Want people to use your AI features? These 5 teardowns show you how
Getting someone to actually try an AI feature is harder than building it. Usage is the new moat. These are the teardowns I keep sending to product teams about how to close that gap.
Everyone is shipping AI features right now, and very few of them are getting used.
So how do you get people to actually use the thing you built? In these popular teardowns, I break down what real products got right or wrong about getting users to try and stick with their AI features.
These teardowns are for you if: you’re planning to ship or relaunch an AI feature in the next six months, or you work in product and you’re focused on adoption and engagement.
Curious about working with Irrational Labs on AI adoption? Reach out: kristen@irrationallabs.com.
Descript Underlord: Why you should make your AI features “wild”
👉 When AI is the new normal, normal isn’t good enough
How do you get someone to try your new features? Conventional logic would tell you: make the features better than other things on the market. But that can be a slow game. You have to convince people your feature is better.
Descript took a completely different approach. Descript Underlord earned attention when it did something genuinely surprising (turning your voice into a cartoon character’s), the kind of “holy moly, wow” capability that makes people want to play with it. Novelty, not polish, is what drives the first click.
What you’ll learn:
Why incremental AI features get ignored when AI is everywhere
How one or two bold, unexpected capabilities can get users excited
Slack AI: What product teams can learn from one banner
👉 One message, three lessons on the psychology of feature adoption
Slack’s AI features haven’t caught fire. Why? They are operating in an old-school way: the product team builds something and then the marketing team is told to promote it. They get one banner in the product to tell you about it (“AI is turned on”) and the banner isn’t hooked into the workflows. This won’t fly in the AI world. In the AI world, first use is the most important moment and no one wants to try something just because it’s AI.
The fix is to show the immediate benefit, get users to take a concrete first step while they’re still motivated, and make sure a failed first step doesn’t quietly teach them that “AI search doesn’t work.”
What you’ll learn:
Why turning a feature on doesn’t guarantee people will use it
How to give people a reason to try, help them act, and support them when it goes wrong
How to get users to actually try your AI feature: Insights from Zoom’s AI Companion
👉 4 takeaways for AI feature adoption
Zoom’s AI features also haven’t caught fire. This teardown gives you the answer why. The path to Zoom’s AI Companion is a maze of opt-ins that asks for commitment before showing any payoff, so a lot of people bail. The better move is to deliver something valuable first (a transcript, a summary) and ask for the commitment after, treating the whole adoption flow as a product to design rather than a marketing pop-up to dismiss.
What you’ll learn:
How stacked opt-ins kill adoption before the value ever lands
Why you should show value first, then ask for commitment
Mistake-proofing: The secret weapon for successful AI adoption
👉 The Japanese design principle your AI adoption strategy is missing
“Poka yoke” is a term all designers and product managers should know. It comes from manufacturing: it is mistake-proof design. It doesn’t allow the user to make a mistake.
When fintech app Mint launched for the first time, they required that you link your accounts. Why? If you didn’t, the product wouldn’t work. This was risky because it was high friction, but they decided to do it to ensure the user would have a successful journey. For AI, that means embedding it directly in the workflows people already live in instead of building a separate destination they have to remember to visit (the trap Gemini fell into).
What you’ll learn:
How shifting responsibility from users to your system design dramatically increases adoption
Why the most reliable way to get a behavior is to make it the path of least resistance
Want your teams to use AI more? Try this
👉 How narrowing the scope can turn AI ambition into real behavior change
You’re a leader at a growing company and under pressure to get your teams to be AI-native. While it would be easy, the thing you do NOT want to do is tell your teams to “Use AI more.” This is unhelpful. It doesn’t give people a place to start.
What should you do instead? Try a specific, bounded, measurable challenge to motivate your teams: every PM drafts a PRD with one click in the next 30 days, or the research team ships synthetic users by quarter end. Constraints make it easier to experiment and see progress, which is what actually moves a group.
What you’ll learn:
Why fuzzy AI mandates stall and concrete challenges don’t
How to set a bounded first task a team will actually complete
That’s it for this best-of edition. If an AI feature is on your roadmap, whether it’s a brand-new launch, a big relaunch, or finally getting an existing feature used, these five teardowns cover the behavioral science you need.
Want to work with Irrational Labs on AI adoption? We’ve helped companies like LinkedIn, Google, and Intuit get their users to actually adopt new features. Reach out: kristen@irrationallabs.com.
Have a friend who would enjoy these teardowns? Click the button below to refer them.👇
For more deep dives into product psychology, browse the full archive here. A new teardown drops every week. Subscribe so you don’t miss any.
We design products that change behavior, using behavioral science. Check out our case studies to see it in action.
Shh! I have no interest in using AI. Please please do NOT help them anymore.