Short answer
In the age of generative AI (GenAI), a developer’s value lies less in writing code than in making sound engineering decisions, taking responsibility for their quality and delivering business results.
Generating functional code does not, by itself, guarantee the scalability, reliability or maintainability of a software product.
Developers therefore need not only to implement technical solutions but also to explain to clients why those solutions are worth adopting, their potential consequences, limitations and risks, while remaining responsible for the quality of the final result.
The test of a developer’s value
Almost everyone in software development will have to take an “exam” we never signed up for. The exam is about proving to clients why they need our skills.
At first glance, this seems like a strange task. But you can understand the clients, too: they look at the end result, the surface, because that is the only way they can assess it.
Most clients never needed to understand what happens under the hood. But now that a result can be produced with just a few prompts, explaining the value of what remains invisible is becoming much harder.
Why a prompt alone is not enough
Some of us will have an easier time. It may be enough to say, for example: “You see, it can generate two normalization functions for the same data type in different places, or put four forms into a single component.”
Or: “A prompt alone is not enough. Solving the problem needs to start with a conversation in the chat. During planning, we have to read through five or more pages of LLM responses, agree with some points, say ‘no’ to others, and add our own comments and clarifications before approving code generation and integration. After that, we still have to check the result.”
Or: “GenAI can generate code, but it does not take responsibility for how that code will work a year from now or how the system will scale.”
Or: “Today, you can generate something that looks like a finished product very quickly. But there is an enormous gap between ‘it works’ and ‘you can build a business on this for years,’ and that gap is not always obvious at first glance.”
Each of us will have our own arguments for why GenAI is a tool rather than a magic button for every kind of task. I have only given a few examples that are easy to spot.
Technical arguments may not be enough
But we need to remember that there are two sides to this process: you and the client. And, unfortunately, technical arguments may not be enough.
Clients decide for themselves what they are building: a predictable system with long-term support, or simply an attractive facade.
From implementation to business value
Perhaps this exam will be useful for us as well. Unfortunately, “I know how it works under the hood” will no longer always be enough. We will have to explain exactly what value our knowledge creates for the business, emphasize our responsibility, and get even more deeply involved in the business itself.
Perhaps the main change is that we used to focus on explaining how we would implement a solution. Now, increasingly, we will have to explain why that particular solution is worth implementing and what consequences that choice will have.
Where engineering responsibility begins
I hope we all find the right arguments. Not to prove that GenAI is bad and we are irreplaceable, but to explain where the tool’s capabilities end and the engineer’s responsibility begins. Because when both sides take a sensible approach, everyone wins.
Updated: