A machine-learning productivity system went live across 32 government entities this month. The agentic AI it will eventually judge has completed one publicly announced transaction.
A productivity system built before the technology it will judge spends its first years measuring the people who are waiting, and their numbers become the baseline the machines are scored against.
Issue 020 followed the chain to where it breaks, 021 found the aggregate that hides its own composition, 022 left the corridor for the one country arguing about who pays, 023 watched a duty get assigned before its scope was defined. 024 stays with the machinery and asks what has actually been delivered.
Start with the one that got counted. On 23 July, Ajman renewed a trade licence using an agentic AI system, and it was reported as a first for United Arab Emirates government. Six weeks later it is still the only completed autonomous government transaction announced anywhere in the country. We looked hard for a second, because a lead that rests on an absence has to earn it: every result returned was the same July event carried by a different outlet. The nearest thing to a falsifier is a Khaleej Times piece of 20 May announcing four government AI agents for procurement, tax auditing, customer happiness and technical support, and it predates Ajman and announces initiatives rather than completed transactions.
Two things about that transaction are worth holding onto, because they are the difference between a fact and a slogan. There is no number attached to it. Not licences renewed, not transactions completed, not users served, in any account we could find, and none on the federal portal either, which is where a published result would sit. And the phrase doing most of the work in the coverage, that this is a proactive, headless government service model in which a transaction is fully automated, belongs to Ajman's authorities rather than to a reporter. Khaleej Times' own description of the same system has the agent identify customer needs, prepare the service journey, request approval and coordinate procedures, with human oversight where required. There is an approval step in the middle of the fully automated transaction.
Now the thing that did ship. On 23 August Dubai adopted a unified Workforce Productivity Measurement System across 32 government entities, built by the Dubai Government Human Resources Department under Executive Council Resolution No. 67 of 2025, using, in the release's own words, analytical tools and digital platforms powered by machine learning and advanced artificial intelligence. It was five years in development. What it measures is the relationship between human resources, financial resources and government service outputs, entity by entity. We checked the release for the words individual, employee performance and staff performance, and none of them is in it, so nobody should read this as per-employee surveillance and we are not writing it that way.
Put the two beside each other and the shape of the week appears. The instrument that will judge the productivity of government work is live across 32 entities and took five years to build. The instrument that is supposed to do the work has one completed transaction to its name, no volume figures, and a task-classification framework that has not been published. The federal portal lists a practical guide for designing agentic government services, in Arabic only, and lists no framework, no liability rules, no recourse rules, no completed services and no performance metrics. Five years for the meter. Eleven weeks and counting for the machine.
1
On 23 August Dubai adopted a machine-learning system to measure workforce productivity across 32 government entities, five years in the building and backed by an Executive Council resolution. The agentic AI that system will eventually have to account for has one publicly announced completed transaction.
The argument here is about order of arrival, and it needs one distinction made carefully at the outset. The 90 days everyone is counting is not a delivery deadline. The programme launched on 10 June with a 90-day sprint for each entity to create its own agentic AI service, and the clearest account of what that means describes a window for identifying which services and procedures suit agentic AI. That is scoping. No source states an end date, and the date that has been circulating is an inference from a launch, so we do not print it. Anyone attacking the United Arab Emirates for missing a deadline is attacking a deadline nobody set.
What can fairly be observed is a sequence. Since June the programme has produced governance material: a launch workshop with more than 300 participants from 50 federal entities, an executive committee, a national committee, and in late August a unified guide for designing government services powered by agentic AI, released at a workshop attended by more than 100 officials. That guide has no English title. The only formal title anywhere is on the federal portal and is in Arabic, which is worth knowing before anyone cites it as a named document.
And the thing that would let an agent actually act has not appeared. The sharpest account of the gap puts it as a conditional rather than an accusation, and we quote it that way: the task classification frameworks are the document to watch, and when they are published they will be the first serious attempt by any state to write down which decisions a government is prepared to let a machine make on its behalf. The same analysis notes that the material released so far does not describe what a citizen does when an autonomous system gets their case wrong, or where liability lands when it does. Governance about how to design the service exists. Governance about what the machine may decide does not.
