ping are small line items next to an en­ter­prise auto­ma­tion pro­gramme, and they de­term­ine whether that pro­gramme's budget is aimed at the steps that ac­tu­ally cost time and money. Skip­ping the dia­gnosis to save on the dia­gnosis is the ex­pens­ive eco­nomy.

There is a second reason the un­der­stand­ing lags. Map­ping felt, for years, like an ad­min­is­trat­ive chore, a com­pli­ance arte­fact pro­duced once and left to rot. That fram­ing is now ob­sol­ete. A liv­ing pro­cess pic­ture is the sub­strate every auto­ma­tion and every agent reads from. It has moved from fil­ing cab­inet to op­er­at­ing de­pend­ency.

Nin­tex sits on the un­der­stand­ing side of this mar­ket rather than only the ex­e­cu­tion side, which is the ar­gu­ment of the next three parts. First the map. Then the mine. Then the bridge between them.

  • $16.81B. IPA mar­ket size, 2024. Mar­ket.us, Sep 2025
  • $61.23B. Pro­jec­ted mar­ket, 2034. Mar­ket.us, Sep 2025
  • 13.8%. Com­pound an­nual growth rate. Mar­ket.us, Sep 2025

Three places automation breaks, all of them invisible on the map

Auto­ma­tion pro­grammes fail in a small num­ber of pre­dict­able ways, and each is in­vis­ible to any­one work­ing from the

Process IntelligenceVoice of Volge04 •

de­signed pro­cess alone. Name the three and you have a check­list for why an ini­ti­at­ive un­der­delivered; see them in your own data and you have a work­list.

01 Un­doc­u­mented vari­ants

A pro­cess de­scribed as one path is usu­ally many. Re­gions adapt, teams im­pro­vise, and a single "stand­ard" pur­chase-to-pay quietly grows a dozen live vari­ants no manual re­cords. Auto­mate the doc­u­mented path and every un­doc­u­mented vari­ant either breaks the bot or by­passes it, which is how a shiny pro­gramme ends up cov­er­ing a minor­ity of real volume.

02 Hid­den bot­tle­necks

Delay con­cen­trates. In most pro­cesses a few steps hold most of the wait­ing, an ap­proval that queues for days, a han­doff between two sys­tems no one owns, a re­work loop triggered by bad up­stream data. These do not show up on a swim­lane dia­gram, which has no axis for time. They show up in event logs.

03 Si­lent de­vi­ations

The costly de­vi­ations are the ones nobody flags. A con­trol skipped "just this once" that be­came routine. An ap­proval threshold every­one treats as ad­vis­ory. A step per­formed out of order be­cause the of­fi­cial order was in­con­veni­ent. Each is de­fens­ible alone and dan­ger­ous in ag­greg­ate, and each is in­vis­ible until you com­pare the mine against the map.

Exhibit 3 Failure mode, symptom, and the instrument that surfaces it

Failure modeWhat the board seesWhat actually happenedSurfaced by
Undocumented variantsAutomation covers less than promisedReal work runs on paths the map omitsEvent-log analysis
Hidden bottlenecksCycle time refuses to fallA few steps hold most of the waitEvent-log analysis
Silent deviationsAn audit finding, eventuallyControls quietly skipped at scaleMap vs mine comparison

Source: Volge ana­lysis, 2026.

The uncomfortable arithmetic

If the doc­u­mented path car­ries 60 per cent of volume and your auto­ma­tion cov­ers only the doc­u­mented path, your ceil­ing is 60 per cent be­fore you have writ­ten a line of con­fig­ur­a­tion. The other 40 per cent is not an edge case. It is the case you did not meas­ure.

The dis­tinc­tion that mat­ters is between fail­ing loudly and fail­ing quietly. Tra­di­tional auto­ma­tion is brittle. It stops when real­ity de­parts from its script, and a human no­tices. That brit­tle­ness is an­noy­ing and also pro­tect­ive. It is a smoke alarm. An

agent, by design, does not stop. It reas­ons its way past the gap and pro­duces an an­swer, which may be right, may be wrong, and will in either case look equally pol­ished.

AI agents make the invisible-process problem worse before they make anything better

An AI agent in­her­its every blind spot in the pro­cess pic­ture you hand it, and then acts on those blind spots with con­fid­ence a rule-based sys­tem never had. This is the cent­ral shift of 2026. The or­gan­isa­tions rush­ing agents into op­er­a­tions are, in many cases, point­ing them at the de­signed map, the one doc­u­ment guar­an­teed to be in­com­plete, and call­ing the res­ult trans­form­a­tion.

So the value of a com­plete pro­cess pic­ture rises ex­actly as agents ar­rive. The agent needs to know the vari­ants and the ex­cep­tions, the con­trols that apply and the true se­quence of events. That know­ledge does not live in the manual. It lives in the event logs, wait­ing to be read. Meas­ur­ing the work as per­formed is there­fore the thing to do be­fore the agent, not after it.

This re­frames the min­ing cat­egory. Find­ing sav­ings in ex­ist­ing pro­cesses is part of it, but the lar­ger role is as a read­i­ness layer for any­thing you plan to hand to a ma­chine, the dif­fer­ence between de­ploy­ing an agent against your op­er­a­tions as they are and de­ploy­ing it against a flat­ter­ing de­scrip­tion of how they used to be.

Nin­tex's own fram­ing places pro­cess in­tel­li­gence at the front of the AI ques­tion rather than the back. We agree, with one caveat we will keep re­peat­ing. Min­ing tells you what happened, not why, and the "why" still needs people. The in­stru­ment sharpens the judge­ment. It does not re­place it.

The AI entry point no one is talking about: Process intelligence.

Title of an art­icle by Nir­an­jan Vi­jayaragavan, then Nin­tex chief product and tech­no­logy of­ficer, For­bes, 3 Feb­ru­ary 2026

A caveat we will not bury

Pro­cess min­ing re­con­structs the se­quence and tim­ing of work from data. It does not ex­plain in­tent. A step that looks like a de­vi­ation may be a sens­ible local fix, and a bot­tle­neck may be a de­lib­er­ate qual­ity gate. The in­stru­ment tells you where to look. De­cid­ing what it means is still a human job.

Process IntelligenceVoice of Volge05 •

Diagnosis first, then automation: the order that saves the budget

The se­quence below is de­lib­er­ately un­glam­or­ous, and its whole value is that it re­fuses to auto­mate be­fore it un­der­stands. Each step nar­rows the field of what is worth auto­mat­ing, so that the ex­pens­ive ex­e­cu­tion budget lands on the few steps that ac­tu­ally move cycle time and cost.

  • Dis­cover. Sur­face the pro­cesses that exist, of­fi­cial and shadow
  • Map. Doc­u­ment the design in busi­ness lan­guage
  • Mine. Re­con­struct the real­ity from event logs
  • Com­pare. Loc­ate the gap: vari­ants, waits, de­vi­ations
  • Act. Sim­plify, auto­mate, or leave alone: on evid­ence

Why the order is non-negotiable

Re­verse any two steps and the eco­nom­ics col­lapse. Auto­mate be­fore you mine and you scale the cur­rent mess. Mine be­fore you map and you have real­ity with noth­ing to com­pare it against. Act be­fore you com­pare and you are back to guess­ing, which is where most pro­grammes star­ted. The order is the product.

  • Map. Pro­cess Man­ager holds the de­signed pro­cess
  • Mine. Event logs re­con­struct the real one
  • Gap. The dif­fer­ence is the cos­ted work­list

What this part established

Auto­ma­tion mul­ti­plies whatever it is poin­ted at. Most firms point it at the map. The map is in­com­plete in the exact places that cost money, among them vari­ants and bot­tle­necks as well as out­right de­vi­ations, and agents make that worse. The fix is to see the pro­cess as per­formed be­fore you change it. Part II sup­plies the map. Part III sup­plies the mine.

Pro­cess Man­ager at­tacks that fail­ure by mov­ing own­er­ship to the front line. Maps are writ­ten in read­able steps, not ar­cane sym­bols, so the per­son who runs the pro­cess can read it, chal­lenge it, and own it. Feed­back and change re­quests sit in­side the pro­cess it­self. Own­er­ship is as­signed, so every pro­cess has a named human ac­count­able for keep­ing it hon­est. The res­ult is a map that stays close to the truth be­cause the people closest to the truth main­tain it.

Tra­di­tional pro­cess map­ping fails for a human reason, not a tech­nical one. The maps are drawn by spe­cial­ists in a nota­tion only spe­cial­ists read, stored some­where the doers never

look, and up­dated on a cycle meas­ured in years. So they drift from real­ity the mo­ment they are signed off, and every­one re­verts to tri­bal know­ledge. The dia­gram be­comes a com­pli­ance arte­fact, pro­duced and filed and then ig­nored.

The map

Nintex Process Manager, formerly Promapp, turns the process as designed into a living, business-readable document your whole organisation can own, in three verbs: identify, map, and model.

Nintex Process Manager turns the designed process into a document the business actually maintains

Nin­tex Pro­cess Man­ager ex­ists to make the pro­cess as de­signed a liv­ing asset rather than a dead dia­gram, writ­ten in plain busi­ness lan­guage, owned by the people who run the work, and kept cur­rent be­cause keep­ing it cur­rent is easy. Formerly Promapp, rebranded to Pro­cess Man­ager in 2022, it car­ries a de­lib­er­ately mod­est tagline, which is to identify, map and model your pro­cesses. The mod­esty is the point.

That read­ab­il­ity has a second pay­off. A map any­one can read is a map any­one can im­prove, which turns pro­cess doc­u­ment­a­tion from a pro­ject into a habit. New join­ers learn the real pro­ced­ure in­stead of shad­ow­ing a col­league, and aud­it­ors read the same doc­u­ment the op­er­at­ors use. Change be­comes vis­ible and ver­sioned rather than whispered.

"Model", the third verb in the tagline, mat­ters as much as the first two. A map re­cords the steps; a model cap­tures the logic and the de­cisions, the branches and the roles and the han­doffs, in a form pre­cise enough to reason about and to hand down­stream. That pre­ci­sion is what makes the later moves pos­sible. The model can be com­pared against mined real­ity, and the arte­facts that drive auto­ma­tion gen­er­ated from it.

Exhibit 4 Nintex Process Manager at a glance

AttributeDetail
What it isBusiness process mapping and modelling: the process as designed
Former namePromapp (rebranded to Nintex Process Manager, 2022)
Now surfacing asNintex Processes, launched 26 October 2026, native to the Nintex CE Platform
TaglineIdentify, map, and model your processes
Owned byFront-line process owners, in business language
Core valueA living, current, readable map rather than a filed diagram
Process IntelligenceVoice of Volge06 •

Source: Nin­tex product pages, read 8 Novem­ber 2026; rebrand con­firmed 2022; Nin­tex Pro­cesses launch re­por­ted 26 Oc­to­ber 2026.

Three verbs, three jobs, one living document

The tagline hides three dis­tinct jobs, and it is worth sep­ar­at­ing them be­cause or­gan­isa­tions tend to be good at one and poor at the oth­ers. Identi­fy­ing, map­ping, and mod­el­ling are not the same task, and a pro­cess pro­gramme that skips any of them leaves a pre­dict­able gap.

01 Identify: know­ing which pro­cesses you ac­tu­ally have

Most or­gan­isa­tions can­not name their own pro­cesses. Work is de­scribed by de­part­ment or by sys­tem or by pro­ject, and the pro­cess, the end-to-end thing the cus­tomer ex­per­i­ences, falls between those cat­egor­ies. Identi­fy­ing means build­ing the re­gister, the list of pro­cesses that exist, who owns each, and how they con­nect. Dull, found­a­tional, and usu­ally miss­ing.

02 Map: writ­ing down how it is meant to run

Map­ping re­cords the de­signed se­quence in read­able steps, nam­ing what hap­pens, in what order, by whom, with what in­puts and out­puts. Done in busi­ness lan­guage, the map be­comes us­able by the front line rather than filed by the qual­ity team. This is the arte­fact new join­ers learn from and aud­it­ors read.

03 Model, cap­tur­ing the logic, not just the steps

Mod­el­ling adds the de­cisions and branches, the roles and con­di­tions and han­doffs, in a struc­tured form. A model can be reasoned about, com­pared against mined real­ity, and handed to down­stream tools. It is the dif­fer­ence between a pic­ture of a pro­cess and a spe­cific­a­tion of one.

