Capture First, Automate Second: Friendly Coffee Talk About AI Adoption

By Liesa Witt · Published 2026-05-22 · Updated 2026-10-09

AI cannot automate what was never written down. Why capturing knowledge first beats automating second, and what companies lose when people leave.

Why companies are quietly re-hiring the people they replaced with AI, and what it tells us about the real bottleneck.

Had coffee with a friend last week who told me something I can't stop thinking about.

We were talking about the AI bubble. The fear of being replaced. And how a sudden shift in workforce can quietly backfire.

Some companies have already gone there, replacing a chunk of their team with AI tools earlier this year. The goal was saving costs and being more efficient. Honestly? I think most of us are aiming for some version of that.

But here's the thing: when you don't fully understand the work a human does, replacing them with AI will eventually backfire. Because you're missing the one question that actually matters:

How is the work done?

Fast forward six months, and the company in question is quietly hiring some of those same people back. Not all of them. But enough to repair what broke.

So instead of saving money, they're looking at lost productivity, lost money, and, in the worst case, lost clients. All for chasing the AI replacement gold rush.

It wasn't that the AI didn't work

Here's the part that stuck with me: the models are genuinely capable. The agents do what they say on the tin.

The problem was upstream.

AI can't automate what isn't written down.

And did you know that roughly 42% of the knowledge inside a company lives only in people's heads?

The exact way Sarah handles that one client who always emails on Fridays. The unspoken rule that finance reviews anything over €5K. The reason you stopped using that integration two years ago.

When those people walked out the door, the documentation they never wrote walked out with them. The AI was left trying to automate workflows that, technically, didn't exist anywhere a model could read them.

So the company did what companies do: they hired the knowledge back. Sometimes as contractors. Sometimes as advisors. Sometimes just quietly, with a different title. A few things I've been sitting with The teams getting real value from AI right now aren't the ones with the biggest budgets. They're the ones who already had a writing culture.

"Let's pilot an AI tool" often turns into "let's finally write down how we actually work." The pilot becomes the documentation project in disguise.

Replacing people before capturing what's in their heads isn't an efficiency play. It's an expensive way to discover how much of your business runs on tacit knowledge. Not an argument against AI I don't think this is an argument against AI. I think it's an argument for sequencing.

Capture first. Automate second.

Otherwise you're just paying twice, once to let the knowledge leave, and once to bring it back.

Curious if others are seeing the same pattern. Companies quietly re-hiring after AI rollouts that looked clean on paper?