Everyone is debating whether AI will take jobs.
Wrong question. The right one: where is it showing up first, and how do you tell signal from noise?
Read three pieces this week. The Economist on labour history. Bloomberg on Korea's AI dividend debate. Reuters on India IT.
One signal jumps out.
The Economist's case: no technology in history has produced sustained mass unemployment. Diffusion is slow. Aggregate OECD employment is at record highs. Industrial Revolution wages stagnated because of food prices and the Corn Laws, not because machines stole labour's share.
Solid history. Wrong altitude.
OECD aggregate employment is a lagging indicator. By the time that chart breaks, the political response is already drafted.
The action is one layer down. Sector. Firm. Function. That is where the signal lives now.
TCS FY26 dollar revenue: down 0.5% YoY, down 2.4% in constant currency.
First full-year decline since listing.
This happened while TCS won $12 bn in Q4 deals alone and AI services revenue crossed $1.3 bn annualised.
Big deals, smaller revenue. That gap is the story.
The gap is called AI deflation. Clients pay less per outcome because AI does more of the work.
Productivity gain is real. It is just not flowing to the services exporter. It is flowing to the client.
Same pattern at Wipro: third straight year of revenue decline despite TCV up 14% to $16.4 bn.
The Engels question, restated for 2026: who captures the productivity gain.
In Britain 1790-1840, capital captured it via cost-of-living inflation. Workers stagnated. Today, the early evidence in Indian IT is that clients capture it via lower prices. Neither the exporter nor labour wins.
What about jobs?
Nasscom: IT-BPM employee base is 5.80 million as of FY25, up just 2.2% YoY. FY25 net hiring was 126,000.
Compare to FY21 net hiring of 138,000 on a smaller base. The absolute number is similar. The base has grown 30%. The hiring rate has roughly halved.
But aggregate hiring is the wrong place to look. Watch attrition.
Indian IT attrition runs 15-20% in normal years. A firm with 15% attrition can shrink headcount by a third in two years without firing anyone.
No layoff press release. No backlash trigger. No Economist chart.
Just quiet shrinkage.
This is why I keep saying: aggregate data will mislead for years.
The adjustment in services-led economies runs through three channels in this order:
1. Hiring freezes
2. Attrition-driven shrinkage
3. Contract repricing
Only the fourth channel, mass layoffs, is visible. By then it is late.
For the investor: this changes how to read Indian IT multiples.
TCS at 26x trailing is not a growth stock being temporarily disrupted. It may be a structural transition story. The business model that built it, thousands of engineers, time-and-materials contracts, USD pricing power, is under pressure.
What to actually watch, in order of leading-ness:
Revenue per employee at top-5 Indian IT
Operating margin vs wage bill growth, same firms
GCC headcount additions in Nasscom data
Campus placement salaries at IITs and NITs
If three of these four turn together, you have your signal. Two to three years before any Economist chart shows it.
One blind spot in my own thinking.
India is not patient zero for AI displacement generally. It is patient zero for displacement of outsourced services work. Different thing.
US white-collar work bundled with client relationships, regulation, or physical presence will not follow the same curve. Don't flatten it.
Watch the facts, not the statements.
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.