The Sensing Report: The Confession

This week, the AI workforce story stopped being a forecast and became a confession. Gartner has surveyed 350 large enterprises and reported what every honest CHRO already knew: companies are cutting people in the name of AI, and the cuts are not generating returns. The pretext has caught up with the strategy. The interesting question is no longer whether AI will displace work. It is what boards do when they realise the displacement has not produced the value they bought it for.

---

SIGNAL 1: Gartner confirms 80 per cent of AI-deploying enterprises have cut staff. The cuts are not delivering returns.

SIGNAL TYPE: Strong STRENGTH RATING: 5 TIMELINE: Immediate

On 5 May, Gartner published the results of a survey of 350 global enterprises, each with revenues above one billion dollars and each piloting or deploying agentic AI, intelligent automation, or autonomous systems. Around 80 per cent had reduced their workforce as a result. The headline finding, delivered by lead analyst Helen Poitevin: workforce reduction rates were nearly identical between companies reporting strong AI returns and companies reporting modest gains or negative outcomes. Cutting people did not correlate with making money.

This is the first piece of large-sample, non-vendor evidence that the AI-as-cost-takeout playbook does not work on its own terms. It does not say AI is failing. It says the layoff-first strategy is failing. Poitevin's framing, that organisations improving ROI are not those eliminating people but those amplifying them, is a direct rebuke to a generation of CEO communications that linked AI deployment to headcount reduction as if the two were the same project. They were never the same project. The market was just too distracted to notice.

For boards in India, the GCC, and Africa, the finding removes the air cover that Western CEOs have been providing for aggressive cost-cutting. Until this week, "we are doing what Salesforce, Meta, and Microsoft are doing" was a defensible boardroom argument. After this week, it is an argument that fails the basic ROI test. The smart organisations in our regions will pause, reopen their AI business cases, and ask whether the headcount line was a forecast or a wish.

Why this matters for your work: If your AI investment thesis assumed people savings as the primary return, you now have third-party evidence that the assumption is wrong. Open the model. Replace the people line with capability amplification metrics. The board will eventually ask.

---

SIGNAL 2: Cognizant signals 12,000 to 15,000 layoffs under Project Leap, with the majority falling in India.

SIGNAL TYPE: Strong STRENGTH RATING: 5 TIMELINE: Immediate

Cognizant disclosed Project Leap on 29 April with a severance budget of 230 to 320 million dollars. Industry reporting through this week has converged on the implication: the budget covers between 12,000 and 15,000 redundancies globally, with the bulk falling in India where the company employs more than 250,000 of its 357,000 staff. CEO Ravi Kumar S has framed the restructuring as a move toward a "broader and shorter pyramid" combining digital labour and human labour. Translation: fewer freshers, fewer middle managers, more agentic systems doing what entry-level associates and manual QA testers used to do.

The Cognizant move is not isolated. It sits inside a pattern that includes Freshworks cutting 11 per cent, Atlassian cutting 10 per cent, Coinbase cutting 14 per cent, PayPal targeting 4,760 jobs across two to three years, and Meta beginning 8,000 cuts on 20 May. What is distinct about Cognizant is the geographic concentration. The Indian IT services pyramid, which has fed the country's middle class for two and a half decades, is being dismantled by its own architects. Clients are no longer willing to pay for large entry-level training cohorts. The pyramid that fed campus hiring at IITs, NITs, and tier-two engineering colleges is becoming a flat top with sharp edges.

For Indian organisations and Indian workers, this is the first time the signal cannot be interpreted as someone else's problem. Infosys, TCS, Wipro, Tech Mahindra, and HCL are watching Cognizant's framing and severance accounting with attention. None of them have announced numbers of this scale, but each of them has a similar pyramid problem. The shape of Indian IT services employment is going to change in the next eighteen months, and it will not change at the top of the pyramid. It will change at the bottom and the middle.

Why this matters for your work: If your organisation hires from Indian IT services as a vendor or as a talent source, the people you have been buying will not exist in the same form by 2027. Map your dependencies now and identify which roles you are sourcing through staff augmentation that you should be sourcing through capability transfer.

---

SIGNAL 3: UAE launches the world's first agentic AI work permit system, effective 1 May 2026.

SIGNAL TYPE: Emerging STRENGTH RATING: 4 TIMELINE: Immediate

The UAE has activated a new framework under which every work permit application submitted from 1 May 2026 onward is automatically evaluated by an autonomous AI system, governed by a national policy on agentic AI in government services issued by Sheikh Mohammed bin Rashid Al Maktoum. The framework treats AI agents not as tools assisting human decisions but as the primary decision-maker on a category of public administration that affects the working lives of millions of expatriates. Human review is the exception, not the default.

This is the first time a major economy has put autonomous AI in the critical path of labour market access at scale. The implications go beyond efficiency. The UAE has effectively decided that the gating function on who works in the country, and on what terms, is now an algorithmic function. That decision is consistent with the UAE's broader bet on becoming the global capital of agentic governance, but it also creates a new category of risk. Errors, biases, and edge cases in the work permit algorithm are no longer human errors with human accountability. They are system errors with diffuse accountability, and the population most affected has the least leverage to challenge them.

For HR leaders in the UAE, this changes the negotiation dynamics on hiring foreign talent. Permit timelines may compress dramatically, which is good. But the criteria for permit issuance now sit inside a model whose decision logic is not fully transparent, which is harder. Companies that hire heavily into UAE will need a new internal capability: understanding how the agentic system reads job descriptions, salary benchmarks, and qualifications, and tuning their applications accordingly. This is a new form of compliance, and the firms that learn it first will hire faster.

Why this matters for your work: If you hire into the UAE, audit your job description templates and offer letters in the next thirty days. The agentic system will rank some of your applications above others based on signals you did not know it was reading.

---

SIGNAL 4: Microsoft Work Trend Index identifies the "Transformation Paradox" as the binding constraint on AI value.

SIGNAL TYPE: Weak STRENGTH RATING: 3 TIMELINE: 6 months

Microsoft released its 2026 Work Trend Index on 5 May, surveying 20,000 AI users across global workplaces. The central finding sits well alongside the Gartner data and adds the missing piece. The barrier to AI value is not the technology and not the workforce. It is the organisation around them. Only 13 per cent of AI users say they are rewarded for experimenting with AI in their jobs. The employees are ready. The systems and incentives around them are not.

This finding matters because it identifies the layer of the problem that boards typically ignore. Companies are willing to spend on AI infrastructure (hundreds of billions across hyperscalers this year) and willing to fund AI training programmes (Microsoft alone has committed to skilling three million Saudis by 2030). What they have not done is redesign the performance management, rewards, and approval architectures that govern day-to-day work. So employees with new AI capabilities run into the same approval workflows, the same career ladders, and the same recognition systems built for a pre-AI organisation. The transformation paradox is the gap between individual readiness and institutional inertia.

For HR leaders, this is the most actionable finding of the week. The infrastructure problem is too big and too expensive to solve in a quarter. The skilling problem is funded and underway. The organisation design problem is the one that is both unaddressed and within the reach of the HR function. Boards that ask their CHROs "where is our AI-era performance management system" will be asking a question that almost no organisation can answer.

Why this matters for your work: Audit your performance management cycle for AI-era misalignment. If your top performers in 2026 still look like your top performers in 2022, your reward system is not capturing AI value. It is hiding it.

---

SIGNAL 5: ServiceNow and Accenture launch forward deployed engineering programme to push agentic AI from pilot to production.

SIGNAL TYPE: Emerging STRENGTH RATING: 3 TIMELINE: 1 to 2 years

At ServiceNow's Knowledge 2026 conference on 6 May, ServiceNow and Accenture announced a forward deployed engineering programme designed to take agentic AI from pilot status to enterprise-wide production. The framing matters because it is a tacit admission of a sector-wide failure. Multiple analyst sources, including Gartner and MIT, have reported that 86 to 89 per cent of agentic AI pilots have not reached production at scale. The pattern is consistent. Pilots impress. Production breaks. The gap is not technological but operational: integration with legacy systems, identity and permissions, audit trails, change management, and the parts of enterprise IT that nobody puts on stage at a vendor keynote.

The Accenture and ServiceNow programme essentially packages the operational scaffolding into a service. Build the agent natively inside the system where work happens, then scale through a single continuous motion. It is also a sign that the agentic AI market is moving from sale of capability to sale of deployment. The next twelve to eighteen months will reward vendors who can demonstrate not pilot velocity but production survival rate.

For organisations in our regions, the lesson is to be sceptical of any agentic AI presentation that does not include an integration architecture diagram. The board demos are easy. The handoff to operations is where most projects collapse. If your AI strategy does not have a named owner for the operational layer, you do not have a strategy. You have a shopping list.

Why this matters for your work: Before approving any agentic AI investment in the next two quarters, require the vendor to walk you through a customer in production for at least six months. If they cannot, the pilot is the product.

---

This week's signals fit together. Gartner says the layoff-first strategy does not deliver. Cognizant says it is happening anyway. The UAE says agentic systems are entering critical-path decisions on labour itself. Microsoft says the binding constraint is organisation design, not technology. ServiceNow and Accenture say the production gap is the new ROI problem. Read the rest of this issue with all five in mind. They are one story.