Exhibit 5 What each verb produces, and who reads it

VerbProducesPrimary readerFailure if skipped
IdentifyA process register with ownersOperations leadershipYou improve processes you cannot name
MapA readable procedureFront-line staff, auditorsKnowledge stays tribal
ModelStructured process logicAutomation and analysisNothing downstream to build on

Source: Volge ana­lysis of Nin­tex Pro­cess Man­ager, 2026.

The ownership rule

Every pro­cess in Pro­cess Man­ager has a named owner. This single dis­cip­line does more for ac­cur­acy than any fea­ture. A doc­u­ment with an ac­count­able human at­tached stays

hon­est; an own­er­less dia­gram rots. Volge con­fig­ures own­er­ship be­fore it con­fig­ures any­thing else.

Between March 2025 and December 2025, Nintex removed most of the manual labour from mapping

Three AI fea­tures shipped in­side ten months have changed the eco­nom­ics of get­ting a map on the page, be­cause each at­tacks the single most ex­pens­ive part of pro­cess doc­u­ment­a­tion, the human ef­fort of writ­ing it down. The old ob­jec­tion to map­ping was al­ways cost, since it took spe­cial­ists weeks. That ob­jec­tion is now much weaker. Ex­hibit 6 dates the re­leases; the para­graphs below ex­plain what each one does.

01 AI Pro­cess Gen­er­ator: from blank page to draft

The AI Pro­cess Gen­er­ator entered beta in the March 2025 re­lease and reached gen­eral avail­ab­il­ity in June 2025, for Pro­fes­sional tier and above. It drafts a pro­cess map from a de­scrip­tion, turn­ing the ter­ri­fy­ing blank page into an ed­it­able first draft. The draft is not the de­liv­er­able, a human owner still re­fines it, but start­ing from eighty per cent com­plete rather than zero changes who can par­ti­cip­ate and how fast.

02 Pro­cess Cap­ture: doc­u­ment­a­tion that re­cords it­self

Pro­cess Cap­ture reached gen­eral avail­ab­il­ity in the Septem­ber 2025 re­lease, in­cluded in Ex­pert and En­ter­prise tiers, an add-on else­where, and not offered in the AEN re­gion. It re­cords the ac­tual activ­ity of a task as it is per­formed and auto-doc­u­ments it, so the map is drawn from what someone did rather than from what they later re­membered doing. This closes part of the memory gap that makes in­ter­view-based map­ping un­re­li­able, the map in­her­its the doer's real clicks, not their ti­died re­col­lec­tion.

03 AI Model to Map: tech­nical logic made busi­ness-read­able

AI Model to Map, named AI Model to Map Con­verter in the product, reached gen­eral avail­ab­il­ity in the Decem­ber 2025 re­lease, Pro­cess Man­ager 4.12.27.0, hav­ing entered beta that Septem­ber. It con­verts tech­nical BPMN, the nota­tion ana­lysts and min­ing tools speak, into busi­ness-read­able pro­ced­ures that a pro­cess owner can ac­tu­ally use. This is the fea­ture that bridges the mine back to the map, and it earns a full treat­ment in Part IV be­cause it is the mech­an­ism that closes the loop.

Exhibit 6 Nintex Process Manager AI features: release timeline

  • March 2025: beta. AI Pro­cess Gen­er­ator: drafts a pro­cess map from a de­scrip­tion.
Process IntelligenceVoice of Volge07 •
  • June 2025: GA. AI Pro­cess Gen­er­ator, Pro­fes­sional tier and above.
  • Septem­ber 2025: GA. Pro­cess Cap­ture: activ­ity re­cord­ing that auto-doc­u­ments a task as per­formed.
  • Septem­ber 2025: beta. AI Model to Map Con­verter, Ex­pert and En­ter­prise tiers.
  • Decem­ber 2025: GA. AI Model to Map Con­verter: con­verts tech­nical BPMN into busi­ness-read­able pro­ced­ures.

Source: Nin­tex Pro­cess Man­ager re­lease notes, 4.12.10.0 to 4.12.27.0, read 8 Novem­ber 2026.

This is a major step forward in making process intelligence accessible, actionable, and scalable across the organization.

Chris Ellis, Dir­ector of Solu­tion En­gin­eer­ing, Nin­tex, on the gen­eral avail­ab­il­ity of AI Model to Map; re­por­ted by PEX Net­work, 6 Janu­ary 2026

What changed: mapping went from specialist project to shared habit

The prac­tical ef­fect of read­able maps plus AI draft­ing is that pro­cess doc­u­ment­a­tion stops being a pro­ject and be­comes a habit, which is the only state in which it stays ac­cur­ate. A pro­ject ends; a habit per­sists. Ex­hibit 7 con­trasts the old mode with the new one, and the dif­fer­ence is less about tech­no­logy than about who is al­lowed to par­ti­cip­ate.

Exhibit 7 Process documentation: before and after

DimensionTraditional mappingNintex Process Manager
AuthorProcess specialistsFront-line owners, AI-assisted
LanguageFormal notationPlain business steps
Starting pointBlank pageAI-generated draft
Update cadenceEvery few yearsContinuous, in-flow
Where it livesA shared drive, unreadThe tool people work in
Accuracy over timeDecaysMaintained by owners

Source: Volge ana­lysis, 2026.

The map is ne­ces­sary and, on its own, in­suf­fi­cient. How­ever cur­rent and read­able, it re­mains a re­cord of in­ten­tion. It tells you how the work is meant to run, and, be­cause own­ers keep it hon­est, it tells you that bet­ter than most or­gan­isa­tions man­age today. What it can­not tell you is whether the work ac­tu­ally runs that way in the mil­lions of cases nobody watched.

That is not a cri­ti­cism of Pro­cess Man­ager; it is a de­scrip­tion of what a map is. A map of a city does not show you the traffic. For the traffic you need a dif­fer­ent in­stru­ment, one that reads the re­cord of what ac­tu­ally moved. A map with no mine is a plan with no feed­back, and the mine is the sub­ject of Part III.

What this part established

Pro­cess Man­ager makes the de­signed pro­cess a liv­ing, owned, busi­ness-read­able doc­u­ment, and re­cent AI fea­tures have cut the cost of pro­du­cing it. That is the map. The next part re­con­structs the ter­rit­ory.

The mine

Mining reconstructs how work actually runs from the event logs your systems already keep, and the gap it exposes against the map is where the money is. Nintex has retired its own mining product, which changes who supplies the instrument but not the argument.

Mining reconstructs the real process from data you already generate

Pro­cess min­ing re­con­structs how work ac­tu­ally runs, case by case, de­vi­ation by de­vi­ation, from the event logs your ERP and CRM sys­tems already pro­duce as a by-product of doing busi­ness. Noth­ing new needs to be re­cor­ded. Every time an order is raised, an ap­proval clicked, or a re­cord up­dated, the sys­tem stamps it with an iden­ti­fier and a time. Min­ing reads those stamps back into the story of the pro­cess, told by the pro­cess it­self.

The method is worth un­der­stand­ing be­cause its cred­ib­il­ity rests on it. En­ter­prise sys­tems log events, each with a case iden­ti­fier, an activ­ity, a timestamp and some­times an actor. Pro­cess min­ing takes mil­lions of those events and re­as­sembles them into the paths cases ac­tu­ally fol­lowed, which steps, in which order, with what waits between them. The out­put is not an opin­ion about the pro­cess. It is the pro­cess, as re­cor­ded by the ma­chines that ran it.

This is why min­ing beats in­ter­views for fac­tual ques­tions. Nobody has to re­mem­ber, es­tim­ate, or de­fend any­thing. The log does not round its cycle times, for­get its re­runs, or omit the awk­ward vari­ant. It shows every path every case took, in­clud­ing the ones no one would have de­scribed in a work­shop.

The hon­esty cuts both ways, which is the caveat this book keeps re­turn­ing to. The log re­cords what happened, not why, and it re­cords only what the sys­tems cap­tured. Work done in email, on paper or in a spread­sheet leaves no event to mine,

Process IntelligenceVoice of Volge08 •

so a pro­cess pic­ture built only from sys­tem logs has its own blind spots. Good min­ing pairs the data with the people who can ex­plain it, and that pair­ing is part of the method rather than an op­tional extra.

Nin­tex's own po­s­i­tion has changed, and that change is the most im­port­ant fact in this part. On 4 May 2026 Nin­tex an­nounced end-of-life for Nin­tex Pro­cess Dis­cov­ery, the product that de­livered pro­cess min­ing and task min­ing: end-of-life on 1 May 2027, end of sup­port in May 2028, with ex­ist­ing cus­tom­ers moved to Pro­cess Man­ager. Ex­hibit 8 sets out what Nin­tex re­tired and what sur­vives.

Exhibit 8 What Nintex retired, and what survives

CapabilityStatus at November 2026
Process miningDelivered by Nintex Process Discovery. End-of-life 1 May 2027; support ends May 2028
Task miningSame product, same dates. No current Nintex page positions it as shipping
Process CaptureShipping. Desktop activity recording inside Process Manager, GA September 2025
Process mapping and modellingShipping. Process Manager, now surfacing as Nintex Processes

Source: Nin­tex com­munity, Pro­cess Dis­cov­ery end-of-life an­nounce­ment, 4 May 2026, up­dated 17 Au­gust 2026; Nin­tex Pro­cess Man­ager re­lease notes, read 8 Novem­ber 2026.

What the retirement does to the argument

The ana­lyt­ical case is un­af­fected, be­cause the pro­cess as per­formed is still the only hon­est baseline. What has changed is who sup­plies the in­stru­ment. Nin­tex no longer sells a min­ing product, so the mine now comes from a third-party min­ing tool, from Pro­cess Cap­ture at the desktop, or from a sampled log ana­lysis run by hand. The map, and the com­par­ison that gives it value, stay in­side Nin­tex.

From event log to reconstructed process, in four moves

Pro­cess min­ing is easier to trust once you can see the mech­an­ism. There is no black box worth being nervous about; there is a pipeline that reads timestamps and re­builds se­quences. The four moves below are what hap­pens between your sys­tems and the pic­ture on the screen.

  • Ex­tract. Pull event logs from ERP, CRM, line sys­tems
  • Re­con­struct. Re­as­semble events into the paths cases took
  • Ana­lyse. Meas­ure vari­ants, waits, and de­vi­ations
  • Com­pare. Set the real­ity against the de­signed map

Exhibit 9 What each move answers

MoveQuestion it answersTypical finding
ExtractWhat did the systems record?Millions of timestamped events per process
ReconstructWhat paths did cases follow?Many more variants than anyone expected
AnalyseWhere is time and money lost?A few steps hold most of the delay
CompareWhere does reality break from design?Deviations no one had flagged

Source: Volge ana­lysis, 2026. Find­ings are typ­ical pat­terns, not guar­an­teed out­comes.

Process Capture fills the desktop gap

Sys­tem logs miss work done between sys­tems, the copy-paste, the manual re­con­cili­ation, the swivel-chair step where someone reads one screen and types into an­other. Pro­cess Cap­ture re­cords that desktop activ­ity dir­ectly, so the pic­ture in­cludes the work that never touched a logged trans­ac­tion. It reached gen­eral avail­ab­il­ity in the Septem­ber 2025 re­lease. To­gether with log ana­lysis it leaves fewer blind spots than either alone.

Delay concentrates, and mining shows you exactly where

In most pro­cesses a hand­ful of steps hold the large ma­jor­ity of the elapsed time, and the whole value of min­ing is that it names those steps in­stead of leav­ing you to guess. Wait­ing, not work­ing, is usu­ally the vil­lain. A case spends far longer sit­ting in a queue than being act­ively handled, and the queues are in­vis­ible to a swim­lane dia­gram that has no axis for time.

Exhibit 10 Where cycle time goes in a typical purchase-to-pay process (illustrative)

  • Ap­proval wait­ing. 38% of elapsed time
  • Re­work loops. 24%
  • Sys­tem han­doffs. 18%
  • Manual data entry. 13%
  • Ac­tual value-add work. 7%

Source: il­lus­trat­ive com­pos­ite by Volge, 2026, of pat­terns com­monly sur­faced by pro­cess min­ing. Not drawn from a spe­cific named cli­ent; fig­ures are in­dic­at­ive of the shape min­ing typ­ic­ally re­veals, not a bench­mark.

