The way I work has fundamentally changed over the past week and a bit. With Cowork — fully integrated Work IQ paired with Anthropic's models — I've seen a step-change in what I can get done in a day. But the more I use it, the more convinced I am that the conversation around AI productivity is missing the point.

What Has Actually Changed (And What Didn't)

What hasn't changed is the hard part: I still sit with the problem definition. I still own the challenge, the nuance, and the understanding of where my clients and I are heading. That work — thinking, reading, listening, anticipating — is mine. No model is going to do it for me, and frankly, I don't want it to.

What has changed is the execution. Once the initial understanding is built, the distance from idea to artifact has collapsed. Models, analyses, drafts, and structured outputs — things that used to take half a day now take twenty minutes. That's not a small productivity gain. That's a different kind of working day.

"Shit In, Shit Out" Is the Whole Story

There's a lot of skepticism right now that AI will make people lazy, that we'll stop thinking, that auto-generated Excel models and worksheets are a shortcut to mediocrity and that we will create a lot of mistakes and errors within our business. These are concerns and scepticism that I hear both internally and externally. I understand the concern, but I think it inverts the actual dynamic.

The people I see getting exponentially better outcomes are the ones who already had deep expertise. They know what good looks like, they recognise when the output is wrong, and they iterate fast because they've already done the thinking. The people struggling are those who hoped the tool would do the thinking for them and when the output disappoints, they conclude AI doesn't work and stop experimenting.

The tool amplifies what you bring to it. It does not substitute for what you don't.

The Comprehension Debt

A piece I read recently put words on something I've been wrestling with. AI-assisted software development has exploded — Claude Code, Codex, others — and it's undeniably useful. But it creates what the author called comprehension debt: code that exists without anyone fully understanding it.

The insight that stuck with me:

It isn't the code. It is the knowledge that went into writing the code. Most of what an experienced senior does isn't writing; it's thinking, reading, comprehending, planning, anticipating, and verifying their own understanding. Writing is just the mechanical operationalization of understanding.

This generalises far beyond software. The Excel model, the client deck, and the strategy memo — the artifact is not the value. The understanding behind it is. And human comprehension is still the bottleneck. It is not one you can skip.

It reminds me of Malcolm's warning in Jurassic Park:

"I'll tell you the problem with the scientific power that you're using here — it didn't require any discipline to attain it. You read what others had done and you took the next step. You didn't earn the knowledge for yourselves, so you don't take any responsibility for it."

That line is uncomfortably relevant right now.

The Split I'm Worried About

This brings me to the concern I keep hearing from clients — and feeling myself. The workforce is splitting. Those who are jumping on the train are pulling ahead at a pace that compounds daily. Those who are falling behind are falling further behind, faster than at any point I can remember in my career. The stretch within organizations and teams is widening exponentially.

We've been here before. The shift to digital. The arrival of smartphones. Each time, we eventually accepted a collective obligation: everyone, every age, every role, and every level needed to reach a baseline of understanding and capability. Not mastery. A floor.

We owe our teams and partners the same now. Not because everyone needs to become an AI power user, but because the cost of being on the wrong side of this gap is no longer abstract. It shows up in who can contribute, who feels confident, and who quietly disengages.

The Honest Takeaway

AI does not replace the work that matters. It replaces the work around the work that matters. If you bring expertise, judgement, and a clear problem definition, these tools will multiply you. If you bring vagueness and hope, they will multiply that too.

So this week, my Friday perspective is simple: keep earning the knowledge. Keep doing the thinking. And help the people around you reach the floor — because the gap is real, and it's growing while we debate whether it exists.