Week 6, 2026—Issue #399

🌐 Automation Shifts Value from Execution to Oversight

File this under #customervalue

Photo by Shamin Haky on Unsplash

Over the past few weeks, I’ve been writing about how software engineering is changing. Not disappearing, but changing faster than most of us expected (see #397, #398). I find this interesting not only because I run a software company, but because it feels indicative of where knowledge work is heading more broadly.

What feels different this time is not just better tools, but a shift in how work itself is organised. Sure, AI code quality has improved markedly in the last few months, but the more consequential change is that agents can now retain context across days, sometimes weeks, under the right conditions. That alone changes what is possible. Work no longer has to be broken into tiny, tightly supervised steps. Agents can carry a thread, check their own progress, and keep moving.

Engineers at the bleeding edge are already adapting to this. Instead of writing most of the code themselves, they set multiple agents to work in parallel and outsource execution. Their own attention shifts upward, toward outcomes, context, constraints, and validation. The limiting factor becomes cognitive rather than mechanical: how much intent one person can hold, articulate, and oversee at once.

This is where things get interesting beyond engineering. This way of working is akin to what a proper OKR workflow should look like. We define what results we want to see, set boundaries, and deliberately postpone decisions about how to get there. Engineers just happen to be the first group for whom it has become possible to automate this at scale. In that sense, they are showing the way for what may soon be possible for the rest of us.

Which brings us to the real bottleneck. These workflows only work when intent is clear. And that has always been the hard part. In more than twenty years in professional services, I have yet to read an RFP that truly was well specified. Ambiguity, hidden assumptions, and shifting goals are the norm. Code was never the scarce resource. Clarity was.

Software engineers, then, are not going away. But they are changing. And the same will soon be true for many other forms of knowledge work. As execution becomes cheaper and more automated, value shifts upward: toward sense-making, goal-setting, constraint design, and responsibility for outcomes. Less time spent doing the work; more time spent deciding what work should be done.

That, I suspect, is the deeper pattern worth paying attention to.

/Andreas

Sidenotes

🚀 100× Contributor

Andrej Karpathy recently noted that coding with tools like Claude Code is shifting where leverage comes from. With agent orchestration, the old 10× engineer may become a 100× contributor — and leverage becomes even less evenly distributed.

🧶 Balls and Springs

A few weeks ago, I compared organizational change to phase transitions in physics (#379), drawing on ideas from my friend Tal Rotbart (CTO, me&u). He has since pushed the metaphor further; think balls, springs, and tension.

How can we build better organizations? That’s the question I’ve been trying to answer for the past 10 years. Each week, I share some of what I’ve learned in an email newsletter called WorkMatters. Back issues are marinated for three months before being published to Medium. This article was originally published on Feb 6, 2026. If you are reading this your are missing out. Subscribe to email updates at www.workmatters.studio.