Process IntelligenceVoice of Volge09 •

The pat­tern in Ex­hibit 10 is il­lus­trat­ive, and we have la­belled it as such, but its shape is fa­mil­iar to any­one who has mined a trans­ac­tional pro­cess. The value-add work, the part a per­son is genu­inely paid to do, is a thin slice. The rest is wait­ing, re­work­ing, and shuff­ling data between sys­tems. That com­pos­i­tion is in­vis­ible in in­ter­views, be­cause people de­scribe the work, not the wait­ing between it.

This changes where auto­ma­tion should point. Faced with a pro­cess like this, the in­stinct is to auto­mate the value-add step, be­cause that is the part every­one can see. The data says oth­er­wise. The re­turn is in the ap­proval queue and the re­work loop, where auto­ma­tion and pro­cess re­design re­move wait­ing rather than speed­ing up work that was never the bot­tle­neck.

It also dis­cip­lines the AI con­ver­sa­tion. An agent set loose on the seven per cent of value-add work leaves ninety-three per cent of the elapsed time un­touched. An agent, or a re­design, aimed at the ap­proval wait and the re­work loop at­tacks the ac­tual cost. Min­ing is what tells you which is which, be­fore you com­mit the budget.

The reframe

Stop ask­ing "which task can we auto­mate?" and start ask­ing "where does a case wait, and why?" The first ques­tion flat­ters the tools you already bought. The second finds the money. Min­ing an­swers the second.

What mining cannot see

The mine is only as com­plete as your di­gital ex­haust. Work done on paper, over a call, or in a sys­tem the logs never reach stays in­vis­ible to it, which makes the mine a truer pic­ture than the map, not the whole truth.

The gap between the map and the mine is the product

The single most valu­able arte­fact Nin­tex can pro­duce is not the map and not the mine but the dif­fer­ence between them, the gap where the de­signed pro­cess and the per­formed pro­cess dis­agree. That gap is not a fail­ure to be ti­died away. It is a cos­ted, pri­or­it­ised list of everything worth im­prov­ing, auto­mat­ing, or de­lib­er­ately leav­ing alone, and it is the reason to buy both in­stru­ments rather than either.

Neither in­stru­ment de­liv­ers this alone, which is the com­mer­cial heart of the ar­gu­ment. The map without the mine is a plan you can­not check. The mine without the map is a pile of real­ity with no baseline to judge it against, you can see that cases took el­even paths, but not which of the el­even you

in­ten­ded. Only the com­par­ison pro­duces a work­list, and only a work­list jus­ti­fies a budget.

Exhibit 11 Map versus mine: reading the gap

The map saysThe mine showsThe gap is worth
One approval pathEleven live variantsStandardisation, then automation
Approval takes hoursApproval waits daysThe largest cycle-time prize
Controls always appliedControls skipped at scaleAudit and compliance exposure
Handoffs are cleanCases loop back repeatedlyRework elimination
Process is stableBehaviour drifts by quarterContinuous monitoring

Source: Volge ana­lysis, 2026. Rows are rep­res­ent­at­ive gap types, not a single cli­ent's find­ings.

The map tells you what you meant to build. The mine tells you what you built. You get paid for closing the distance.

Volge: on the map-versus-mine gap

The gap is also self-pri­or­it­ising, which C-level read­ers will ap­pre­ci­ate. Each entry car­ries a size, the cycle time it wastes, the volume it touches, the con­trol it by­passes. That turns pro­cess im­prove­ment from a mat­ter of taste and polit­ics into a mat­ter of arith­metic. You work the gap in order of value, and you can say why each item sits where it does.

One dir­ec­tion of the gap re­mained awk­ward until re­cently. Min­ing speaks BPMN, the tech­nical nota­tion ana­lysts read and pro­cess own­ers do not. So the mined real­ity, how­ever valu­able, ten­ded to stay trapped in the ana­lyst's tool, un­read by the people who owned the work. The fea­ture that un­blocked this reached gen­eral avail­ab­il­ity in the Decem­ber 2025 re­lease, and it is the sub­ject of Part IV.

The variant explosion is the finding that surprises boards most

The first mined res­ult that un­settles a lead­er­ship team is al­most al­ways the vari­ant count, the dis­cov­ery that a pro­cess every­one de­scribed as one path was, in the logs, dozens. This is not chaos, or not only chaos. Some vari­ants are sens­ible local ad­apt­a­tions; some are qual­ity gates; some are genu­ine mess. The point of min­ing is that you can fi­nally tell which is which.

Process IntelligenceVoice of Volge10 •

Exhibit 12 How variant volume typically distributes (illustrative)

  • Main path (as de­signed). 58% of cases
  • Top 5 vari­ant paths. 27%
  • Long tail of vari­ants. 15%

The "long tail" here bundles dozens of rare paths that in­di­vidu­ally carry little volume but col­lect­ively res­ist auto­ma­tion and hide risk.

Source: il­lus­trat­ive com­pos­ite by Volge, 2026. Shape is rep­res­ent­at­ive of mined trans­ac­tional pro­cesses; not a spe­cific bench­mark.

Read Ex­hibit 12 the way an op­er­ator would. The de­signed path car­ries a com­fort­able-look­ing ma­jor­ity, which is why in­ter­view-based map­ping feels ad­equate, most of the time, most cases really do fol­low it. The trouble is the rest. The top hand­ful of vari­ants carry real volume, and the long tail, though thin per path, col­lect­ively de­feats any auto­ma­tion built for the main path alone.

The stra­tegic move is not to elim­in­ate vari­ants on sight. Some earn their place. The move is to sort them. Stand­ard­ise the ones that are need­less di­ver­gence, pro­tect the ones that are de­lib­er­ate, and auto­mate the main path only once it is stable enough to be worth the build. Min­ing sup­plies the sort; judge­ment sup­plies the de­cision.

This is where the map earns its keep a second time. Com­pared against the mine, it tells you which vari­ants were in­ten­ded and which crept in, sep­ar­at­ing a de­signed branch from an un­doc­u­mented work­around. Without the map, every vari­ant looks the same. With it, the mess sorts it­self into sig­nal.

Why boards should not panic at the variant count

A high vari­ant count is not proof of a broken or­gan­isa­tion. It is proof that you can now see the or­gan­isa­tion. The count was al­ways that high; you simply could not meas­ure it. Meas­ure­ment is the first im­prove­ment.

You now have reality; the problem is who can read it

At the end of min­ing you hold some­thing most or­gan­isa­tions have never had, which is a fac­tual, com­plete, case-level ac­count of how the work truly runs. The re­main­ing obstacle is one of trans­la­tion. That ac­count is ex­pressed in BPMN, and BPMN is a lan­guage of ana­lysts, not of the pro­cess own­ers who must act on what it says.

This is a real bar­rier, not a cos­metic one. A find­ing that only the ana­lyt­ics team can read is a find­ing that stalls. The pro­cess owner in fin­ance or pro­cure­ment, the per­son with the au­thor­ity and the con­text to change the pro­cess, can­not act on a dia­gram they can­not in­ter­pret. So the mined truth sits in a spe­cial­ist tool, ad­mired and un­used, while the map in Pro­cess Man­ager stays com­fort­ably out of date.

The whole sys­tem Nin­tex has as­sembled only pays off when the two dir­ec­tions con­nect, with design flow­ing into real­ity for com­par­ison, and real­ity flow­ing back into design for ac­tion. For years the second dir­ec­tion was the weak link. Min­ing could ex­pose the truth, but re­turn­ing that truth to the busi­ness in a form it could use re­quired manual trans­la­tion that rarely happened.

That link closed in the Decem­ber 2025 re­lease, when AI Model to Map reached gen­eral avail­ab­il­ity and turned tech­nical BPMN into busi­ness-read­able pro­ced­ures auto­mat­ic­ally. A second link opened in the Oc­to­ber 2026 re­lease, when the Pro­cess Man­ager MCP server gained au­then­tic­a­tion for el­even AI cli­ents, among them ChatGPT, Claude, Gem­ini and Mi­crosoft 365 Co­pi­lot. Part IV be­gins there.

  • Map. De­signed pro­cess, busi­ness lan­guage
  • Mine. Real pro­cess, BPMN
  • Model to Map. BPMN → busi­ness-read­able
  • Act. Own­ers im­prove and auto­mate

What this part established

Min­ing re­con­structs the real pro­cess from data you already keep, ex­poses where time is lost and how many vari­ants exist, and, set against the map, pro­duces the cos­ted gap that is the ac­tual product. Nin­tex no longer sells the min­ing half, which is a sourcing ques­tion rather than an ar­gu­ment against it. The one miss­ing piece was trans­la­tion. Part IV sup­plies it.

This is why we held the fea­ture back from Part II des­pite it being a Pro­cess Man­ager cap­ab­il­ity. Its sig­ni­fic­ance is not that it drafts a map; the AI Pro­cess Gen­er­ator already does that. Its sig­ni­fic­ance is dir­ec­tional. It runs mine-to-map, clos­ing the re­turn path that turns a dia­gnostic ex­er­cise into an op­er­at­ing loop. Chris Ellis, Nin­tex's dir­ector of solu­tion en­gin­eer­ing, called the re­lease “a major step for­ward in mak­ing pro­cess in­tel­li­gence ac­cess­ible, ac­tion­able, and scal­able across the or­gan­iz­a­tion”, as re­por­ted by PEX Net­work on 6 Janu­ary 2026, and the ac­cess­ib­il­ity he names is ex­actly that, real­ity rendered read­able.

Re­call the im­passe from Part III. Min­ing pro­duces truth in a tech­nical dia­lect. The people em­powered to act on that truth

Process IntelligenceVoice of Volge11 •

do not speak the dia­lect. The trans­la­tion between them was, until this year, a manual job that com­peted with everything else on an ana­lyst's desk and usu­ally lost. So in­sight ac­cu­mu­lated faster than any­one could con­vert it into ac­tion.

Closing the gap

AI Model to Map bridges the mined reality back to the business-readable map. From there, insight becomes action: DocGen and Forms carry the improved process into daily work, and the proof justifies the spend.

AI Model to Map turns the mined truth into something the business can act on

AI Model to Map is the fea­ture that makes the whole map-and-mine sys­tem work in prac­tice, be­cause it car­ries the mined real­ity across the one bar­rier that used to stop it. That bar­rier is lan­guage. Gen­er­ally avail­able since the Decem­ber 2025 re­lease, it con­verts tech­nical BPMN, the nota­tion min­ing pro­duces, into busi­ness-read­able pro­ced­ures that a pro­cess owner can read and ques­tion and change. The mine flows back into the map, auto­mat­ic­ally, and the loop closes.

Model to Map re­moves the trans­la­tion cost. The mined pro­cess, ex­pressed in BPMN, be­comes a pro­ced­ure in plain busi­ness steps in­side Pro­cess Man­ager, the same tool the front line already uses to own its maps. What was locked in the ana­lyst's screen be­comes a doc­u­ment the pro­cess owner can read on Monday and change by Fri­day. The bot­tle­neck between know­ing and doing is the one being auto­mated.

With the loop closed, the se­quence from Part I fi­nally runs end to end without a manual gap in the middle. Dis­cover and map, then mine and com­pare, then act and map again, be­cause act­ing changes the pro­cess and the cycle re­peats. Ex­hibit 13 draws the closed loop.

A caveat we owe you

AI Model to Map has been gen­er­ally avail­able only since the Decem­ber 2025 re­lease. We de­scribe it on Nin­tex's fram­ing and our own read­ing, not on a track re­cord across dozens of en­ter­prises. Auto­mated trans­la­tion also smooths, a mis­la­belled step reads as cleanly as a cor­rect one, so treat the out­put as a strong draft an owner still checks, not an or­acle.

Exhibit 13 The closed loop, design, reality, and back again

  • Pro­cess Man­ager. The de­signed map, in busi­ness lan­guage
  • Mined real­ity. Event logs and desktop cap­ture, in BPMN
  • AI Model to Map. BPMN rendered busi­ness-read­able
  • Own­ers act. Sim­plify, auto­mate, mon­itor
  • Pro­cess Man­ager MCP. Ap­proved pro­cess know­ledge, read by AI as­sist­ants

The out­put of "own­ers act" be­comes the next de­signed map, and the loop runs again, pro­cess un­der­stand­ing as a habit, not a pro­ject.

Source: Volge ana­lysis of Nin­tex cap­ab­il­it­ies, 2026. AI Model to Map GA, Pro­cess Man­ager 4.12.27.0, Decem­ber 2025; MCP au­then­tic­a­tion, re­lease 4.26.23, Oc­to­ber 2026.

