The Receipts Arrive

The Sensing Report

A 96-hour window in late April flipped the application layer from agent-optional to agent-default, and within five trading days the layoff letters started catching up. The pretence that AI is "augmenting, not replacing" knowledge work did not survive this week's earnings calls.

If you spent the last eighteen months reading thought leadership that told you AI would create more jobs than it destroyed, this is the issue where the receipts arrive.

---

SIGNAL 1: Microsoft, Google, OpenAI and Adobe flipped agents to default in 96 hours

Signal Type: Strong
Strength Rating: 5 / 5
Timeline: Immediate

Between 22 and 24 April 2026, four of the largest enterprise software vendors moved AI agents from preview to default inside the tools knowledge workers already pay for. Microsoft made Copilot Agent Mode generally available in Word, Excel and PowerPoint, where the agent now drafts documents end-to-end, builds Excel analyses, and rebuilds decks against company templates without prompting at each step. Google used Cloud Next 2026 to launch Gemini Enterprise Agent Platform and Workspace Studio, a no-code agent builder that runs across Gmail, Docs, Sheets, Drive, Meet and Chat. OpenAI shipped GPT-5.5 with a one-million-token context window, launched Workspace Agents in ChatGPT for Business and Enterprise, and made agent mode a dropdown inside ChatGPT for Pro, Plus and Team users. Adobe rebranded Experience Cloud as CX Enterprise, anchored around persistent agents the company is openly calling "Coworkers."

This is not the announcement of a new product category. It is the application layer flipping from agent-optional to agent-default in the same week, inside tools that already sit on roughly half a billion knowledge worker desktops. Custom GPTs are being deprecated. Migrations carry hard deadlines, with OpenAI's billing transition arriving on 6 May. The vendor story has shifted from "AI assistant you can choose to use" to "AI agent that is already running unless you turn it off."

For organisations, the practical question is no longer whether to roll out AI tooling. It is whether your workforce policies, data governance, audit trails and licence allocations are written for a world where the default state of every Word document, every Slack channel, every Salesforce record is that an agent is already operating inside it.

Why this matters for your work: If your AI policy was written for a world where employees opted in, it is already obsolete. Audit your Microsoft 365 and Google Workspace agent permissions this week, before the agents do work no one approved.

---

SIGNAL 2: Cognizant put AI restructuring on the income statement

Signal Type: Strong
Strength Rating: 5 / 5
Timeline: Immediate

On 29 April 2026, alongside its Q1 results, Cognizant announced Project Leap, a restructuring programme with 200 to 320 million dollars in severance and personnel charges through December, expected to generate 200 to 300 million dollars of in-year savings. The CEO's framing was unusually direct for a firm of this scale: the programme exists "to transform the operating model by investing more in AI capabilities and offerings, upskilling the workforce, realigning it to market demands, and improving productivity." Q1 revenue rose 5.8 per cent to 5.4 billion dollars. Trailing-twelve-month bookings hit 29.6 billion, up 11 per cent. This is not a downturn cut. It is a deliberate AI-led restructuring booked alongside a record bookings quarter.

The precedent matters more than the numbers. Three years ago, Cognizant's previous restructuring affected roughly 3,500 employees at a comparable cost. Project Leap is the first time a firm in the WITCH tier (Wipro, Infosys, TCS, Cognizant, HCL) has formally booked AI-led workforce restructuring as a P&L line item with the rationale stated openly. TCS already shed 23,460 roles in FY26 (Q4 results, April 2026), but framed it as "limited deployment opportunities and skill-mismatch." Cognizant has dropped the euphemism. Every CFO at every Indian-heritage IT services firm now has the template, the precedent, and the analyst expectation that they will follow.

For Indian IT, this is the moment the productivity-vs-headcount conversation moved from quarterly earnings call hedging to formal restructuring with named savings targets. The 5.5 million people directly employed in India's IT services industry are about to discover what "operating-model transformation" means in practice.

Why this matters for your work: If you lead an India delivery centre, a captive operation, or any function that bills clients for human time, you have between now and Q3 to decide whether your operating model will be restructured deliberately or reactively. The next four WITCH quarterly results will tell you which way the sector is moving.

---

SIGNAL 3: Six US banks shed 15,000 jobs while booking 47 billion dollars in profit

Signal Type: Strong
Strength Rating: 4 / 5
Timeline: Immediate

JPMorgan, Citi, Bank of America, Goldman Sachs, Morgan Stanley and Wells Fargo collectively cut around 15,000 jobs in Q1 2026 while booking roughly 47 billion dollars (35 billion pounds) in profit, up 18 per cent year-on-year. Bank of America CEO Brian Moynihan credited AI directly for 1,000 attrition cuts, four months after telling staff that AI was "not a threat to their jobs." Citi has pledged a 20,000 headcount reduction. JPMorgan's Mary Callahan Erdoes described an AI-transformed controls review process that previously required 200 people. The bank's 2026 technology budget is 19.8 billion dollars, up 10 per cent.

The Q1 reduction equals roughly one per cent of the six banks' workforce. If the ratio holds for three more quarters, the sector sheds another 50,000 jobs by year-end without any single bank making a dramatic announcement. The structural effect is not at the front line. It is in the middle. The vice-president and director ranks that historically led to partner positions are being thinned at exactly the rung where pitchbook automation, controls review, proxy advisory and analyst-grade financial modelling now operate. JPMorgan Asset Management has already eliminated its proxy advisor function entirely, replaced by an internal AI tool. The career staircase that produced Wall Street's senior bench for forty years is missing several steps.

For finance, the signal is that the sector has stopped pretending. The same banks publishing record profits and historic technology budgets are admitting in the same earnings cycle that the headcount line is going down because of AI. The macro narrative that "banks always say layoffs and never do them" no longer fits the data.

Why this matters for your work: If you are a mid-career banker on the VP-to-MD track, the path you assumed existed has narrowed. If you lead talent at a financial institution, your succession pipeline planning needs a five-year rebuild starting now, not at the next strategic review.

---

The pattern across these three signals is not three separate stories. It is one story told from three angles. The model layer consolidated (Google's 40 billion dollar Anthropic commitment on 24 April). The application layer turned on by default. The workforce got the invoice in the same week. The rest of this issue tracks where that invoice landed by sector, by region, and by the people closest to it.