AI took the easy half of software engineering. The hard half is bigger

Something real happened to software engineering, and it is not a vibe. The part of the job that was paid for throughput — writing the forms, the endpoints, the glue, the tests, the thousand small integrations that make a system usable — has become the cheapest thing in the building. An AI writes it faster than a team can review it, and the market has noticed.

That is uncomfortable for anyone whose career was built on being fast at exactly that. The old contract was simple: learn a stack, get paid for output. It held for twenty years. It is not holding now, and pretending otherwise does not help the people living through it.

Two markets, not one

There is a second, stranger market on top of the first. The people building the foundations — the models, the agent frameworks, the developer tools that automate other developers — are doing extremely well. That work is concentrated in a small number of companies, in a small number of places, and it pays accordingly. A narrow group is being paid a great deal to build tools that remove work from a much larger group.

That is a structural fact, not a moral one. The tool builders are not villains; they are solving the most valuable problem available. But it explains why the aggregate numbers look fine while individual engineers feel the floor move: the value has moved up the stack and into fewer hands.

What actually changed

The bottleneck moved. It used to be production: how many features can we ship this quarter. Now it is judgment: what should exist, what is correct, what is safe, what will still be maintainable in three years. AI is very good at producing plausible code and very bad at being accountable for it. Verification, security, architecture, data design, operations and taste are the scarce inputs now.

That is why so many engineers describe the job as converging on AI engineering. It is partly true. Building systems around models — retrieval, evaluation, guardrails, cost, latency — is the new core skill. But convergence is only half the story. The other half is divergence: software is becoming the instrument rather than the product, and the people who can point it at a hard domain are suddenly much more valuable than the people who can only point it at a CRUD app.

The bigger door

Here is the part that we think is underrated. The reason software engineering paid so well for so long is that software was the bottleneck for everything else. Biology, materials, energy, medicine, physics — every field was waiting on someone to build the tool, run the simulation, clean the data, wire the instrument.

That bottleneck just got much cheaper to clear. A small team can now attempt work that used to need a department. Drug discovery, protein design, new materials, climate models, instrumentation, analysis of data nobody could afford to process — these are no longer fantasies reserved for well-funded labs. They are engineering problems with a shorter path from idea to prototype.

This is why we do not accept the framing that AI is simply destroying opportunity. It is destroying a particular kind of opportunity — the one where being fast at boilerplate was enough — and opening a much larger one for people who can combine engineering with a domain. The catch is that the new work is not a job title yet. Nobody can hand you a description for "engineer who finds something worth building with a model and a domain expert". You have to make that role, and that is uncomfortable.

The honest problem

None of this makes job losses less real. A person whose work is automated still has rent to pay, and "learn AI" is not a plan for everyone. The question of how people get paid in the transition is genuinely open, and we should say so rather than wave at reskilling.

Our view is that the answer will not come from defending the old throughput work. It will come from doing new things that only became possible recently, and discovering, in the doing, what people will pay for. That is how previous general-purpose technologies played out: the jobs that replaced the lost ones were not visible from the day the old ones disappeared. They were found by people who started building before the market existed.

What we are doing about it

ABN Asia is not waiting for the job description to settle. We build with these tools every day, and we are deliberately pointing them at domains where software is the instrument: security, banking systems, data platforms, and increasingly the sciences. We are excited by the size of what is now attemptable.

We are also keeping the commitments that do not change with the technology: do the work properly, make it secure, hold the quality bar, and do it with care for the people on the other side. Commercial success is the reward for that, not the substitute for it.

The old software engineer is not coming back. The next one has more to build, in more fields, than any generation before. We intend to be among the people building it.

Commentary by ABN Asia. This is an editorial view, not a report of new findings.

Author

Ai-Banking-NetworkSecurity (ABN), ABN ASIA was founded by people with deep roots in academia, with work experience in the US, Holland, Hungary, Japan, South Korea, Singapore, and Vietnam. ABN Asia is where academia and technology meet opportunity. With our cutting-edge solutions and competent software development services, we're helping businesses level up and take on the global scene. Our commitment: Faster. Better. More reliable. In most cases: Cheaper as well.

Feel free to reach out to us whenever you require IT services, digital consulting, off-the-shelf software solutions, or if you'd like to send us requests for proposals (RFPs). You can contact us at [email protected]. We're ready to assist you with all your technology needs.