A readable process is only useful if it drives the daily work: DocGen and Forms make it do so

Un­der­stand­ing a pro­cess pays noth­ing until it changes what hap­pens on the ground, and two Nin­tex en­han­cers carry the im­proved pro­cess from the dia­gram into the day-to-day work. Nin­tex DocGen gen­er­ates doc­u­ments from pro­cess data. Nin­tex Forms cap­tures data into the pro­cess. Named here as en­han­cers rather than head­line products, they are what stop an im­proved map from stay­ing a map.

01 Nin­tex Forms: clean data in

Forms cap­tures the data a pro­cess needs at the point the pro­cess needs it, in a struc­tured shape rather than a free-text email. Clean struc­tured input is what makes down­stream auto­ma­tion pos­sible; garbage en­ter­ing a well-de­signed pro­cess still pro­duces garbage. Forms is the front door that keeps the pro­cess fed with data it can ac­tu­ally use.

02 Nin­tex DocGen: fin­ished doc­u­ments out

DocGen gen­er­ates the doc­u­ments a pro­cess pro­duces, among them con­tracts and quotes, state­ments and let­ters, dir­ectly from pro­cess data, without the copy-paste step that in­tro­duces er­rors and delay. The pro­cess be­comes the source of the doc­u­ment rather than a pre­lude to someone as­sem­bling it by hand. That re­moves a whole class of the manual data entry Ex­hibit 10 flagged as pure waste.

Neither en­han­cer is the reason to buy. We are de­lib­er­ately not over-fea­tur­ing them, be­cause the ar­gu­ment of this book is about see­ing the pro­cess, not about any single tool for run­ning it. But they mat­ter to the C-level reader for a spe­cific reason. They are how the im­prove­ment you iden­ti­fied in the gap ac­tu­ally reaches the people doing the work. In­sight that never touches the daily task is a re­port, not a res­ult.

This is the hon­est end of the "in­sight to ac­tion" prom­ise. There is no single but­ton marked "act". There is a read­able pro­cess, a clean way to feed it, and an auto­matic way to pro­duce what it

Process IntelligenceVoice of Volge12 •

owes the cus­tomer. Above all, there is the dia­gnosis that told you which part was worth the ef­fort.

Placed in the loop, Forms sits at the pro­cess en­trance and DocGen at its exit, with the im­proved, mined-and-mapped pro­cess in between. Data enters clean, moves through a pro­cess you can now see and have de­lib­er­ately shaped, and leaves as a fin­ished doc­u­ment. The un­der­stand­ing from Parts II and III be­comes op­er­a­tional rather than ana­lyt­ical.

The role of the enhancers

DocGen and Forms are en­han­cers, not the head­line. They turn an im­proved pro­cess into im­proved work. Forms feeds it clean data, and DocGen pro­duces its doc­u­ments. Buy them for what they do to a pro­cess you already un­der­stand, not as sub­sti­tutes for un­der­stand­ing it.

The evidence that this is a mature platform, not a promising demo

Nin­tex is a widely de­ployed, in­de­pend­ently rated plat­form rather than an emer­ging bet, and the num­bers say so on scale and on sat­is­fac­tion. Nin­tex's com­pany page re­por­ted more than 7,000 or­gan­isa­tions in over 100 coun­tries when it was last mod­i­fied in Septem­ber 2026, though the news­room boil­er­plate of 22 Oc­to­ber 2026 drops the count. For­res­ter named Nin­tex a Strong Per­former in The For­res­ter Wave: Agen­tic Pro­cess Or­ches­tra­tion Plat­forms, Q3 2026. This is the part of the ar­gu­ment that de-risks the pur­chase.

  • 7,000. +. or­gan­isa­tions across 100+ coun­tries. Nin­tex, Sep 2026
  • 4.2 /5. G2 rat­ing, 1,326 veri­fied re­views. G2, Nov 2026
  • 8 /10. TrustRadius, 530 re­views. TrustRadius, Nov 2026

Exhibit 14 Independent proof points, with source and vintage

Proof pointFigureSource · date
Deployed scale7,000+ organisations, 100+ countriesNintex company page, Sep 2026
G2 rating4.2 / 5 across 1,326 reviewsG2, Nov 2026
TrustRadius rating8 / 10 across 530 reviewsTrustRadius, Nov 2026
RecognitionTop Rated, seventh consecutive yearTrustRadius, 2025 award
Analyst placementStrong Performer, not a LeaderForrester Wave, Q3 2026
Logo-wall namesL'Oréal, Panasonic, Sonos, CARFAXNintex solutions pages, Jul 2026

Source: as stated per row. For­res­ter placed Nin­tex as a Strong Per­former, not a Leader; no Gart­ner Magic Quad­rant place­ment is as­ser­ted here.

Read the rat­ings for what they are, which is dur­able third-party sig­nal from large re­view bases. A 4.2 out of 5 across 1,326 G2 re­views and an 8 out of 10 across 530 on TrustRadius are not the scores of a product that frus­trates its users, and a TrustRadius Top Rated award in 2025 for a sev­enth con­sec­ut­ive year, re­peated in 2026, is a state­ment about con­sist­ency rather than a single good quarter. These are noisy meas­ures, but the sample sizes make the dir­ec­tion re­li­able.

The logo-wall names mat­ter for a dif­fer­ent reason, and they carry a smal­ler claim than they ap­pear to. L'Oréal, Panasonic, Sonos and CARFAX ap­pear as cus­tomer logos on Nin­tex's solu­tions pages, with no pub­lished case study, met­ric or quo­ta­tion be­hind any of them. Read them as evid­ence that the plat­form is bought by large, de­mand­ing or­gan­isa­tions, not as evid­ence of what it achieved in­side them. Nin­tex's pub­lished air­line story is Vir­gin Aus­tralia, a dif­fer­ent com­pany from the Vir­gin At­lantic an earlier edi­tion of this re­port named.

We should be equally clear about what we are not claim­ing. Nin­tex is not a Gart­ner Magic Quad­rant leader on any des­ig­na­tion we have sourced, and For­res­ter placed it as a Strong Per­former rather than a Leader in the Agen­tic Pro­cess Or­ches­tra­tion Plat­forms Wave of the third quarter of 2026. The proof above stands on its own terms, which is the only kind of proof worth put­ting in front of a board.

The economics: one documented return, and why your own number is the one that matters

The fin­an­cial case rests on one doc­u­mented study, at­trib­uted and dated so you can weigh it your­self. For­res­ter's Total Eco­nomic Im­pact study of Nin­tex Work­flow found a 213 per cent re­turn on in­vest­ment with a nine-month pay­back. Nin­tex states that the study was con­duc­ted in late 2022, which makes the vin­tage old, and it cov­ers Nin­tex Work­flow rather than the map­ping cap­ab­il­ity that is this book's sub­ject. No newer For­res­ter study of Nin­tex ex­ists.

Exhibit 15 The return, in two figures

  • For­res­ter TEI ROI. 213% re­turn on in­vest­ment
  • Pay­back period. 9 months

ROI bar shown to scale against a 100% ref­er­ence; pay­back bar shown against a 24-month ref­er­ence. The For­res­ter study was con­duc­ted in late 2022 and de­scribes Nin­tex Work­flow spe­cific­ally.

Process IntelligenceVoice of Volge13 •

Source: For­res­ter Total Eco­nomic Im­pact of Nin­tex Work­flow, study con­duc­ted late 2022 per Nin­tex. Pay­back and ROI as re­por­ted.

Treat the For­res­ter fig­ures with the care their age de­serves. A 213 per cent ROI and a nine-month pay­back are strong num­bers, but the study was con­duc­ted in late 2022 and con­cerns Nin­tex Work­flow rather than the map­ping and min­ing cap­ab­il­it­ies that are this book's sub­ject. We in­clude it as dir­ec­tional evid­ence that Nin­tex auto­ma­tion pays back quickly, not as a fore­cast of your spe­cific res­ult. The vin­tage is on the label for a reason.

Against that single dated study, the use­ful num­ber is your own. A map­ping pro­gramme's re­turn shows up as hours not spent re­writ­ing pro­ced­ures and as ap­provals that stop queueing, and both are meas­ur­able in the logs be­fore and after. Nin­tex pub­lishes cus­tomer stor­ies in local gov­ern­ment and else­where, but none we could source car­ries both a dated total and an in­de­pend­ent audit, so we quote none of them. The cal­cu­la­tion is cheap to run and the an­swer is spe­cific to you.

The hon­est sum­mary is that the re­turns are real and the exact num­ber is yours to dis­cover. What the map-and-mine ap­proach changes is the con­fid­ence with which you pur­sue it. In­stead of auto­mat­ing on faith and hop­ing for a For­res­ter-shaped out­come, you mine the gap, size each op­por­tun­ity, and spend against evid­ence. The ROI stops being a prom­ise and be­comes a cal­cu­la­tion.

How to read these numbers

The 213 per cent re­turn and nine-month pay­back are real but dated, from a study Nin­tex says was con­duc­ted in late 2022, and spe­cific to Nin­tex Work­flow rather than to map­ping. It is not a guar­an­tee for your or­gan­isa­tion. It is a reason to run the cal­cu­la­tion on your own pro­cesses, which is what Volge does first.

Handing the map to an agent is a governance act, not an integration

The Pro­cess Man­ager MCP server turns the ap­proved map into some­thing an AI as­sist­ant can read, and that is a smal­ler and more con­sequen­tial change than it sounds. Nin­tex opened it in beta on 13 Au­gust 2026 and shipped au­then­tic­a­tion for el­even cli­ents in the Oc­to­ber 2026 re­lease, 4.26.23, among them ChatGPT, Claude, Gem­ini and Mi­crosoft 365 Co­pi­lot. The server is read-only.

Read-only is the re­as­sur­ing part and also the part that mis­leads. An as­sist­ant that can read the pro­cess lib­rary can­not

change a pro­cess, but it can re­peat what the lib­rary says to any­one who asks, in­clud­ing a pro­cess that is out of date, a draft that was never ap­proved, or a pro­ced­ure that ex­ists only for one re­gion. The ex­pos­ure is not write ac­cess. It is the au­thor­ity a stale an­swer ac­quires when a ma­chine de­liv­ers it flu­ently.

Three de­cisions fol­low, and all three be­long to the pro­cess func­tion rather than to the plat­form team. Which pro­cesses are ex­posed, mean­ing whether draft and re­tired maps are reach­able or only pub­lished ones. Which audi­ences are ex­posed, be­cause a pro­ced­ure that is cor­rect for one jur­is­dic­tion and wrong for an­other should not be answered the same way for both. And what the cur­rency floor is, be­cause a map last ed­ited three years ago should not be served as the an­swer at all.

The third is meas­ur­able, which makes it the use­ful one. Set a cur­rency threshold, ex­clude any­thing past it from the server, and the ex­clu­sion list be­comes the back­log for the own­ers. A lib­rary that can­not be served to an as­sist­ant be­cause most of it is stale is a lib­rary that was already fail­ing its human read­ers. The ma­chine simply asked more often.

Process intelligence in the Gulf and India

Nintex runs a United Arab Emirates instance, and it is for exactly this subject. The status register lists the UAE for Process Manager production and Nintex Analytics, and for nothing else, which makes it the only in-region capability any non-Microsoft vendor in this series operates in the Gulf. Against that, Bahrain requires prior written authorisation for the kind of cross-system automated processing that process discovery is, and India requires logs to be maintained within its own jurisdiction. Classification comes before discovery, and discovery before any export.

The Gulf has one in-re­gion home for pro­cess in­tel­li­gence, and it be­longs to this half of the port­fo­lio rather than the other. Nin­tex's own ser­vices status re­gister, read on 8 Novem­ber 2026, lists a United Arab Emir­ates in­stance for Nin­tex Pro­cess Man­ager pro­duc­tion and for Nin­tex Ana­lyt­ics. It lists the UAE for no other ser­vice. Nin­tex Auto­ma­tion Cloud, the work­flow plat­form, runs in seven re­gions and none of them is in the Middle East or India.

That asym­metry is the right way into the re­gional pic­ture, be­cause it cuts in this sub­ject's fa­vour. The map can sit in the Gulf. The work­flow that acts on the map can­not. For a pro­gramme whose first de­liv­er­able is an hon­est ac­count of how the work runs, the in­stru­ment that holds that ac­count has an in-re­gion home and the in­stru­ment that auto­mates against it does not. Se­quence the pro­gramme ac­cord­ingly.

