Last month, three different people sent me the same LinkedIn post claiming that 5 million developer jobs would disappear by 2027 due to AI. The month before, someone else sent me the same format of post but about marketing jobs. The month before that, designers.
Here's what I've noticed: the frequency of "AI will destroy [category of work]" content has dramatically outpaced the actual evidence that it's happening at the scale claimed. That's not the same as saying nothing is changing — a lot is changing. It's saying the narrative is running significantly ahead of the reality, in ways that are actually harmful for the people trying to make career decisions based on it.
Let me tell you what the numbers I've been tracking actually say.
What's actually happened to developer employment since 2023
Large tech company developer headcount has contracted — significantly at some firms. The 2022–2023 wave of tech layoffs happened primarily because of the hiring bubble of 2020–2021, where companies massively over-hired assuming pandemic-era growth would continue indefinitely. That correction would have happened regardless of AI.
Post-2024, a different pattern has emerged: large companies are not rebuilding developer headcount at previous rates even as revenues recover. The explanation offered by most is specifically AI tooling — if a team of 15 engineers with AI assistance can produce what 25 engineers produced before, you don't rehire the 10. This part is real.
But the "developer employment is collapsing" narrative misses several countervailing forces. Software development itself is still expanding. Every business in every industry is now a software business to some degree. The number of digital products being built globally in 2026 is higher than in 2023 — significantly so. The "fewer developers per product" efficiency gain is being partially offset by "more products being built." Net employment outcomes look more like "stable to slight growth overall" rather than "mass displacement."
What has concentrated is the pain. Junior developer hiring is the entry category hit hardest. Big tech is not significantly expanding junior developer headcount because AI tools reduce the senior-to-junior headcount ratio required. This makes the beginning of a developer career harder in 2026 than it was in 2019. That's a real problem for a specific cohort — but it's not the same as "developers have no future."
The India-specific picture
India's IT services sector is the most consequential domestic labour story in this shift. Firms like Infosys, TCS, Wipro, and HCL built their business model on large teams of developers delivering defined scope — volume-based, margin on scale. AI tools that reduce the required developer headcount for equivalent output threaten that model structurally, not just cyclically.
I've spoken to engineering managers at two major Indian IT firms off the record. Both said versions of the same thing: their clients — US and EU enterprises — are starting to question whether they need to pay for the same team size when the team has access to AI coding tools that multiply individual output. The conversation is happening at renewal time. Rate pressure is increasing. The model is under stress.
But. The Indian IT sector still employs 5.4 million people directly and the projection for 2026–2028 is not mass displacement — it's upgrade pressure. The engineers who can work effectively with AI tools, understand AI systems, and build AI-adjacent products are in shortage, not surplus. The engineers doing commodity implementation work are under pressure. These are not the same population, and conflating them produces a false picture of the total sector.
The roles where AI is creating demand
This part gets less coverage because it's less emotionally engaging than layoff stories.
AI agent development — building, deploying, and maintaining autonomous AI agents for business use — is the most in-demand technical skill category of 2025–2026. Not "prompt engineering" in the superficial sense — actual agent architecture, MCP server development, tool use integration, multi-agent coordination. The demand for this skill significantly exceeds the supply.
AI integration engineering — taking existing business systems and wiring them to AI capabilities (embedding models, generation APIs, retrieval-augmented generation) — is a growing specialty. Almost every mid-to-large company has AI integration on its roadmap and isn't sure how to execute it. People who can do it reliably are valuable.
AI-assisted product management and design — PMs and designers who understand AI capabilities, can spec AI-native features, and can test and evaluate AI outputs as part of their workflow. Not a new job title, but a significant skill expansion of existing roles.
Indian founders building AI-native products — the startup formation rate for AI-native applications in India is accelerating. People who understand both the technology and Indian market distribution are building companies, not just finding jobs. This is a directional opportunity, not just a career one.
The question worth asking instead of "will AI take my job?"
Better question: what specifically do I do that AI can't replicate in the next 3 years?
Be honest answering it. If the answer is "write boilerplate code to spec" — that's under pressure. If the answer is "understand what non-technical clients actually need, translate it into a technical approach, and hold the ambiguity long enough to get the right solution" — that's much safer.
The useful career response to the current moment isn't panic and it isn't denial. It's an audit: of which parts of your current work have a high substitution risk from AI tools, and a deliberate move toward building irreplaceable judgment in the parts that don't. That's not a comfortable or easy exercise. It's the accurate one.
For developers who are mid-career in India: the knowledge of how systems actually work in production — reliability, observability, incident response, the political and organisational context of software decisions — is not going away. What's changing is which developers have it and which substituted expertise for speed. The ones who built real things for years know things the AI doesn't. That knowledge is your competitive moat. Extend it deliberately.
The career audit worth doing today: list the last 10 things you shipped or solved at work. Categorise each one: was this primarily an AI-substitutable task (writing code to spec, implementing a defined design) or a judgment task (deciding what to build, how to structure it, what to deprecate, how to communicate the trade-offs)? If more than 7 of 10 were implementation tasks, the repositioning work needs to start now, not after the market has already repriced your role.
Also see: Backend developers: your future is building AI agents, not just APIs.