Every few months a headline announces that AI will replace programmers. Every few months another one says the fear is overblown. Both get clicks and neither helps you decide what to do with your own career.
The useful question is not "will AI take my job". It is: which parts of my job is AI already good at, and what does that leave me doing? That question has a real answer, and the answer is different depending on where you sit.
What AI is genuinely good at now
Being honest about this is the whole point. These are not hypothetical:
- Writing routine CRUD code — the forms, the endpoints, the database calls that look the same in every project.
- Converting between formats and languages.
- A first draft of unit tests.
- Explaining unfamiliar code, which used to cost a new joiner weeks.
- Documentation, which most teams never wrote anyway.
- Mechanical refactors across a lot of files.
- First-line support answers for questions that have been asked a thousand times.
Notice what those have in common: the task is well specified, small, and has been done before. That is the shape of work AI compresses.
What it is still bad at
- Deciding what to build. Turning a vague business complaint into a specification is most of the job, and it is the part nobody writes down.
- Debugging live production with partial information, pressure, and three systems interacting in a way no document describes.
- Your company's undocumented rules — the ones that exist only because of a decision someone made in 2019 and a customer who shouts.
- Accountability. When it goes wrong at 2am, a person has to own it. That is not a technical limitation and it will not be solved by a better model.
- Long-horizon architecture, where the cost of a bad choice shows up two years later.
So who is actually exposed?
The exposure is not even, and pretending it is helps nobody.
| Where you sit | Exposure | Why |
|---|---|---|
| Entry-level developer doing well-defined tickets | High | That work is small and specified — exactly AI's strength |
| Manual QA | High | Repetitive, scripted, describable |
| First-line support | High | Most questions have been answered before |
| Mid-level developer who only implements | High | The generic middle is where the 2026 cuts actually landed |
| Senior engineer who owns a system | Lower | Judgement, context and accountability |
| Engineer with deep domain knowledge | Lowest | The knowledge is not on the internet to be learned from |
There is something specific to say about Indian IT services, and it is not what the headlines suggest. That industry has historically billed by headcount for well-specified work delivered to someone else's specification — precisely the shape AI compresses hardest. But through 2026 the squeeze did not land where most people expected. The large firms went on hiring freshers in tens of thousands while cutting middle and senior grades. The pyramid is thinning in the middle, not at the base.
Two things follow from that, and both are more useful than the usual panic. Fresher hiring continues — but the filter has moved. Intake is increasingly screened on skills you can demonstrate rather than on a degree plus a training programme, and the skills being asked for are specific: AI, cloud, data pipelines, test automation. And the generic mid-level role — the one that takes a written specification and implements it — turns out to be the most exposed position in the industry, not the safest.
The uncomfortable part nobody says plainly
The tasks juniors traditionally learned on are the tasks AI now does. That is not an immediate problem for today's seniors. It is a serious problem for how the next generation of seniors gets made.
If you are early in your career, you can no longer rely on "do tickets for three years and absorb it by osmosis". The osmosis has been automated. You have to go and get the harder work on purpose — ask for the design discussion, volunteer for the production incident, take the piece nobody understands.
And notice how this fits with the hiring pattern above. Getting in is not the hard part; the industry is still taking freshers. The hard part is the rung after that, because the layer you were supposed to climb into is the layer being cut. Plan for the second job, not the first one.
What to actually do
- Use AI harder than the person next to you. In the near term the competition is not human versus AI. It is people who got twice as fast versus people who did not. Refusing the tools on principle is a luxury position.
- Move from "implements" to "owns". Owning an outcome — this feature works in production and I am responsible for it — is a different job from completing tickets, and it is a much harder one to automate.
- Go deep in one domain. Trading systems, payments, health records, logistics, lending. Generic coding is becoming a commodity. Generic coding plus ten years of knowing how NSE settlement actually works is not.
- Learn to operate, not just to write. Monitoring, deployment, incidents, performance. Systems fail in ways nobody specified, and that is where people earn their seniority.
- Make the work visible. Inside your company and outside it. The people who are fine in a hiring freeze are the ones whose work other people can already name.
- Build a financial buffer. This is not career advice, it is career insurance — and it does more for your negotiating position than any certificate.
What not to do
- Do not panic-switch to "AI engineer" with no foundation. The shortage is in people who can build real systems. A prompt course does not create one, and the market works that out fast.
- Do not assume your employer's stability is your stability. A company can be perfectly healthy and still stop hiring, or restructure the team you are in.
- Do not take on a large EMI on the assumption that your income only goes up. This is the single most common way a job scare turns into a genuine crisis. A fall in income is survivable. A fall in income with a fixed obligation against it is not.
The honest part
Nobody knows the ten-year picture. Anyone giving you a confident number about how many jobs go is selling something — a course, a newsletter, or an opinion they enjoy having.
So plan for resilience rather than for a prediction: skills that transfer to another employer, savings that buy you months instead of weeks, and a network that actually answers when you message it. Those three things pay off whether the pessimists or the optimists turn out to be right.
Educational only. This article explains how things work. It is not investment, tax, legal or career advice, and it is not a recommendation to buy or sell anything. What suits you depends on your own situation - check with a qualified adviser before acting on money decisions.