INDIA'S IT MACHINE IS BREAKING | How is the industry growing while the jobs are vanishing?
Executive Summary
A Think School case study on why India's IT services industry is growing without adding jobs. The presenter traces the industry from Y2K work to a pyramid of cheap, high-volume labour billed by the hour, argues large language models are strongest at exactly that repetitive base work, and explains the "AI deflation" problem for firms that bill for people instead of results. He ends with four steps for engineers. The caption text covers the whole 17 minute video. The figures (Infosys 8.2 percent AI revenue, TCS cutting 8,000 roles, fresher hiring falling from about 600,000 to 120,000, 2.3 million AI openings by 2027) are the presenter's claims and were not independently checked. This is career commentary, not financial, legal or career advice. Captions are automatic with errors. Topics are untimed; metadata follows YouTube. Source: https://youtubetotranscript.com/transcript?v=9k40TcOyG9U
Key Takeaways
Growing without jobs
The presenter says Infosys reports AI at 8.2 percent of revenue yet cut its yearly growth forecast to 1.5 to 3 percent, and that TCS, HCL and Tech Mahindra grew profit while headcount fell. His point is that the industry is growing without jobs.
How Y2K built the industry
He explains that the Y2K fix needed huge manual work, and Indian firms with English-speaking engineering graduates did it at 27 to 30 dollars an hour versus 60 to 70 for US contractors, which created the offshore model.
The pyramid and selling time
He says the model sold time at cheap rates in high volume: many freshers at the base, team leads in the middle, architects and client partners on top. Billing is engineers times hours, so the profit is the gap between billing and pay.
Why AI hits the base
He lists five bottom-layer tasks: testing, boilerplate, maintenance, translation between systems, and documentation. He calls them repetitive, patterned and low-ambiguity, which he says is where language models are strongest.
AI deflation
Because clients pay per person, he argues a 30 percent productivity gain lowers a services firm's billing by about as much, while a product company like Microsoft keeps the gain. He says the industry now calls this AI deflation.
What the numbers show
He cites fresher hiring down from about 600,000 to 120,000 a year, June entry-level openings down 44 percent and senior openings down 67 percent, and a 28-month low in active tech jobs. These are his cited figures, unverified here.
What engineers can do
He says AI cannot sit with a nervous executive, choose between architectures, or take responsibility at 3 AM. He gives four steps: build the judgment layer, learn real AI tools and not certificates, communicate with clients, and explain your work clearly.
The Conversation Breakdown
Growing without jobs
⏱ Untimed topicThe presenter says Infosys reports AI at 8.2 percent of revenue yet cut its yearly growth forecast to 1.5 to 3 percent, and that TCS, HCL and Tech Mahindra grew profit while headcount fell. His point is that the industry is growing without jobs.
How Y2K built the industry
⏱ Untimed topicHe explains that the Y2K fix needed huge manual work, and Indian firms with English-speaking engineering graduates did it at 27 to 30 dollars an hour versus 60 to 70 for US contractors, which created the offshore model.
The pyramid and selling time
⏱ Untimed topicHe says the model sold time at cheap rates in high volume: many freshers at the base, team leads in the middle, architects and client partners on top. Billing is engineers times hours, so the profit is the gap between billing and pay.
Why AI hits the base
⏱ Untimed topicHe lists five bottom-layer tasks: testing, boilerplate, maintenance, translation between systems, and documentation. He calls them repetitive, patterned and low-ambiguity, which he says is where language models are strongest.
AI deflation
⏱ Untimed topicBecause clients pay per person, he argues a 30 percent productivity gain lowers a services firm's billing by about as much, while a product company like Microsoft keeps the gain. He says the industry now calls this AI deflation.
What the numbers show
⏱ Untimed topicHe cites fresher hiring down from about 600,000 to 120,000 a year, June entry-level openings down 44 percent and senior openings down 67 percent, and a 28-month low in active tech jobs. These are his cited figures, unverified here.
What engineers can do
⏱ Untimed topicHe says AI cannot sit with a nervous executive, choose between architectures, or take responsibility at 3 AM. He gives four steps: build the judgment layer, learn real AI tools and not certificates, communicate with clients, and explain your work clearly.
Core Frameworks & Ideas
Selling time versus selling results - presenter's model
Firms that bill hours lose revenue when AI saves hours; product firms keep the gain.
The IT pyramid - presenter's model
Freshers at the base, supervisors in the middle, decision makers on top; AI erodes the base first.
The judgment layer - presenter's advice
Learn to solve problems without the tools so you can catch a wrong or lazy AI answer.
Actionable Next Steps
- ✓Pick one repetitive task in your work and test how far an AI tool gets it done.
- ✓Practise solving a problem by hand before using a tool.
- ✓Write a one-page explanation of a project as if for a client spending 100 crore.
- ✓Replace a certificate course with a small real project that uses an AI tool.
Notable Quotes
“It says that AI is extremely good at the bottom and extremely bad at the top”
“selling time at cheap rates in high volumes”
“Moral of the story is you don't need to be scared. You just need to spot the right opportunity to excel”
The One Thing to Remember
AI is extremely good at the bottom of the pyramid and extremely bad at the top, so build for the top.