Put Authority Close to the Work

A company can get everything else right and still fail, because the decisions are being made by the wrong people, too far from the problem.

This is part three of a five-part series on how the companies seeing real returns on their AI investment are actually working.

We’re about half way through this series. So far we’ve talked why it’s important to start with, and get wedded to, the problem you’re solving We have explored three things the companies getting a return all do: Put the solutions they build where their people work, measure the cost of the problem before they start, and let the people using it make it better.

But I ended the last article teasing a fourth thingIt isn’t about AI at all. It’s about who’s allowed to decide.

Most big companies are set up something like this. The AI project goes to a project group. The people on it are borrowed part-time from their real jobs. The actual decisions sit with a committee that meets once a fortnight. The group recommends; the committee approves. So a suggestion waits two weeks for a yes. Then the next one waits another two. By the time anything’s agreed, the AI has moved on and the problem has shifted underneath it. Nobody on that group is slow or foolish. The structure is designed to be slow. They’re just stuck inside it.

And look at who’s deciding. The committee is senior people, far from the daily work. The person who actually feels the problem, the one drowning in it every day, isn’t in the room. They get asked for input, maybe, and then they wait to hear what was decided. The people who know the problem best have the least say in fixing it. That’s the part that’s broken.

The companies seeing positive returns on their AI investments have turned that round. They put a small group in charge, three or four people, and let them decide for themselves, within agreed limits, without going back up the chain each time. And crucially, the people in that group are the ones who live the problem. The person who runs the workflow. The one who handles the tricky compliance call. The one who deals with the customer when it goes wrong. They’re not asked for input afterwards. They’re in the room, making the call, because they’re the ones who understand what they’re deciding.

This is the harder shift, much harder than buying a tool, because it asks leaders to give something up. To let a few people decide without the committee’s blessing. To trust that the person closest to the work knows more about it than the org chart suggests. Most companies can’t quite bring themselves to do it. That’s exactly why so few are seeing a return.

But the ones that have done it are clear proof it works. Every, a company that is part publisher, builder of software and consultant, runs their business with teams of three or four who are able to turn their hands to most challenges. Block has stripped itself down to a handful of clear roles and now lets AI do the chasing-up and coordinating that used to keep a layer of managers busy. Shopify told its teams they have to show why AI can’t do a job before they’re allowed to hire a person to do it. Different companies, different trades, same instinct: keep the group small, keep the skills broad, and make the decisions close to the work instead of far above it.

What changes when you do this isn’t only speed, though things do speed up. What changes is that deciding and doing stop being two separate jobs done by two separate groups. In the committee way of working, one group works it out and another group approves it, on different days, and the wait between them is where the momentum drains away. When the people who know the problem can also decide, the gap closes. The understanding and the decision live in the same heads, in the same room, on the same day.

That’s a change in how a company behaves before it’s a change in what it builds. It asks senior people to hand real decisions to people below them. It asks the people who live the problem to step up and own it, not just describe it and wait. And it asks everyone to get comfortable aligning and making progress quickly, out in the open, instead of hiding behind a fortnightly sign-off. For a company used to careful approval at every step, that’s uncomfortable. It’s also where great work starts.

So you can get the three principles right, put the tool in the right place, measure your starting point, listen to the people using it, and still get nowhere if every decision has to climb the ladder and come back down. Get the deciding right and you’ve cleared the thing that stops most companies.

Which leaves the obvious question. If a small group close to the work is the answer, what does that group actually do? How do a few people build something real, and build it fast enough to matter? That’s the next piece.

AUTHOR FOOTER

David Beath runs BuildFirst, where he helps organisations in regulated industries build with AI without losing their footing. He writes The AI Cookbook, notes from two months further along the trail. After a quiet stretch, he is writing again. If you are seeing your own version of this pattern inside your organisation, he would like to hear about it.

Next: how a small group, given the room to decide, actually builds, and why a few people can now do in two days what a big project takes six months to even agree on.