So the meter arrives first, and it is worth being precise about what it meters. Dubai's system measures entities, not people, and it measures the relationship between staff, money and service outputs. Today the staff in that relationship are civil servants. The agents are not in the denominator, because on the public record there is one of them and it has done one thing. A productivity instrument that predates the productivity technology will spend its first years measuring the humans who are waiting for it, and whatever it finds about them will be the baseline the machines are eventually judged against.
Building the machine.
Building its meter.
Governance about how to design an agentic service exists. Governance about what the machine may decide does not. In between them a measurement system went live across 32 entities, and what it meters today is the civil servants who are still doing the work by hand.
Our claim is narrow, and the boundaries matter more than usual because this rests partly on absences. We are not claiming the United Arab Emirates has failed at anything. There is no missed deadline, because no deadline was published. We are not claiming that building a measurement system before a delivery system is a mistake; a case can be made that it is the right order, and that a government which can measure its own service outputs is better placed to judge what an agent is worth than one which cannot.
We are claiming that the machinery of measurement, design and committee is shipping considerably faster than the transactions, that the gap is now visible enough to name, and that the one instrument which would settle what an autonomous system is allowed to decide is the one that has not been published. That is a statement about sequence, and it is checkable in both directions. If the classification framework appears next week, we were early. If a second entity announces a completed transaction, the count of one moves and we will move it.
The honesty screen. Four of this issue's load-bearing findings are negatives: one transaction, no second entity, no volume figure, no end date. Negatives are the easiest thing in journalism to get wrong, so here is how they were established. Each was searched for its own falsifier rather than for its confirmation, the federal portal was read directly because it is where a published result would live, and the searching is described in the register so a reader can judge the effort rather than take our word for the conclusion. If any of the four is wrong, it is wrong because something exists that we could not find, and that is the failure mode we would want reported to us.
One more thing belongs in the verdict rather than in a panel, because it cuts against the whole week's reading. The single most-cited measure of AI's effect on employment just went quiet. Challenger's August report puts AI fourth among stated reasons at 3,462 cuts, its lowest month since December 2025, ending a five-month run as the leading monthly reason. The report is careful, and so are we: AI remains the leading reason year to date at 116,175. But the man who compiles that series told us in July that as regulations take shape companies will be even more careful in their announcements, which would make tracking the impact of AI on jobs more opaque. We printed that warning in advance. We cannot yet tell you which of the two things August was.
Five years for the meter. Eleven weeks and counting for the machine.
CEO
Productivity allocation
CHRO
Workforce transition
CFO
Economic attribution
BOARD
Governance threshold
A government has committed to putting agents into public services and has published how to design them. What it has not published is the classification that says which decisions a machine may take. Most organisations are in the same position with less excuse, because nobody is asking them for the list yet.
One completed autonomous transaction, no volume figure, and no published framework for what the next one may decide.
Ajman renewed a trade licence with an agentic system on 23 July and it remains the only one announced in the country. Even there the account includes an approval step: the agent identifies the need, prepares the journey, requests approval and coordinates procedures.
If nobody in the organisation can produce that list, the honest answer is that the boundary is wherever the vendor put it.
The roles below come out of a specific gap: a government has committed to putting agents into public services, has published how to design them, and has not published what they are allowed to decide. Every job in that gap is about writing down a boundary, and none of them existed as a titled function two years ago.
The first two are the boundary itself. Somebody has to author a task-classification framework, which means deciding, in writing and in advance, which decisions a state will let a machine make on its behalf. Somebody else has to own what happens when it gets one wrong, because the material published so far does not describe what a citizen does in that case or where liability lands. Those are not the same job and in most administrations neither is filled.
The middle pair come from the meter. A productivity system spanning 32 entities that measures the relationship between people, money and outputs needs analysts who can read it without over-reading it, and it needs somebody who can tell the difference between an entity-level relationship and an individual's performance, because the second thing is what everyone will assume it measures. Dubai's release is careful on exactly this point. Whoever operates the system will spend a good deal of time being careful about it in public.
