Why AJ asks instead of answers

This is the last piece in a series about AI and how we think. I’ve argued that AI quietly does our thinking for us, that it’s making everyone think the same, and that it should be a mech suit, not a replacement. Fine words. Here’s what they actually look like in a product, because it’s easy to say and surprisingly hard to build.

The tempting version

When we started building AJ, the AI inside Clearahead, the obvious product was right there. Plug into every tool the company uses. Pull the metrics. Write the weekly update for everyone. Score each goal red, amber or green. Tell the leadership team what to focus on.

People would have loved the demo. I genuinely think it would have sold. And I think it would have made every team that used it slowly worse at running their company, for all the reasons in the research. Nobody thinks about an update they didn’t write. Nobody questions a score that sounds confident. Nobody owns a priority a machine picked.

So we did the opposite, and it’s been the single most important design decision we’ve made.

Context, not metrics

Most teams aren’t short of numbers. They’ve got dashboards coming out of their ears. What they’re short of is the story behind the numbers: why it moved, whether it matters and what anyone’s doing about it.

So Clearahead doesn’t build charts or pull your metrics in automatically. That surprises people. Instead, when something moves, the person closest to it says what changed in a sentence, in their own words. “Churn ticked up because two customers lost their champion.” “Pipeline’s down but the deals are bigger.” AJ saves that as evidence, with who said it and when, against the constraint or move it belongs to. I’ve set out the full reasoning on What Clearahead doesn’t do.

That one sentence is where the thinking happens. Writing it forces the person to work out what’s actually going on. Reading it tells everyone else something no dashboard ever will. And because it’s in their words, they remember it and stand behind it.

AJ asks before it tells

When a team sets a goal with AJ, AJ doesn’t hand them a plan. It asks what’s holding the goal back, and then it keeps asking until they’re past the symptoms and down to the handful of constraints that really matter. That conversation is uncomfortable sometimes. It’s meant to be. Working out what’s really in the way is the most valuable thinking a leadership team does, and it’s the bit I’d never let a machine do on its own.

AJ will suggest moves to clear each constraint, and the team keeps the ones that fit and bins the rest. Some of AJ’s ideas will be good. Some will be the sensible, average answers any model would give. That’s fine, because they’re suggestions, and the team’s own moves are usually better.

Guide, don’t block

The other half of this is that AJ never takes the wheel. It’ll tell you a move looks done, a constraint looks cleared or the goal looks achieved, and offer the one-click way to close it. But a person presses the button. If the team wants to do something AJ wouldn’t, it happens. AJ says once what it would keep an eye on, and gets out of the way.

That matters more than it sounds. The moment an AI becomes a gate, people either work around it or start deferring to it, and deferring is exactly the automation bias the research warns about. I’d rather AJ was occasionally ignored than routinely obeyed.

The admin is where AI earns its keep

None of this means AJ is shy. It does a huge amount of work, it’s just the right work. It chases the updates nobody wants to chase. It keeps the picture current from dozens of small updates, from each move all the way up to the goal. Every Monday it writes a short brief for everyone: where you stand, what moved, what you learned and up to three things that need a decision. It remembers what worked last time so nobody has to.

That’s the coordination admin that eats a founder’s week, and handing it to AI saves hours. The point is what you do with those hours. With an autopilot, you spend them not thinking. With AJ, you spend them on the decisions only your team can make.

Why I’d keep it this way

AI is going to keep getting better, and I know there’ll be pressure to let AJ do a bit more of the thinking each year. Customers will ask for it. Competitors will ship it. I think it would be a mistake. Not because AJ won’t be clever enough, but because a company is a group of people who’ve chosen to work towards something together. The thinking and the choosing are what make it theirs.

Use AI to save hours. Don’t use it to skip thinking. If you want to see how that feels in practice, Meet AJ walks through a real conversation.

Questions people ask

What is AJ?

AJ is the AI assistant inside Clearahead. It helps a team set a goal, find what is holding it back and plan the moves to clear it, then keeps the picture current from people’s updates and writes a short weekly brief. AJ suggests and people decide.

Does AJ write status updates for my team?

No, on purpose. People say what changed in a sentence or two, in their own words, and AJ turns that into evidence and keeps the picture current. The thinking behind the update stays with the person who did the work, and AJ does the collating and the reporting.

Can I connect my dashboards and metrics to Clearahead?

Clearahead doesn’t build charts or pull metrics automatically. Teams report a number with the context behind it: what it is, what changed and what they think it means. AJ saves that as evidence against the constraint or move it supports.