Role of a human

I was, and still am, very enthusiastic about using AI to improve our productivity. I always have been, as all engineers, with every automation we saw and embraced. And each of them required us to change our role and position in the larger world, career and industry. But with the advent of LLMs, the change came quicker and hit harder. I see that in my colleagues questioning their roles and capacities, in graduates struggling in the job market, and in friends left jobless after their organisations swapped them for some agentic workflows.

I don’t try to fool myself into thinking I don’t have a role in that turmoil and that I’m not on the chopping block ahead. Everyone is. Just as we had to adapt to changing programming languages and frameworks to keep up with the market, we need to adjust again. But again, quicker and harder.

Not long ago, my problem in team management was how to fit all the work into a finite team. Today, the problem is how to divide work so people don’t step on each other’s toes. At the same time, we don’t want to put anyone in a silo with a critical bus-factor. In that world, I cannot imagine large teams working on a single component, nor can I they can work on different aspects of it. It feels like we need to come back to old school pair programming - with the difference that engineers prepare the specification, not the code itself.

And each individual engineer must transform the perception of their own role. They aren’t artisans of code anymore. They aren’t the ones making changes. They can lead them. They can start work, establish patterns, standards, and guidelines. But at the end of the day, they become managers of their own (agentic) team. It’s an easier transition for someone who has had their own reports like me. It’s a larger mental transformation for someone who was the file-and-rank contributor.

I have absolutely no idea what space it leaves for graduates and other beginners - the next generation of engineers. I feel they still need to learn the ropes, but the market may not be interested in those skills. They will need to do it to establish their own experience and judgement. The same as no one (including a teacher) is hired just for knowledge of the maths curriculum, engineers won’t be hired for knowledge of languages and frameworks. But they will need to know that foundation to steer their digital delegates.

And there are so many aspects outside of coding. Just in creating my own projects and playing with this supercharged automation, I invested in self-owned cookie-less analytics, social media automation, market research, and blog writing (not this post), work and life planning, weight loss, nutrition, mentoring, and AuDHD support. For each use case, I needed my experience in systems thinking and at least a basic understanding of the topic. Only then could I pick an LLM, with a healthy amount of algorithmic guardrails, to get good-enough results. The more personal I wanted the outcome to be, the less automation I had to use. That is why I wrote this post myself.

That seems to be our place. We can write essays, blog posts, and books using AI. But they will be nothing other than just a blend of thoughts written before. Semi-random reconfiguration of words that were already spoken. I won’t improve my communication or take a fresh take if I just prompt it and get results. Same as a newbie won’t become an AI-enabled engineer, just by giving commands to Lovable to Claude Code and not trying to make their hands dirty with code.

Our role is to debate, argue, and challenge. We must make the first few steps to give the algorithm directions. Our role is still in reaching first frontiers, even if we send a probe deep into space - like Voyager - where we cannot reach ourselves.