The last two are about the record rather than the work. As the AI label fades from layoff announcements, attributing a reduction becomes a research task rather than a reading task, and somebody inside every large employer will end up doing it. And on the sovereign side, the deals signed in Riyadh this week bundle agentic layers into productivity software for a target of a million users across the Middle East and Africa, with no employment figure attached to any of them. Somebody has to sit between a partnership announcement and a payroll, and at the moment nobody does.
6 rising role categories, each with a sourced hiring signal.
The framework that would say which decisions a government lets a machine make on its behalf is described as the document to watch and has not been published. Somebody has to author the equivalent inside every large organisation.
The material released so far does not describe what a citizen does when an autonomous system gets their case wrong, or where liability lands when it does. That is an unfilled role rather than an unanswered question.
Dubai's new system measures the relationship between human resources, financial resources and service outputs across 32 entities. Reading that without over-reading it into individual performance is a job, and the release is careful about the difference.
Ajman's agent identifies customer needs, prepares the service journey, requests approval and coordinates procedures, with human oversight where required. Somebody owns that approval step, and at scale it becomes a control function rather than a courtesy.
Artificial intelligence fell to the fourth most cited reason for US job cuts in August at 3,462, ending a five-month run at the top. As the label fades, attributing a reduction stops being a reading task and becomes a research one.
Microsoft and HUMAIN will offer an agentic layer bundled with Microsoft 365 Copilot, initially targeting one million users across the Middle East and Africa, with no employment figure in the release. Somebody has to sit between that announcement and a payroll.
Named HUMAIN
One set of figures, measured one way, so there is nothing here to compare it against.
AI-attributed announced job cuts
This week's signal through the India, Middle East and Africa lens.
BUILDING
Signal
Dubai adopted a machine-learning Workforce Productivity Measurement System across 32 entities on 23 August, five years in development. The federal agentic programme produced a design guide and a workshop of 100-plus officials. Ajman's July trade licence remains the only completed autonomous transaction announced.
Why it matters
The region measuring government work hardest is the one with the least of it done by machine so far.
BUILDING
Signal
The one crore AI skilling pledge reached day twenty with no scheme, no ministry and no budget. The only delivery announced is a ministry release covering 150,000 learners in foundational literacy and 10,000 in job-ready training. ICRIER finds no evidence of large-scale job losses.
Why it matters
Twenty days on, the largest skilling pledge in the region still has no owner, no scheme and no money.
EMERGING
Signal
At GITEX Nigeria the government restated 3MTT as 1.87 million registrations and over 125,000 trained, with no placement figure. That 125,000 is lower than a 135,000 ministerial figure from February. Ghana ran train-the-trainers for One Million Coders against 12,623 completions recorded in May.
Why it matters
Two national programmes publish registrations and completions, and neither publishes a placement.
Short read · this week's signal across the nine sectors we cover
This is where the sequence is clearest. Dubai adopted a unified Workforce Productivity Measurement System on 23 August across 32 government entities, built over five years by the Dubai Government Human Resources Department under Executive Council Resolution No. 67 of 2025, using analytical tools powered by machine learning to measure the relationship between human resources, financial resources and service outputs. The federal agentic programme, meanwhile, has produced a launch workshop, an executive committee, a national committee and a design guide, and one completed transaction, in Ajman, on 23 July.
The most-cited measure of AI's effect on employment moved sharply. Challenger's August report puts total announced US cuts at 52,881, up 58% on July, with restructuring leading at 16,173. Artificial intelligence fell to the fourth most cited reason at 3,462, its lowest monthly total since December 2025, ending what the report calls a five-month run beginning in March in which AI was the leading monthly reason. Technology announced 6,103, its lowest month of 2026, against a year-to-date 155,126 that is up 52% on 2025.
LEAP 2026 in Riyadh opened with more than 15 billion dollars in investments and partnerships: an AWS and HUMAIN AI Zone at about 5 billion, xAI scaling from 50 to 500 megawatts, an AMD inference cluster, and HUMAIN's own Arabic model and voice products. Microsoft and HUMAIN announced an agentic layer bundled with Microsoft 365 Copilot, initially targeting one million users across the Middle East and Africa. Checked deal by deal rather than in aggregate: not one states a hiring, headcount or job-creation figure.
Hyundai Motor's union ratified its wage agreement in a vote held on 31 August and announced the next day, with 19,183 of 31,166 ballots cast in favour, 61.55%, on a turnout of 78.63% of 39,638 eligible members. The AI demand that opened this thread three issues ago closes as a consultation right. The agreement calls the introduction of physical AI and robotics essential to the company's long-term competitiveness and survival, with commitments to share developments transparently and manage the transition together, and 500 new production workers from 2027.
The skilling ladder lost a rung in public. At GITEX Nigeria on 1 September the Secretary to the Government of the Federation restated 3MTT as 1.87 million registrations across all 774 local government areas and over 125,000 trained, with no placement figure at all. That 125,000 is lower than the 135,000 a minister gave in February and which was still circulating in August. India's ministry release of 23 August pairs 150,000 learners in foundational AI and cloud literacy with 10,000 in job-ready training, a ratio of fifteen to one.
Platform work produced the sharpest institutional move of the week. Uber, Eternal and Porter withdrew from Karnataka's Platform-Based Gig Workers Welfare Board, saying they did not wish to remain part of a statutory body created under a law they are challenging in the High Court. The state recruited Delhivery, Namma Yatri and Yulu to replace them. Six platforms deposited about 4 crore rupees of April to June welfare fees with the High Court rather than with the board, answering a July direction from the court. Yulu paid the board directly.
The attitude data arrived before the outcome data here too. A BCG X survey published on 2 September reports that 60% of frontline employees and 66% of managers and leaders in the Gulf believe AI agents could perform at least half of their current job responsibilities within the next three years, and that 93% use AI at least several times a week. The countries surveyed are the United Arab Emirates, Saudi Arabia, Kuwait and Qatar. It is the only new workforce datapoint the region produced in the window.
A single report became a week of headlines. Around 1 September roughly 3,000 Oracle job cuts in India were reported, and every account located traces to one Economic Times story: two outlets credit it explicitly and the rest attribute to industry estimates, reports or people familiar. Oracle has not confirmed the cuts. Its own annual filing states approximately 141,000 full-time employees as at 31 May 2026, of whom about 92,000 are outside the United States, and carries no comparative headcount for the prior year.
Worker protection in the AI supply chain stays open from 019 through 023 with nothing to add. The one thread that moved is the Karnataka litigation, and it moved sideways rather than forward: money is accumulating with the court registry under a July direction while the platforms that are paying it have walked off the welfare board created by the law they are challenging. Three weeks after the hearings listed for 14 August and on or after 24 August, no outcome of either has been published anywhere we can find.
For Editor reAImagine · curated to this issue's signal · 90-day horizon
Why now
A government has published how to design agentic services and not what they may decide. The task classification owner is whoever writes that list.
Do this
Why now
The published material does not say what a citizen does when an autonomous system errs. The automated recourse owner is the role that gap creates.
Do this
Why now
Dubai's system measures the relationship between people, money and outputs across entities and not individuals. The workforce productivity analyst holds that line in public.
Do this
Why now
AI fell to fourth among stated reasons in August, and the series compiler warned the label would fade. The AI attribution analyst replaces it.
Do this
Why now
More than fifteen billion dollars of deals were announced in Riyadh without one hiring figure between them. The sovereign AI partnership manager is who asks.