Process IntelligenceVoice of Volge14 •

It is also the only such cap­ab­il­ity either half of the re­search be­hind this chapter found. Flex­era has no Gulf or In­dian host­ing re­gion. Nin­tex Auto­ma­tion Cloud has none. Against that back­ground a live United Arab Emir­ates in­stance for the map­ping and ana­lyt­ics ser­vices is a genu­inely dis­tin­guish­ing fact, and it rests on an op­er­a­tional re­gister rather than on a bro­chure.

The United Arab Emirates appears on the Nintex status register for Process Manager production and Nintex Analytics, and for nothing else.

Nin­tex Ser­vices status re­gister, read 8 Novem­ber 2026

One dis­tinc­tion has to be kept, and it is the kind a pro­cure­ment team will test. The re­gional foot­print is op­er­a­tional fact, not a con­trac­tual com­mit­ment. Nin­tex's trust centre is a gated portal that re­turned no text, and no data res­id­ency art­icle could be loc­ated in its product doc­u­ment­a­tion. No con­trac­tual res­id­ency com­mit­ment by Nin­tex is as­ser­ted here. Where res­id­ency is a bind­ing ob­lig­a­tion, ob­tain the com­mit­ment in the con­tract, by name, and do not infer it from a status page.

Bahrain regulates the act of discovery itself

Bahrain is the jur­is­dic­tion that reaches pro­cess dis­cov­ery dir­ectly, and al­most nobody reads it that way. Art­icle 15 of the Per­sonal Data Pro­tec­tion Law, Law No. 30 of 2018, re­quires prior writ­ten au­thor­isa­tion from the Au­thor­ity for pro­cessing con­sist­ing of visual re­cord­ing used for mon­it­or­ing, and for auto­mated pro­cessing of per­sonal data files held by two or more con­trol­lers pro­cessing for dif­fer­ent pur­poses. Read the second limb against what cross-sys­tem pro­cess min­ing does. It as­sembles per­sonal data from the re­cords of sev­eral sys­tems, held for sev­eral pur­poses, and pro­cesses it auto­mat­ic­ally to re­con­struct a se­quence.

Min­is­terial Order No. 43 of 2022 adds a second gate. Art­icle 3 re­quires a data pro­tec­tion im­pact as­sess­ment for sys­tem­atic mon­it­or­ing of a pub­licly ac­cess­ible area on a large scale. Desktop activ­ity cap­ture is not that, but the draft­ing is a re­minder that Bahraini law treats ob­ser­va­tion of be­ha­viour as a cat­egory re­quir­ing as­sess­ment rather than as or­din­ary pro­cessing.

The prac­tical con­sequence is a se­quen­cing one rather than a pro­hib­i­tion. In Bahrain, es­tab­lish whether the dis­cov­ery design re­quires prior au­thor­isa­tion be­fore the first ex­tract is taken, not after the find­ings are cir­cu­lated. An au­thor­isa­tion sought in ad­vance is a pre­con­di­tion. The same ques­tion

raised by a su­per­visor after the fact is a find­ing, and the find­ing at­taches to pro­cessing that has already happened.

Desktop capture is the part that needs a stated basis everywhere

Event-log min­ing and desktop cap­ture sit dif­fer­ently in law, and a pro­gramme that treats them as one activ­ity will mis­state its own po­s­i­tion. Log min­ing reads re­cords the sys­tems already keep as a by-product of doing busi­ness. Desktop cap­ture ob­serves a named per­son work­ing. The second is mon­it­or­ing, and every jur­is­dic­tion in this chapter treats mon­it­or­ing more care­fully than it treats re­cord-keep­ing.

Four de­cisions settle it, and all four be­long in writ­ing be­fore de­ploy­ment. Which pop­u­la­tions are in scope, and whether con­sulta­tion with them is re­quired or merely ad­vis­able. What is cap­tured, mean­ing whether key­strokes, screen con­tent or only ap­plic­a­tion and win­dow metadata. How long the cap­ture is re­tained, which is where the In­dian re­ten­tion floor dis­cussed below col­lides with a pri­vacy in­stinct to de­lete quickly. And who may read the raw cap­ture, as against the ag­greg­ated map de­rived from it.

The fourth is the one that most often goes wrong. A map built from cap­ture is an op­er­a­tional arte­fact. The raw cap­ture is a re­cord of an in­di­vidual's work­ing day, and the two should not carry the same ac­cess list.

India requires the logs, and requires them to stay

India is the jur­is­dic­tion where the raw ma­ter­ial of pro­cess in­tel­li­gence is man­dated and its loc­a­tion reg­u­lated, which makes it un­usu­ally fa­vour­able ground and un­usu­ally ex­act­ing. CERT-In's Dir­ec­tions of 28 April 2022, No. 20(3)/2022-CERT-In, are in force and un­amended. Dir­ec­tion (iv) re­quires every ser­vice pro­vider, in­ter­me­di­ary, data centre, body cor­por­ate and gov­ern­ment or­gan­isa­tion to en­able logs of all their in­form­a­tion and com­mu­nic­a­tions tech­no­logy sys­tems and main­tain them se­curely for a rolling period of 180 days, "main­tained within the In­dian jur­is­dic­tion". Dir­ec­tion (ii) re­quires re­port­ing of lis­ted in­cid­ents within six hours of no­ti­cing them, which is what the log loc­a­tion is ul­ti­mately for.

The Di­gital Per­sonal Data Pro­tec­tion Rules, 2025, no­ti­fied on 13 Novem­ber 2025 as G.S.R. 846(E), add a re­ten­tion floor from the other dir­ec­tion. They com­mence in tranches, with rule 4 on con­sent man­agers from 13 Novem­ber 2026 and rules 3, 5 to 16, 22 and 23 from 13 May 2027. Rule 6(1) then re­quires en­cryp­tion, ob­fus­ca­tion, mask­ing or vir­tual tokens, ap­pro­pri­ate ac­cess con­trol, and "vis­ib­il­ity of ac­cess through logs,

Process IntelligenceVoice of Volge15 •

mon­it­or­ing and re­view". Rules 6(1)(e) and 8(3) im­pose a min­imum one-year re­ten­tion of per­sonal data, as­so­ci­ated traffic data and pro­cessing logs, bind­ing on a pro­cessor as well as on the fi­du­ciary.

CERT-In's own fre­quently asked ques­tions of May 2022 soften the loc­a­tion rule. Ques­tion 35 states that logs "may be stored out­side India also as long as the ob­lig­a­tion to pro­duce logs to CERT-In is ad­hered to by the en­tit­ies in a reas­on­able time". Ques­tion 37 enu­mer­ates the log types in scope, in­clud­ing fire­wall, se­cur­ity mon­it­or­ing, web, data­base, mail, proxy, ap­plic­a­tion and vir­tual private net­work logs, to­gether with event logs of crit­ical sys­tems. The bind­ing text says India. The cla­ri­fic­a­tion says any­where, provided pro­duc­tion is timely. Cite both, de­cide de­lib­er­ately, and re­cord the de­cision.

For a pro­cess-min­ing en­gage­ment with its nearest Nin­tex re­gion in Singa­pore, that is the op­er­at­ive ten­sion. An event-log ex­tract taken for dis­cov­ery is de­rived from the logs dir­ec­tion (iv) cov­ers, and the ex­tract crosses a bor­der. Settle whether the ex­tract is it­self a log for these pur­poses, settle the pro­duc­tion time, and put both an­swers in the design re­cord.

So In­dian law re­quires the vis­ib­il­ity that pro­cess in­tel­li­gence con­sumes, and then re­quires it to be kept for a year. A dis­cov­ery pro­gramme in India has more raw ma­ter­ial than one any­where else in this chapter, and less free­dom to dis­card it. Rule 7 com­pletes the pic­ture with breach in­tim­a­tion to each af­fected data prin­cipal without delay, to the Data Pro­tec­tion Board without delay, and again in de­tail within sev­enty-two hours. There is no de min­imis threshold.

What India does not give you, and it matters for what mining may conclude

India of­fers no re­sid­ual bal­an­cing test, which con­strains what a pro­cess in­tel­li­gence pro­gramme may do with what it finds. Sec­tion 4 of the Di­gital Per­sonal Data Pro­tec­tion Act, 2023 per­mits pro­cessing only with con­sent or for a le­git­im­ate use under sec­tion 7, and sec­tion 7 is ex­haust­ive. There is no equi­val­ent of a le­git­im­ate-in­terests basis, so a busi­ness can­not self-cer­tify a new pro­cessing pur­pose as le­git­im­ate once dis­cov­ery has sug­ges­ted one.

The re­search and stat­ist­ical ex­emp­tion is nar­rower than it looks. Sec­tion 17(2)(b) is avail­able only where the data is not used to take any de­cision spe­cific to a data prin­cipal, which ex­cludes most com­mer­cial de­ploy­ment of a find­ing. A mined in­sight that changes how a par­tic­u­lar cus­tomer's case is routed is a de­cision spe­cific to a data prin­cipal.

Nor is there an auto­mated-de­cision-mak­ing right in In­dian law to con­strain what min­ing may infer, which cuts the other way. The ab­sence of such a right is not per­mis­sion. It means the con­straint sits on the law­ful basis for the pro­cessing rather than on the out­put, and the law­ful basis has to be es­tab­lished be­fore the ex­tract, not after the in­sight.

Exhibit The regional position for a process intelligence programme

JurisdictionNintex process mapping in regionThe gate on discovery itselfThe supplier gate
United Arab EmiratesYes, Process Manager production and Nintex AnalyticsFederal PDPL in force with no Executive Regulations as at 8 November 2026; health data may not leave the StateDESC Cloud Service Provider Security Standard for Dubai government and semi-government
BahrainNoPDPL Article 15: prior written authorisation for automated processing of files held by two or more controllers for different purposesNo national accreditation scheme; international certifications specified instead
OmanNoClassification before cloud processing, and a record of processing covering any movement of personal data outside OmanTRA authorisation before contracting
Saudi ArabiaNoProhibition on transferring content comprising government agency data outside the KingdomCST registration at Class A, B or C by data classification
KuwaitNoLevel 3 and Level 4 data and government-held personal data may not be hosted outside KuwaitCITRA cloud licence, requiring a data centre inside Kuwait
IndiaNo; nearest region is SingaporeCERT-In direction (iv): logs maintained within the Indian jurisdiction for 180 days, with FAQ 35 to the contraryMeitY empanelment and STQC audit status for SEBI regulated entities under the 2023 cloud circular

Sources: Nin­tex Ser­vices status re­gister, read 8 Novem­ber 2026, for re­gional avail­ab­il­ity. 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.

Three fur­ther Omani ob­lig­a­tions shape a dis­cov­ery design. The Ex­ec­ut­ive Reg­u­la­tions re­quire a re­cord of pro­cessing

Process IntelligenceVoice of Volge16 •

activ­it­ies from both con­trol­ler and pro­cessor, and that re­cord must cover any move­ment or pro­cessing of per­sonal data out­side Oman. A data pro­tec­tion of­ficer is man­dat­ory for con­trol­lers, with the of­ficer's name and con­tact de­tails pub­lished. And the Cloud Com­put­ing First Cir­cu­lar, Min­istry of Trans­port, Com­mu­nic­a­tions and In­form­a­tion Tech­no­logy Sum­mary 10/2026 of 28 June 2026, re­quires gov­ern­ment en­tit­ies to clas­sify data be­fore stor­ing or pro­cessing it in a cloud en­vir­on­ment.

Oman prices the export, not the discovery

Oman has no in­stru­ment reach­ing pro­cess dis­cov­ery dir­ectly, and it has the re­gion's most severe pen­alty for mov­ing the res­ult­ing data out. The Per­sonal Data Pro­tec­tion Law, Royal De­cree No. 6/2022, has been fully en­force­able since 5 Feb­ru­ary 2026, when the trans­ition period under the Ex­ec­ut­ive Reg­u­la­tions closed. Crim­inal fines are im­posed by the com­pet­ent court, and most con­tra­ven­tions sit between 500 and 20,000 Rial Omani. Art­icle 23, un­law­ful cross-bor­der trans­fer, sits at 100,000 to 500,000.

So in Oman, clas­si­fic­a­tion is a pre­con­di­tion of dis­cov­ery rather than an out­put of it. That in­verts the in­stinct of most pro­cess pro­grammes, which is to dis­cover first and clas­sify what they find. For a gov­ern­ment unit the in­ver­sion is ex­pli­cit. 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.

