Don't force AI into everything
Embedding AI into every product feature is usually a mistake. Using AI as a productivity tool inside your org is not. Knowing the difference is the job.
There's a lot of pressure right now to put AI into everything. Boards ask for it, competitors announce it, and somewhere a roadmap gets a new column titled "AI." Most of it is a mistake.
Not all of it. The distinction is simple, and most organisations get it backwards.
AI in features: usually the wrong instinct
Most products do not need AI embedded in them. A button that does one thing reliably is better than a chat box that does ten things unpredictably. When you bolt AI onto a feature that was already working, you trade certainty for novelty — and users notice.
The honest test is this: would you ship the feature if you couldn't call it "AI"? If the answer is no, the AI was the point, not the user. That's a marketing decision wearing a product costume.
We've seen plenty of teams add a generative feature, watch engagement spike for a week, then watch it settle back to zero while support tickets climb. The feature didn't solve a problem. It announced that the company was "doing AI."
Good AI features are quieter than that. They remove a step the user already hated — search that actually understands intent, a draft that saves ten minutes, a classification that used to need a human. The user doesn't care that it's AI. They care that the work got easier.
AI as a productivity tool: almost always right
Here's the part organisations underweight. Using AI internally — to write code faster, summarise research, draft first versions, triage support, analyse data — is one of the highest-leverage shifts available right now.
This is where the real gains are, and it requires no product changes at all. Your engineers ship faster. Your analysts get to insight sooner. Your team spends less time on the first 80% and more on the 20% that actually needs judgment.
The teams winning with AI aren't the ones with the most AI features. They're the ones who quietly became faster at everything they already did.
The actual job
So the rule we work by is boring on purpose:
- Push AI into your workflows aggressively. This is where the leverage is, and the downside is small.
- Put AI into features reluctantly. Only when it removes a real problem the user feels, and only when you'd ship it even without the label.
Forcing AI everywhere isn't strategy. It's anxiety with a roadmap. The discipline is knowing where it earns its place — and being willing to leave it out of everywhere else.