Do this
Nothing scores this issue, which makes seven consecutive, and the end is now three weeks out rather than five. No entry's resolve-by date has passed or falls in the coming week. LEDGER-001-03 scores on 30 September and LEDGER-001-05 on 1 October, so the issue published on or after 30 September is the one that finally scores, and 004-02 on 31 October and 002-03 on 30 November follow quickly. Two entries move in substance without moving in status. LEDGER-002-02 got its August data and it cuts against the entry: Challenger puts AI-attributed cuts at 3,462 in August, the lowest monthly total since December 2025, ending a five-month run as the leading monthly reason, which takes the second half to 14,432 against a first half of 101,743. September to December would need roughly 87,000 more to clear the bar, against an August of three and a half thousand. That is a strong miss trajectory, and we are going to say the uncomfortable part rather than bank the win: this is exactly the degradation we pre-registered in 022, when the compiler of the series warned that as regulations take shape companies will be even more careful in their announcements. The entry cannot yet distinguish a fading phenomenon from a fading label, and a miss scored on a number that stopped being reported is not the same as a miss scored on a number that fell. LEDGER-005-02 gains the best intelligence it has had in weeks and it is not a hearing outcome. Uber, Eternal and Porter have withdrawn from the Karnataka welfare board while challenging the Act that created it, the state has recruited Delhivery, Namma Yatri and Yulu to replace them, and six platforms have deposited about 4 crore rupees with the High Court rather than with the board. The platforms are paying the court and walking off the machinery, which is a different posture from either compliance or defiance, and three weeks after two listed hearings there is still no published outcome of either. LEDGER-001-03 records a fourth consecutive all-single-vendor week and continues to trend toward a miss on 30 September. No new series opens: nothing in the window carries a date we could hang a resolve-by on without inventing one.
Anthropic's ID-verification policy takes effect and, whatever its stated intent, functions in practice as a citizenship-sorted access path: US consumers regain restricted-tier access first, with no announced parity path for Indian or GCC passport holders. Anthropic says the change is an unrelated appeals update; we forecast the observable outcome and will score it.
Scored 9 July 2026. Fable 5 came back for every consumer on earth on the same day, 1 July, because the US Commerce Department lifted the export controls on 30 June. The restoration ran through diplomacy, not identity checks, and it landed a week before the ID policy took effect on 8 July. The policy itself verifies identity and age for flagged consumer accounts, carries no nationality component at all, and exempts Team, Enterprise and API customers. The disconfirming evidence we carried inside the entry, Anthropic's statement that this was an unrelated appeals update, held up better than our forecast did. To score this a hit we needed restricted access re-sorted by passport through the verification flow. It was not.
At least one further US frontier-model release goes through government pre-release review rather than open launch, extending the pattern already visible in June.
Scored 16 July 2026, a day early, because the pattern resolved ahead of the date. OpenAI previewed GPT-5.6 with the US government for about a month, released it on 26 June as a limited preview to around 20 government-approved organisations, and only opened it to the public on 9 July after a federal evaluation window under Executive Order 14409's voluntary pre-release framework. That is government pre-release review rather than open launch, exactly as forecast. The honest complication belongs on the record: the White House publicly denied giving any green light, approval or clearance, and EO 14409 explicitly bars mandatory licensing or preclearance. The claim required review, not approval; review demonstrably happened, so the hit stands on the wording as published.
At least one of TCS, Infosys, Wipro or HCLTech publicly announces a formal multi-model or sovereign-fallback architecture policy as strategy, not as a procurement footnote.
unchanged and still trending toward a hit, with nothing qualifying from any of the four in the fortnight to 20 August. This issue's sweep was global rather than corridor-focused, so the negative finding here is weaker than in a house-lens issue and we say so rather than presenting it as a thorough check. Forty days remain and the entry scores on 30 September whatever the state of the evidence then.
Accenture's new bookings decline year on year again, confirming the June repricing as structural rather than sentiment.
the resolution date is corrected. Issue 021 carried it as around 24 September on a third-party aggregator's estimate; Accenture's own investor-relations calendar puts Q4 FY2026 results on 1 October 2026, and the entry now resolves on the company's date rather than on an estimate of it. Nothing else changes and no early scoring is attempted.
At least three of India's top four IT firms disclose a named AI-revenue metric, in whatever form each chooses, in their Q2 FY27 results.
India's top four IT services firms, TCS, Infosys, Wipro and HCLTech, in aggregate add net headcount over FY27's first half, April to September 2026, while each scales AI-attributed revenue, confirming the reroute: the work returns offshore even as the Western rhetoric softens.
unchanged from Issue 021. TCS alone added a net 9,279 in the June quarter on analyst arithmetic with annualised AI revenue of $2.6bn, which is one firm and one quarter of a two-quarter window; the entry requires the aggregate and the September-quarter results are the deciding input.
