Hi Reader,

A few weeks ago I gave a 90-second lightning talk at a conference in Miami. Experiment Live - a meetup for people who work in experimentation and product. Ten speakers, two minutes each, one slide.

Mine was titled "AI grows the tree. You prune it."

Here's what I meant.

The opportunity solution tree

Teresa Torres, who writes and teaches about continuous product discovery, has a framework called the opportunity solution tree. The short version: if you're trying to drive an outcome, you have opportunities. To develop those opportunities, you have hypotheses. To test those hypotheses, you run experiments.

It's a useful structure for keeping your work grounded in actual user problems instead of just building things because someone thought of them.

The issue with AI? It blows the bottom of the tree up.

The number of possible solutions to any given opportunity expands exponentially when you're using AI to generate them. A creative session that used to produce five decent ideas now produces fifty. Some of them are genuinely good. Most of them are plausible enough that you can't immediately dismiss them.

That sounds useful, and it is. Until you have 47 experiments you could run, no clear way to prioritize them, and the same amount of time and user attention you had before.

The bottleneck moved

For a long time, the constraint in product work (and in most knowledge work) was generating enough good options. You'd run brainstorms, bring in outside perspectives, whatever it took to get to more and better ideas.

AI removed that constraint almost entirely, which means the bottleneck moved.

The hard part is now convergence: the call about which ideas are worth your time, your users' attention, and your team's energy to test.

And convergence is the part you want to keep for yourself. This is the piece that runs on your context: what you've learned from talking to users, what hypotheses you understand well enough to stake a test on, what you've tried before and why it didn't work.

The tools that help you diverge won't do the deciding for you. And without a clear convergence practice, you end up with a beautiful, sprawling tree and nothing shipped.

What this looks like in practice

The most useful move I've found: treat divergence and convergence as separate modes, with a hard line between them. Diverging and evaluating at the same time kills both: you cut off ideas too early and you can't evaluate things you haven't fully generated yet.

When you're diverging: let AI do what it's good at. Push it for volume, variety, combinations you wouldn't have thought of. Don't evaluate yet.

When you're converging: put the AI output aside and bring in what it doesn't have - your knowledge of your users, your read on the market, the things that failed before and why. Use that to cut, not the AI.

I put together a free prompt pack for the divergence side: five techniques for getting more useful creative output from AI, including the ones I referenced in the talk. You can grab it at chillaborate.com/grow-the-tree.

A question for this week: Where in your current work is the bottleneck convergence - too many options, not enough decision-making - and you're still treating it like a divergence problem?

Work with me directly. If you want a thinking partner on your product strategy, AI rollout, or building your fractional practice - here's how I work with people: https://gamma.app/docs/Chill-Labs-Coaching-Information-ifjuyb0hjgw9xtb​

Come to the Water Cooler 🚰 I host a free monthly gathering for people building fractional and independent practices - honest conversations on positioning, pricing, and pipeline, plus recaps when you can't make it. Join here: https://chillaborate.com/fractional-water-cooler​

Let's Chillaborate,

Dina

Founder, Chill Labs

PS: New to the Automate Yourself podcast? There are plenty of episodes worth digging into - start anywhere: https://chillaborate.com/podcast 💙