Saudi Arabia, where the export rule governs the extract

Saudi Ar­a­bia's con­straint falls on the ex­tract rather than on the ana­lysis. 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.

One 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.

An event-log ex­tract from a Saudi gov­ern­ment agency's sys­tems is con­tent com­pris­ing that agency's data. There is no Nin­tex re­gion in the King­dom, so a dis­cov­ery en­gage­ment for a Saudi gov­ern­ment cus­tomer is run against data that can­not leave. That points to a sampled ana­lysis con­duc­ted in­side the King­dom rather than a hos­ted min­ing plat­form, which is a method de­cision rather than a lim­it­a­tion.

For a Saudi fin­an­cial in­sti­tu­tion the gate is an ap­proval rather than a geo­graphy. 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 where they are not, and 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 with a ra­tionale for why the work can­not be done within the King­dom.

The gates sit on the supplier, not only on the data

Four ac­cred­it­a­tion gates in the re­gion bind the sup­plier rather than the work­load, and a dis­cov­ery pro­gramme for a pub­lic body stalls at pro­cure­ment without them. 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. In Dubai, the Dubai Elec­tronic Se­cur­ity Cen­ter's Cloud Ser­vice Pro­vider Se­cur­ity Stand­ard is man­dat­ory, in the Cen­ter's 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 li­cence from the Com­mu­nic­a­tion and In­form­a­tion Tech­no­logy Reg­u­lat­ory Au­thor­ity, and clause 2.1 of its Cloud Com­put­ing Reg­u­lat­ory Frame­work ver­sion 2.4 per­mits only a li­censed party with a data centre and in­fra­struc­ture in­side Kuwait to con­tract with the pub­lic sec­tor.

Process IntelligenceVoice of Volge17 •

Oman re­quires a Tele­com­mu­nic­a­tions Reg­u­lat­ory Au­thor­ity au­thor­isa­tion.

Bahrain is the ex­cep­tion, and the con­trast is in­struct­ive. Its Cloud First Policy ver­sion 1.0, dated 24 April 2017 on its title page, names no na­tional ac­cred­it­a­tion scheme and spe­cifies in­ter­na­tional cer­ti­fic­a­tions in­stead. It also takes an ex­pli­cit po­s­i­tion against loc­al­isa­tion, stat­ing under Data Sov­er­eignty that res­id­ency re­stric­tions "un­der­mine the eco­nom­ies of scale and se­cur­ity be­ne­fits to be gained from shared com­put­ing in­fra­struc­ture". So Bahrain is re­laxed about where the min­ing plat­form sits and ex­act­ing about the au­thor­isa­tion to run the pro­cessing at all. Those are two dif­fer­ent ques­tions, and a single re­gional policy that merges them will get both wrong.

What this constrains

The re­gional ground changes the order of the five steps in chapter one rather than the steps them­selves. Clas­sify be­fore you dis­cover, be­cause in Oman and Saudi Ar­a­bia the clas­si­fic­a­tion de­cides whether the ex­tract may leave at all. Es­tab­lish the law­ful basis and any prior au­thor­isa­tion be­fore the first ex­tract, be­cause in Bahrain the au­thor­isa­tion at­taches to the act of pro­cessing and can­not be ob­tained ret­ro­spect­ively. Sep­ar­ate the event-log po­s­i­tion from the desktop-cap­ture po­s­i­tion in every doc­u­ment, be­cause the two are dif­fer­ent activ­it­ies in law. And treat the In­dian re­ten­tion floor as a design input rather than a com­pli­ance af­ter­thought, be­cause it runs against the in­stinct to de­lete dis­cov­ery data once the map is drawn.

One line be­longs in every re­gional po­s­i­tion and is usu­ally miss­ing. Name what is not in re­gion. The map­ping in­stru­ment has a United Arab Emir­ates home and the work­flow en­gine does not, and any min­ing tool a cli­ent already owns has a foot­print of its own that has to be es­tab­lished sep­ar­ately. A con­straint dis­closed in ad­vance is a design de­cision. The same fact found by a su­per­visor is a find­ing.

If the an­swer is a hunch rather than a num­ber, start with min­ing in­stead of auto­ma­tion. You are about to spend against a guess.

Own­er­less pro­cesses rot. If you can­not name the owner, you can­not hold the ac­cur­acy, and no tool fixes an ac­count­ab­il­ity gap.

If the pro­cess lives in a nota­tion only ana­lysts read, in­sight will stall at the ana­lyst's desk. Busi­ness-read­able is not a nicety; it is the dif­fer­ence between know­ing and doing.

Soft­ware bought and left un­op­er­ated is shelfware. De­cide up front whether you will run this cap­ab­il­ity your­self or have it de­livered by a part­ner.

The operator's read

The questions to ask before buying, the objections a board raises and the honest replies, and then the operating detail: what must be in place before discovery, the order the work runs in, the controls and how each is tested, the ways a discovery programme fails, and what to measure once the loop is running.

What to ask before you buy: a short interrogation for the C-suite

Be­fore com­mit­ting to any pro­cess-un­der­stand­ing pro­gramme, a C-level buyer should be able to an­swer six ques­tions, and if the an­swers are vague the pro­gramme is not ready. The ques­tions are de­lib­er­ately blunt. They sep­ar­ate a real dia­gnostic ex­er­cise from a tool­ing pur­chase dressed up as one, and any hon­est buyer should ask them be­fore com­mit­ting spend.

01 Which pro­cesses ac­tu­ally cost us the most?

02 How far is our doc­u­mented pro­cess from the real one?

If nobody has com­pared the map to the mine, the hon­est an­swer is "we do not know", and that is the first thing to fix.

03 Who owns each pro­cess, by name?

04 What would we point an AI agent at, and is that pic­ture true?

An agent in­her­its your pro­cess pic­ture. If the pic­ture is the de­signed map, the agent in­her­its its blind spots. Mine first.

05 Can the people who run the work read the pro­cess?

06 Who holds the li­cence, and who op­er­ates it?

The test

A pro­gramme that can an­swer these six with num­bers and names is ready to spend. One that an­swers with in­ten­tions and hopes should start with a dia­gnostic, which costs little and pre­vents the ex­pens­ive mis­take of auto­mat­ing the map.

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

A well-run board does not buy on en­thu­si­asm. The four ob­jec­tions below are the ones we meet most often, each reas­on­able and each an­swer­able, and 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 every case, stronger than the ob­jec­tion.

01 “We already have pro­cess maps.”

You al­most cer­tainly do, and they are al­most cer­tainly out of date in the places that cost money. The ques­tion is not whether you have maps but whether you have meas­ured them against real­ity. Most maps are sin­cere and wrong;

Process IntelligenceVoice of Volge18 •

min­ing tells you by how much. If the com­par­ison shows a small gap, you have spent little to con­firm good news. It rarely does.

02 “Min­ing sounds in­vas­ive: new agents on every ma­chine.”

Pro­cess min­ing reads event logs your sys­tems already keep; it re­cords noth­ing new. Desktop cap­ture does ob­serve how someone works, and it is op­tional, scoped, and de­ployed only where the logged sys­tems leave a gap. We are can­did that it touches how people work, which is ex­actly why we scope it nar­rowly and ex­plain it plainly rather than switch­ing it on every­where by de­fault.

03 “The ROI num­bers are old.”

The For­res­ter fig­ure is, and we flagged it as a late-2022 study spe­cific to Nin­tex Work­flow rather than to map­ping and min­ing. We do not lean on it as a fore­cast. Nor do we sub­sti­tute a cus­tomer story. The ones we checked are un­dated or self-re­por­ted, and they would not sur­vive the scru­tiny we are ask­ing you to apply. Your own dia­gnostic pro­duces a fig­ure spe­cific to you, and that is the num­ber worth ar­guing about.

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 con­nect­ing event logs, sort­ing real de­vi­ations from sens­ible ad­apt­a­tions, keep­ing own­ers en­gaged, and turn­ing a one-off dia­gnostic into an op­er­at­ing habit. That is judge­ment and per­sist­ence rather than a login, and it is where pro­grammes bought as soft­ware alone tend to stall.

The pattern in the answers

Each ob­jec­tion as­sumes the pur­chase is the soft­ware. It is not. The pur­chase is a truth­ful, main­tained un­der­stand­ing of how your work runs, and the dis­cip­line to act on it. The soft­ware is a means; the un­der­stand­ing is the thing you are pay­ing for.

One process, end to end: how a supplier-onboarding review actually runs

A single worked ex­ample shows the whole method more clearly than any amount of de­scrip­tion, so here is sup­plier on­board­ing, a pro­cess every large or­gan­isa­tion runs, and one that re­li­ably hides money. The fig­ures below are il­lus­trat­ive, drawn to match the shape min­ing com­monly re­veals rather than any one cli­ent. This is a walk­through, not a bench­mark.

Step 1, the map says three weeks, five steps

The doc­u­mented pro­cess is tidy. A re­quest leads to a credit check and a com­pli­ance re­view, then to ap­proval and ac­tiv­a­tion. In­ter­viewed, the team con­firms it. On the map, a sup­plier is live in about fif­teen work­ing days, and every­one be­lieves it.

Step 2: the mine says otherwise

Mined from the pro­cure­ment and fin­ance logs, the pic­ture changes. The me­dian case takes twenty-six days, not fif­teen. Com­pli­ance re­view car­ries most of the extra. Cases sit in a re­view queue for days be­fore any­one opens them. And there are not five steps but nine dis­tinct paths, be­cause two re­gions added their own checks that never reached the manual.

Exhibit 16 Supplier onboarding: map versus mine (illustrative)

MeasureMap (as designed)Mine (as performed)
Median cycle time≈ 15 working days≈ 26 working days
Distinct paths19
Largest delayNone assumedCompliance review queue
Control adherenceAssumed 100%Credit check skipped in ~1 case in 12

Source: il­lus­trat­ive com­pos­ite by Volge, 2026, of pat­terns com­monly seen in sup­plier-on­board­ing min­ing. Not a spe­cific cli­ent bench­mark.

Step 3, read the gap, then act

The gap sorts it­self. The com­pli­ance queue is the cycle-time prize, and the fix is a re­design plus a re­minder, not a bot. The skipped credit check is a con­trol ex­pos­ure to close be­fore it be­comes an audit find­ing. Two of the nine paths are sens­ible re­gional checks worth keep­ing; the rest are drift to stand­ard­ise. Only then, on the stand­ard­ised main path, does auto­ma­tion earn its build.

Step 4: close the loop

AI Model to Map re­turns the mined, stand­ard­ised pro­cess to Pro­cess Man­ager as a read­able pro­ced­ure the on­board­ing owner main­tains. Forms cap­tures the sup­plier's data cleanly at the front; DocGen pro­duces the on­board­ing pack at the end. The next min­ing run checks whether the fixes held. They usu­ally do, until the pro­cess drifts again, which is why the loop runs con­tinu­ously rather than once.

What must be in place before discovery starts

A map-and-mine en­gage­ment fails on ac­cess and au­thor­ity far more often than on method, and the gat­ing item is never the li­cence. Six con­di­tions have to hold be­fore the first ex­tract.

The second con­di­tion is where the weeks go. Event logs are owned by the teams that run the sys­tems, and an ex­tract is a

Process IntelligenceVoice of Volge19 •

re­quest to those teams rather than a con­fig­ur­a­tion change. Make the re­quest in week one.

  • A law­ful basis for the ex­tract, settled in writ­ing, to­gether with any prior au­thor­isa­tion the jur­is­dic­tion re­quires. Chapter five is where that ques­tion is con­strained.
  • Ex­tract au­thor­ity from the owner of each source sys­tem, with the fields agreed. A case iden­ti­fier, an activ­ity and a timestamp are the min­imum; an actor field is what makes the ana­lysis con­ten­tious and use­ful.
  • A named owner for each pro­cess in scope, with au­thor­ity to change it. An own­er­less pro­cess pro­duces a find­ing nobody can act on.
  • A re­ten­tion and de­le­tion po­s­i­tion for the dis­cov­ery data it­self, con­sist­ent with the sec­toral floor rather than with in­stinct.
  • An agreed scope for desktop cap­ture, if any, in­clud­ing the pop­u­la­tions, what is cap­tured and who may read the raw re­cord.
  • A de­cision on where the ana­lysis runs and where the ex­tract rests, be­cause in two of the jur­is­dic­tions in chapter five the ex­tract may not leave at all.