The first senior role explicitly titled for AI sovereignty or model continuity, distinct from CISO or Chief AI Officer, is publicly posted by a GCC entity or Gulf sovereign-linked employer.
nothing qualifying through 20 August, and the criteria risk stated in Issue 021 stands unchanged. Every relevant appointment we can find sits in the Chief AI Officer family, so if the sovereignty mandate is being absorbed into CAIO roles rather than generating a distinct title, this resolves as a definitional miss rather than a real-world one. The negative finding remains weak by construction because Arabic-language decrees are under-indexed in the sources we can reach.
At least one multinational publicly names the Philippines, Romania or Poland, India's closest challengers on this index, as the lead location for a new AI-delivery or engineering hub, chosen over India, in a 2026 announcement.
On the next annual refresh of this index, India retains first place on the outsourcing-led composite while staying outside the top three on the capability-weighted view, confirming that its lead rests on delivery scale rather than AI preparedness.
MoHRE publicly adjusts, delays or waives an element of Emiratisation enforcement, citing market conditions, before 31 December 2026.
unchanged and still heading for a miss. Nothing from MoHRE in the fortnight to 20 August. The most recent substantive posture remains July's statement that 95% of mandated companies met their first-half targets, which is compliance-positive and cuts against the forecast, and the Dh10,000 monthly per-role fines have been live since 1 July. Four months remain, so it is not scored, but we continue to expect a miss.
At least one company that attributed 2026 layoffs to AI is publicly reported to have rebuilt the same function in India, the Gulf or Africa, directly or through a capability centre or outsourcing partner, within twelve months of the cut.
A Gulf sovereign-linked or government entity publicly launches an initiative to position the UAE or Saudi Arabia as an AI-work delivery hub, not only a buyer or funder of AI, consistent with the capability-strong, labour-light profile the index assigns the Gulf.
The Karnataka Platform Based Gig Workers Act survives its constitutional challenge, meaning validity upheld, or the petitions dismissed or withdrawn, by 31 March 2027.
a carried discrepancy closes and a new gap opens. The discrepancy first: Issue 021 recorded that Justice Suraj Govindaraj recused from the IAMAI batch on 1 July yet heard Uber's petition on 28 July, and printed it as unexplained. It is now explained. LawBeat reported on 1 July that the recusal cited a conflict of interest arising from IndusLaw, the firm representing the IAMAI petitioners, with the judge stating that it cannot be before us and directing the matter to another roster bench even though counsel indicated no party objected. A conflict grounded in petitioners' counsel does not travel to a differently represented petitioner, so there was no contradiction and we should have established that before printing one. The new gap: LiveLaw's Karnataka High Court weekly round-up for 10 to 16 August, read in full, contains no gig-worker, Uber, IAMAI or platform-aggregator matter of any kind, so the outcome of the 14 August listing has now gone unreported for a week in the publication that covers this court weekly. We print the gap and assert no outcome. The Uber matter remains listed on or after 24 August 2026.
At least two further UAE government entities, emirate-level or federal and excluding Ajman, complete and publicly announce a fully autonomous end-to-end government transaction by 31 March 2027.
Basis: Ajman's live precedent, its 100-initiative three-year executive phase with coordinators now appointed across entities, and the federal directive to convert 50% of federal operations, procedures and services to agentic AI within two years with 80,000 employees in training. Against it: Ajman's own flow retains a customer approval step, fully autonomous is a description governments apply generously, and a first-of-its-kind claim is easier to make once than to repeat with the same language.
no second entity has announced. The nearest Gulf activity this fortnight was Qatar's Civil Service Bureau workshopping AI job classification with Google and Dubai Chambers signing Nasscom, neither of which is the completed autonomous transaction this entry tests.
A further institutional tally of AI-related hiring against AI-related job losses in India, from Nomura or any other bank, consultancy, industry body or official source, published by 31 March 2027, again reports hires exceeding losses.
Basis: the flow that produced the first result is still running, with TCS adding a net 9,279 in the June quarter, Cognizant's first Frontier cohort due by the fourth quarter, and 64% of new global capability centre roles created in 2026 requiring AI, data or automation skills, while the elimination side is concentrated in support functions already well through their automation. Against it, and this is a criteria risk we would rather state now than at resolution: the original is anecdote-count methodology and highly sensitive to which episodes a compiler happens to collect, one large Indian IT redundancy round would swing it, and no institution has committed to repeating the exercise at all. If no qualifying tally is published by the date, we score this a miss and say plainly that it failed for want of a publication rather than for want of the phenomenon.
