The future of software engineering

This article was originally published on LinkedIn.

Over the last three years of building AI products and doing AI-assisted software development, I have become increasingly convinced that the software engineering role is going to change significantly.

This is my current view, not a prediction presented as certainty. It is a forcing function for thinking about how I need to evolve over the next few years. If the argument is wrong, I would like to understand where. If it is partly right, perhaps it helps someone else think earlier.

The execution layer is getting commoditized

AI is already writing production code from instructions and context, refactoring modules, handling large migrations, finding and debugging production issues, performing detailed code reviews, and suggesting architectural improvements.

If model capabilities and agent harnesses continue to compound, implementation will become abundant. Once that happens, execution will no longer be a meaningful differentiator by itself.

Roles will converge around decisions

The convergence of engineering, product, design, and analytics is unlikely to happen because everyone performs the same execution work. It will happen because AI increasingly handles execution across those functions.

What remains scarce is judgment:

• choosing the right problem to solve
• defining success clearly
• setting constraints and making trade-offs
• owning outcomes end to end

The roles will not become identical. Each will still bring a different lens and optimize for different constraints. But the tools and the way of working will look more alike: think across boundaries, make decisions, and own outcomes rather than contributing only within a function.

The job moves upward, but the baseline rises too

Strong senior engineers have always been valued for judgment, ownership, and the ability to operate under ambiguity. What changes is that these stop being senior-level differentiators and become baseline expectations.

As implementation becomes more automated, the bottom of the engineering pyramid shrinks, the middle thins, and what remains looks more like a plateau: a smaller group operating at a higher level of abstraction.

The differentiator becomes the quality of decisions under uncertainty, not the ability to write code quickly.

What survives

The durable archetype is an AI-native product-systems owner: someone who understands distributed systems, users and incentives, constraints, cost, and trade-offs; uses AI to connect those pieces; reduces ambiguity; and owns measurable outcomes.

This person is not necessarily the fastest executor in the room. They are the person who consistently makes clearer decisions, especially when the trade-offs are uncomfortable.

The learning curve may break

Historically, engineers developed judgment through repetition: fixing bugs, shipping small features, and gradually taking on larger ownership. If AI absorbs much of that layer, how does someone develop taste and intuition?

The difficult questions are not easy to automate:

• When should you say no?
• When is something over-engineered?
• Which trade-off will hurt six months from now?

The bottleneck shifts from “can you code?” to “can you reason about systems?” That is a much harder filter.

Agency becomes more important

Agency is the ability to move things forward without waiting for instructions. It appears when you clarify ambiguity instead of escalating it, challenge flawed assumptions, reduce scope to reach the real objective, make a call with incomplete data, and own the outcome rather than just the task.

In an execution-heavy world, someone could survive with low agency if they were reliable. In an AI-heavy world, reliability alone will probably not be enough.

Why I am sharing this

I am not writing this to preach. I am writing it because if this scenario is even partially true, I need to change how I operate:

• move up the abstraction ladder
• strengthen judgment
• become more comfortable with ambiguity
• use AI as leverage rather than a crutch
• increase the surface area of ownership

I am not yet operating at the level this future may demand. This is me thinking in public and trying to shape my career intentionally rather than reactively.