The sequence, and the exit test at each stage

The order of the five steps is fixed by de­pend­ency, as chapter one ar­gued. What chapter one did not sup­ply is the test that says each step is fin­ished, and without those tests a pro­gramme drifts from stage to stage on the cal­en­dar.

Exhibit The five steps, with the test that ends each one

StepWhat it producesThe exit test
01 DiscoverA process register with an owner against every entryEvery process in scope has a named owner who accepts the description. An entry nobody will own is a finding in itself
02 MapThe designed process, in business languageA person who runs the process reads the map and makes no correction. Until they do, the map is a draft
03 MineThe performed process, reconstructed from logs and captureCase counts reconcile against the source system's own totals for the period, within a stated tolerance
04 CompareA costed gap: variants, waits and deviations, rankedEvery item carries a volume, an elapsed-time cost and an owner. An unranked gap is a report
05 ActSimplification, automation or a deliberate decision to leave aloneA re-mine of the same process shows the change held. Nothing is closed on the strength of the fix alone

Source: Volge ana­lysis, 2026. The five steps re­state chapter one; the exit tests are the part worth copy­ing.

Step three car­ries the test most often skipped. If the mined case count does not re­con­cile against the source sys­tem's own totals, the ex­tract is in­com­plete, and every vari­ant count and cycle time de­rived from it is wrong by an un­known amount. Re­con­cile first, ana­lyse second.

The controls, and how each one is tested

Dis­cov­ery touches per­sonal data, op­er­a­tional sys­tems and the way people work, so its con­trols are as much about re­straint as about cov­er­age. Each of the con­trols below can be tested in an af­ter­noon.

Exhibit The control set, with the test for each

ControlWhat it isHow you test it
Extract scopeOnly the agreed fields leave the source system.Compare the extract schema against the agreed field list. Surplus fields are the ones a supervisor asks about.
Identifier handlingPerson-level identifiers are pseudonymised unless the analysis genuinely requires them.Ask which findings depend on knowing who performed a step. Usually two or three, and the rest do not need the field.
Extract reconciliationMined case counts agree with source-system totals.Re-run the comparison for one closed month and state the variance.
Retention and deletionDiscovery data expires on a stated schedule, subject to any statutory floor.Ask for the deletion date of the oldest extract still held, and whether it passed.
Capture consent and scopeDesktop capture runs only on agreed populations, with agreed content.Check the deployed configuration against the agreed scope. They diverge as the pilot expands.
Raw-capture accessThe raw record of an individual's working day has a narrower access list than the map derived from it.List who can read the raw capture today. If it is the same list as for the map, the control does not exist.
Process ownershipEvery mapped process has an owner who maintains it.Count maps with no edit in twelve months. That count is the decay rate.
Finding attributionEach gap item carries a volume and an owner before it is prioritised.Take the top five items and name the owner of each without consulting the build team.
Model-to-map reviewAutomatically converted maps are reviewed by an owner before they are published.Sample converted maps and check for an owner's edit. Automated translation smooths, so a mislabelled step reads as cleanly as a correct one.

Source: Volge ana­lysis, 2026. Con­trol names are gen­eric, so the set ap­plies whichever min­ing tool sup­plies the mine.

Process IntelligenceVoice of Volge20 •

The raw-cap­ture test is the one that sep­ar­ates a care­ful pro­gramme from a con­fid­ent one. Everything else on the list pro­tects the ana­lysis. That con­trol pro­tects the people being ana­lysed, and it is the reason a dis­cov­ery pro­gramme re­tains its man­date.

Failure modes, and what each looks like in production

Dis­cov­ery pro­grammes fail in five re­cog­nis­able ways, and the first four pro­duce a doc­u­ment rather than a change.

01 The un­re­con­ciled ex­tract

An ex­tract misses a sub­set of cases, usu­ally a re­gion or a chan­nel, and the ana­lysis pro­ceeds. In pro­duc­tion this looks like a con­fid­ent vari­ant count and a cycle time that op­er­a­tions does not re­cog­nise. The con­trol is the re­con­cili­ation at step three, and the symp­tom is an op­er­ator say­ing the num­bers feel wrong.

02 The find­ing with no owner

The gap is pro­duced, ranked and cir­cu­lated, and noth­ing hap­pens be­cause no entry has a name against it. In pro­duc­tion this looks like a well-re­ceived re­port that is cited for a year and acted on never. The con­trol is at­tri­bu­tion be­fore pri­or­it­isa­tion.

03 Map decay

Maps are cre­ated dur­ing the en­gage­ment and never ed­ited again, so within eight­een months the de­signed pro­cess is once more a fic­tion. In pro­duc­tion this looks like a healthy map lib­rary with no edit his­tory. The con­trol is the decay rate, coun­ted and re­por­ted, be­cause own­er­ship that is not meas­ured is own­er­ship in name.

04 Cap­ture scope creep

Desktop cap­ture is agreed for one team and ex­pands quietly to sev­eral, be­cause it is use­ful. In pro­duc­tion this looks like a suc­cess­ful pilot and a grow­ing con­sent prob­lem. The con­trol is the con­fig­ur­a­tion com­par­ison, run on a sched­ule rather than at the end.

05 Auto­mat­ing the stand­ard­ised main path too early

The main path is stand­ard­ised, de­clared stable and auto­mated, while the vari­ants it dis­placed are still being worked out by hand. In pro­duc­tion this looks like an auto­ma­tion cov­er­ing a shrink­ing share of volume. The con­trol is to re-mine be­fore build­ing, and to re­quire the vari­ant dis­tri­bu­tion to have settled across two con­sec­ut­ive peri­ods.

Gap clos­ure is the only one of the four that is hard to fake. A closed item re­quires a re-mine show­ing the change held, which is why pro­grammes that re­port clos­ure hon­estly re­port lower num­bers than pro­grammes that re­port in­tent.

What to measure once the loop is running

Four meas­ures tell you whether pro­cess un­der­stand­ing has be­come a habit rather than a pro­ject, and none of them is a count of maps.

MeasureDefinitionDirectionCadence
Map currencyShare of mapped processes edited within the last twelve monthsRising to a stated floorQuarterly
Variant concentrationShare of cases on the designed path, per processRising where standardisation was the decisionPer mining run
Waiting shareElapsed time spent in queues, as a share of total cycle timeFallingPer mining run
Gap closureItems from the ranked gap closed, with the re-mine confirming the change heldRising, and evidencedQuarterly

Source: Volge ana­lysis, 2026. Defin­i­tions are stated so that a fig­ure means the same thing in two con­sec­ut­ive quar­ters.

What done looks like

A pro­cess in­tel­li­gence pro­gramme is fin­ished, in the only sense that mat­ters, when four things are true on the day they are asked about. Every pro­cess in scope has an owner who has ed­ited its map this year. A mined pic­ture ex­ists for the pro­cesses that carry the most volume, re­con­ciled against source-sys­tem totals. The gap between the two is ranked, with an owner and a size against every item. And at least one closed item has been re-mined and shown to have held.

Three of those four can only be demon­strated after a second min­ing run. That is the hon­est reason a dis­cov­ery en­gage­ment should not be scoped as a single pass, and it is also why the first pass should be small enough to re­peat.

Process IntelligenceVoice of Volge21 •

Appendices

What independent research finds about automating blind

The ar­gu­ment of this book, see the pro­cess be­fore you auto­mate it, is not Volge's alone. In­de­pend­ent ana­lysts, work­ing from their own samples, re­port the same pat­tern. Pro­grammes that skip pro­cess truth stall, and the ones that scale are the ex­cep­tion.

Evidence Independent findings on automation and AI that outrun their process understanding

FindingFigureSource
Firms that had scaled RPA to 50 or more robots3%Deloitte Global RPA Survey, fieldwork Sep 2017 (n=424)
Digital transformations that fall short of objectives70%BCG, Flipping the Odds, 29 Oct 2020
GenAI projects forecast to be abandoned after proof of concept by end-2025At least 30%Gartner, 29 Jul 2024
Organisations getting zero return from enterprise GenAI~95%MIT NANDA, The GenAI Divide, Jul 2025
US C-suite respondents calling their GenAI rollouts “mature”1%McKinsey, Superagency in the workplace, 28 Jan 2025

Sources in­de­pend­ent of Nin­tex and Volge; full cita­tions in the Sources annex.

Read down the column and one story re­peats. The tool­ing scales faster than the un­der­stand­ing be­neath it, and the gap is where the money leaks. De­loitte's 3 per cent, from field­work in Septem­ber 2017, is the share of firms that had scaled RPA to fifty or more ro­bots; the rest stalled, and De­loitte's Novem­ber 2018 sur­vey put the fig­ure at 4 per cent. The most com­mon reason, across these stud­ies, is that the pro­cess the soft­ware was poin­ted at was not the pro­cess the busi­ness ac­tu­ally ran.

The 2025 fig­ures sharpen the point, with two qual­i­fic­a­tions. Gart­ner fore­cast on 29 July 2024 that at least 30 per cent of GenAI pro­jects would be aban­doned after proof of concept by the end of 2025, and has pub­lished no out­turn, so that re­mains a lapsed fore­cast rather than a res­ult. The NANDA paper of July 2025 found that 95 per cent of or­gan­isa­tions, not of pi­lots, were get­ting zero re­turn. Both point the same way. Agents are being aimed at op­er­a­tions no one has meas­ured against real­ity, which is the read­i­ness layer this book is about.

Why this page is here

A ser­i­ous ar­gu­ment should sur­vive con­tact with sources that have no stake in it. Every fig­ure above comes from a body in­de­pend­ent of both the vendor and its part­ner. One needs a caveat. The GenAI Di­vide car­ries the MIT NANDA masthead

but is the work of its four named au­thors, who dis­claim in­sti­tu­tional at­tri­bu­tion, and it has not been peer-re­viewed.

Methodology & editorial standards

This book fol­lows a small set of rules de­signed to keep a dis­closed-in­terest ana­lysis hon­est, and we state them so you can hold us to them. Every ex­ternal fig­ure is at­trib­uted and dated. Vendor po­s­i­tion­ing is la­belled as po­s­i­tion­ing. Mar­ket siz­ing is la­belled as es­tim­ate. Il­lus­trat­ive fig­ures are la­belled as il­lus­trat­ive. Our com­mer­cial in­terest is dis­closed wherever it bears on the ar­gu­ment.

On the figures

Where a num­ber comes from Nin­tex, we say so and give the date. Where it comes from a third party, Mar­ket.us, For­res­ter, G2, TrustRadius, we name the source. The For­res­ter ROI fig­ure is flagged as a study con­duc­ted in late 2022 and spe­cific to Nin­tex Work­flow, not to the map­ping cap­ab­il­ity that is this book's sub­ject. The mar­ket-size fig­ure is a Mar­ket.us es­tim­ate of Septem­ber 2025, and the in­ter­pol­ated years in Ex­hibit 2 are our own arith­metic at the stated growth rate, marked as such.

On the illustrative exhibits

Ex­hib­its 10 and 12, where time is lost and how vari­ants dis­trib­ute, are il­lus­trat­ive com­pos­ites rather than meas­ure­ments of a named cli­ent. They show the shape min­ing typ­ic­ally re­veals, and they are la­belled so no reader mis­takes a rep­res­ent­at­ive pat­tern for a bench­mark. Your own num­bers are the only ones that mat­ter, and find­ing them is the first piece of work.

On what we do not claim

We make no claim that Nin­tex is a Gart­ner Magic Quad­rant leader. For­res­ter placed Nin­tex as a Strong Per­former, not a Leader, in The For­res­ter Wave: Agen­tic Pro­cess Or­ches­tra­tion Plat­forms, Q3 2026, and we name the place­ment rather than imply more than it says. We do note that pro­cess min­ing is an ana­lyst-re­cog­nised cat­egory, and Gart­ner has pub­lished a Magic Quad­rant for Pro­cess Min­ing Plat­forms (April 2025). We de­scribe Pro­cess In­tel­li­gence as Nin­tex po­s­i­tion­ing rather than a product, and we re­cord that the min­ing half of it is now end-of-life.

Our conflict, stated plainly

Volge is a Nin­tex de­liv­ery part­ner and be­ne­fits if you buy. We have handled that con­flict by sourcing every ex­ternal claim in­de­pend­ently, in­clud­ing the ana­lyst fig­ures on page 31, which come from De­loitte, BCG, Gart­ner, the MIT NANDA au­thors and McKinsey, none of whom have a stake in this book, la­belling our es­tim­ates, and keep­ing the first step small.