Kenya enacts its AI policy, or an AI Bill, with the data-worker pay provision substantively intact, meaning pay for annotation, moderation or evaluation work calibrated against international rates for equivalent work, by 30 June 2027.
still nothing, now more than two weeks past the 4 August consultation close. No ministry statement, submission count, revised draft or industry response could be located to 20 August. Kenya's visible activity this fortnight was again on the growth side rather than the protection side, with the 4 August cooperation agreement to expand global business services re-confirmed live this week and nothing at all published on the policy. The ministry's own PDF remains unpulled, so the policy's provisions continue to be described only as reported by named outlets.
TechCabal Insights' full-year 2026 tracker records African tech layoffs above the half-year record of 2,574, while still naming AI as a direct cause in under 10% of tracked events: the cuts scale and the attribution does not.
Basis: the H1 record was driven by restructuring and banking consolidation that has not concluded, and the offshore losses that are genuinely AI-driven are decided by foreign clients who file nothing locally, so they cannot enter the tracker at all. Against it: a single large agent-deployment redundancy at a named African employer, of the Zap Africa kind but larger, would move the attribution share quickly off a small base.
None of India's top four IT services firms, TCS, Infosys, Wipro or HCLTech, publishes a redeployment rate by 30 June 2027: that is, any disclosed metric giving, for a defined period, the share of employees whose roles were automated or eliminated in favour of AI who remain employed by the firm, and in what function.
Basis: this is the metric that would settle whether a positive net headcount represents a transition or a replacement, and no firm in any market in this issue discloses it. The four publish quarterly headcount, attrition, AI revenue run rates and training and certification counts, none of which distinguish a redeployed worker from a new hire, and no regulator anywhere requires the figure. Nomura's own finding, that displaced workers rarely transition to AI engineering roles, is precisely what makes the disclosure unattractive. Against it: human-capital reporting in Indian IT is genuinely competitive, the figure would cost little to compute for a firm whose number is good, and a single firm choosing to differentiate on it would resolve this entry immediately.
the Korean levy bill is the first instrument we have seen anywhere that would compel the underlying attribution, since a levy triggered by AI-caused reductions cannot be administered without one. It would not by itself produce a published redeployment rate and it is not Indian, so it does not bear on this entry's resolution; it is noted because it is the first external pressure toward the disclosure this entry bets against.
Challenger's AI-attributed US job-cut count for the second half of 2026 exceeds the first half's 101,743, despite the softened executive rhetoric. The narrative and the number diverge further, not less.
the next data point is now confirmed rather than estimated. Challenger's August report publishes on 3 September 2026 on the company's own publication calendar. July ran 10,970 AI-cited against 112,713 year to date, about 24% of all cuts, with AI leading all reasons for a fifth straight month, so one month of H2 at that rate still leaves the second half short and the entry needs an acceleration it has not yet shown. A new risk to the entry itself belongs on the record: Andy Challenger warned publicly this month that as regulation takes shape companies will stop saying AI in their announcements, which would make tracking the impact of AI on jobs more opaque. If that happens inside our window, this entry could resolve as a miss because the instrument degraded rather than because the phenomenon did, and we would have to say so.
Data & AI remains a top-two growth sector in every remaining 2026 quarterly Cooper Fitch Gulf Employment Index, even if total GCC hiring stays flat or negative.
A monthly board-ready brief tying AI to the workforce and location decisions in front of you. Built by reAImagine.work and InGovern Research Services. Bengaluru and Dubai.
What you receive
A board-ready brief tied to your sector and the workforce decisions in front of you this quarter, this fiscal year, and twelve-plus months out.
Who it's built with
Editorial by reAImagine.work, founded by Debu Mishra. Board governance practice from InGovern Research Services, founded by Shriram Subramanian. Bengaluru and Dubai.
How it lands
Approve. Appoint. Commission. Separate. Deploy. Decide. Establish. Stakes in INR or USD ranges. No theatre.