Process IntelligenceVoice of Volge22 •

Corrections

If any fig­ure here has aged or any at­tri­bu­tion is wrong, we would rather know. A field guide is only use­ful while it is ac­cur­ate, and this is a first edi­tion.

Independent sources (page 31)

De­loitte Global RPA Sur­vey (field­work Septem­ber 2017, n=424) found that 3% had scaled RPA to fifty or more ro­bots; its Novem­ber 2018 sur­vey put the fig­ure at 4%. BCG, Flip­ping the Odds of Di­gital Trans­form­a­tion Suc­cess (29 Oc­to­ber 2020), found 70% of di­gital trans­form­a­tions fall short of ob­ject­ives. Gart­ner (29 July 2024) fore­cast that at least 30% of GenAI pro­jects would be aban­doned after proof of concept by end-2025, and has pub­lished no out­turn. MIT NANDA, The GenAI Di­vide (July 2025), found ~95% of or­gan­isa­tions get­ting zero re­turn from en­ter­prise GenAI; the paper is not peer-re­viewed and its au­thors dis­claim in­sti­tu­tional at­tri­bu­tion. McKinsey, Su­per­agency in the work­place (28 Janu­ary 2025), found 1% of US C-suite re­spond­ents call­ing GenAI rol­louts “ma­ture”. Gart­ner Magic Quad­rant for Pro­cess Min­ing Plat­forms (April 2025). None have a stake in this book.

Figures generated by Volge

Volge-ana­lysis ex­hib­its and the il­lus­trat­ive com­pos­ites (la­belled in­line) are our own; in­ter­pol­ated mar­ket years are Volge arith­metic at the stated CAGR.

The reader's checklist

Tear-out ver­sion. Six ques­tions from the op­er­ator's read (page 27), on a single page you can take into a meet­ing.

Checklist Before you spend on automation or AI

AskIf the answer is vague
Which processes cost us most?Mine before you automate
How far is the map from reality?Run the comparison first
Who owns each process, by name?Assign ownership before tooling
Is our AI target picture true?The agent inherits your blind spots
Can operators read the process?Insight will stall at the analyst's desk
Who holds and operates the licence?Decide, or it becomes shelfware

Source: Volge ana­lysis, 2026. See page 27.

What you need in place to map and mine

Pro­cess in­tel­li­gence is the light­est of the three lay­ers to buy, but not free of pre­requis­ites. Two lists fol­low. The first names the Nin­tex ele­ments to li­cence, the second the non-Nin­tex pieces that must already be in place for a map-and-mine

en­gage­ment to pro­duce any­thing. Pro­cess min­ing is not among the Nin­tex ele­ments, be­cause Nin­tex Pro­cess Dis­cov­ery is end-of-life, so a min­ing tool, where one is needed, comes from else­where. Tiers are in­dic­at­ive for 2026; con­firm cur­rent SKUs with Nin­tex.

Nintex elements to licence

Nintex Process Manager (Promapp)The mapping and operating-procedure library. Platform subscription, priced by user.
Nintex ProcessesProcess management native to the CE Platform, launched 26 October 2026. Nintex calls it the next evolution of Process Manager.
Process CaptureDesktop activity capture into draft maps. Included in Expert and Enterprise tiers; an add-on elsewhere; not offered in the AEN region.
AI Model to Map ConverterConverts BPMN models into readable procedures. GA December 2025. Expert and Enterprise tiers; an add-on on Professional.
Process Manager MCPRead-only process knowledge for AI assistants. Open beta; no charge published.

What else must be in place (non-Nintex)

Source-system event logsERP, CRM, or ITSM logs, or their exports. You cannot mine what you cannot extract.
Identity / single sign-onMicrosoft Entra ID or a SAML provider, for user access.
Data-governance clearanceSign-off to pull operational logs, with privacy and retention rules agreed.
Process owners' timeNamed owners to validate the mined maps against how work really runs.

The point of this page

Noth­ing here needs a server or a data-centre. The gat­ing item is not a li­cence, it is ac­cess to the logs, and the own­ers' time to read what the mine finds.

Process IntelligenceVoice of Volge23 •
Process Intelligence

The United Arab Emirates appears on the Nintex status register for Process Manager production and Nintex Analytics, and for nothing else.

Reference

Glossary

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

TermDefinition
Process as designed (the map)How a process is meant to run, as documented. Held in Nintex Process Manager.
Process as performed (the mine)How a process actually ran, reconstructed from event logs and desktop capture.
Nintex Process ManagerBusiness process mapping and modelling. Formerly Promapp; rebranded 2022. Tagline: identify, map, and model your processes.
Nintex ProcessesProcess management native to the Nintex CE Platform, launched 26 October 2026. Nintex calls it the next evolution of Process Manager, not a replacement.
Nintex Process IntelligenceNintex's positioning term for mining, task mining, modelling and mapping, never a confirmed standalone SKU. The mining half was delivered by Nintex Process Discovery.
Nintex Process DiscoveryThe retired mining and task-mining product. End-of-life announced 4 May 2026: end-of-life 1 May 2027, support ends May 2028, customers migrated to Process Manager.
Process miningReconstructing the real sequence and timing of a process from event logs.
Event logThe timestamped record enterprise systems keep of what happened: the raw material of mining.
BPMNBusiness Process Model and Notation, the technical notation mining produces, which AI Model to Map renders business-readable.
AI Process GeneratorDrafts a process map from a description. Beta March 2025; GA June 2025, Professional tier and above.
Process CaptureActivity recording that auto-documents a task as performed. GA September 2025, release 4.12.22.0.
AI Model to MapAI Model to Map Converter in the product. Converts technical BPMN into business-readable procedures. Beta September 2025; GA December 2025, release 4.12.27.0. Bridges mine to map.
Process Manager MCPA Model Context Protocol server exposing approved process knowledge to AI assistants, read-only. Open beta from 13 August 2026; authentication for eleven clients in release 4.26.23, October 2026.
Nintex DocGenGenerates documents from process data. An enhancer.
Nintex FormsCaptures structured data into a process. An enhancer.
The gapThe difference between the designed map and the mined reality, variants, bottlenecks, deviations, and the costed worklist it forms.
VariantA distinct path cases actually take through a process, often undocumented.
IPAIntelligent process automation: the market category sized in Exhibit 2.
Process IntelligenceVoice of Volge24 •
Sources

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

  • Market.us (September 2025 edition): Intelligent process automation market estimated at US$16.81bn (2024), projected US$61.23bn by 2034 at 13.8% CAGR. Labelled throughout as an estimate.
  • Nintex, Niranjan Vijayaragavan, then CPTO: “The AI entry point no one is talking about: Process intelligence”, Forbes, 3 February 2026. An article title, not a spoken quotation. Five9 appointed him chief technology officer effective 29 June 2026.
  • Nintex (2022): Rebrand of Promapp to Nintex Process Manager confirmed.
  • Nintex: AI Process Generator, beta March 2025 (4.12.10.0), GA June 2025 (4.12.16.0), Professional tier and above.
  • Nintex: Process Capture, GA September 2025 (4.12.22.0); Expert and Enterprise tiers, an add-on elsewhere, not offered in the AEN region.
  • Nintex: AI Model to Map Converter, beta September 2025 (4.12.21.0), GA December 2025 (4.12.27.0). 6 January 2026 is the date of PEX Network's news story, not of the release.
  • Nintex, Chris Ellis, Director of Solution Engineering: “This is a major step forward in making process intelligence accessible, actionable, and scalable across the organization.” Reported by PEX Network, 6 January 2026, on the general availability of AI Model to Map.
  • Nintex community (4 May 2026, updated 17 August 2026): Nintex Process Discovery end-of-life 1 May 2027, end of support May 2028, customers migrated to Nintex Process Manager.
  • Nintex Process Manager release notes (October 2026, release 4.26.23): MCP authentication for ChatGPT, Claude, Cursor, Gemini, GitHub Copilot, LibreChat, Make.com, Microsoft 365 Copilot, Copilot Studio, Salesforce Agentforce and Windsurf. Open beta announced 13 August 2026.
  • IT Brief (26 October 2026): Nintex Processes launched as process management native to the CE Platform; Forrester named Nintex a Strong Performer in The Forrester Wave: Agentic Process Orchestration Platforms, Q3 2026.
  • Nintex company page (modified September 2026): more than 7,000 organisations in over 100 countries. The newsroom boilerplate of 22 October 2026 drops the count.
  • G2 (November 2026): 4.2 / 5 across 1,326 reviews.
  • TrustRadius (November 2026): 8 / 10 across 530 reviews; Top Rated for a seventh consecutive year in 2025, and a 2026 award thereafter.
  • Nintex solutions pages (modified July 2026): L'Oréal, Panasonic, Sonos and CARFAX appear as customer logos, with no published case study behind them.
  • Forrester: Total Economic Impact of Nintex Workflow, study conducted late 2022: 213% ROI, nine-month payback. Specific to Nintex Workflow.
Process IntelligenceVoice of Volge25 •

Volge and our expertise on process intelligence

Volge pro­duces an hon­est ac­count of how a cli­ent's work ac­tu­ally runs, and then works the dis­tance between that and the de­signed pro­cess. The prac­tice sits where this book loc­ates the money, in the gap between the map and the mine, which is to say in the un­doc­u­mented vari­ants, the queues that hold most of the elapsed time, and the con­trols quietly skipped at scale. De­liv­ery is by senior pro­cess and data prac­ti­tion­ers who run the ex­tract and the ana­lysis them­selves, in­side the cli­ent's own en­vir­on­ment.

The work has a usual shape, and it fol­lows the five steps of chapter one with the exit tests of chapter six at­tached. Dis­cov­ery builds the re­gister and names an owner against every entry. Map­ping pro­duces a doc­u­ment a per­son who runs the pro­cess reads without cor­rect­ing. Min­ing re­con­structs the per­formed pro­cess and re­con­ciles its case counts against the source sys­tem's own totals be­fore any find­ing is drawn from it. The com­par­ison then pro­duces a ranked gap in which every item car­ries a volume, an elapsed-time cost and an owner. Act­ing on it closes with a re-mine, be­cause a fix nobody has re-meas­ured is a hope.

Nin­tex no longer sells a min­ing product, and that changes the sourcing rather than the method. Nin­tex Pro­cess Dis­cov­ery was placed end-of-life on 4 May 2026, with end-of-life on 1 May 2027 and sup­port end­ing in May 2028. So the mine comes from Pro­cess Cap­ture at the desktop, from a sampled log ana­lysis run by hand, or from a third-party min­ing tool a cli­ent already owns. The map and the com­par­ison stay in­side Nin­tex Pro­cess Man­ager, now sur­fa­cing as Nin­tex Pro­cesses.

Who does the work. A de­ploy­ment strategist holds the busi­ness out­come and the gov­ernance ca­dence with the cli­ent's op­er­a­tions lead­er­ship. A trans­form­a­tion ar­chi­tect owns the method, the ex­tract scope and the re­con­cili­ation, and signs the find­ings. Pro­cess ana­lysts work along­side the cli­ent's own pro­cess own­ers rather than in place of them, be­cause a map main­tained by a con­sult­ant de­cays on the day the en­gage­ment ends. The same people stay from the first ex­tract through the re-mine, which is what makes the con­trol tests in chapter six 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 se­quen­cing set out in chapter five is ap­plied rather than de­scribed. The law­ful basis and any prior au­thor­isa­tion are settled be­fore the first ex­tract, which in Bahrain means es­tab­lish­ing whether Art­icle 15 of the Per­sonal Data Pro­tec­tion Law reaches the design. Clas­si­fic­a­tion comes be­fore dis­cov­ery in Oman and Saudi Ar­a­bia, be­cause there the clas­si­fic­a­tion de­cides whether the ex­tract may leave at all. And the event-log po­s­i­tion is doc­u­mented sep­ar­ately from the desktop-cap­ture po­s­i­tion, be­cause the two are dif­fer­ent activ­it­ies in law.

An en­gage­ment starts with a scoped dia­gnostic on one pro­cess, usu­ally three to five weeks. It pro­duces a map the pro­cess owner ac­cepts, a mined pic­ture re­con­ciled against source-sys­tem totals, a ranked gap with an owner and a size against each item, a stated po­s­i­tion on the law­ful basis and the res­id­ency of the ex­tract, and a re-mine date. Volge is a Nin­tex 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.

Process IntelligenceVoice of Volge26 •
Version · VVN-PRI-006 · 2026
volgeEnable Powerful
Transformations
About Volge

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

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

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

Terms of use

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