Skip to main content

Latent Demand - Product Discovery When Agents Are in the Loop

· 3 min read
Calvin Cheng
Calvin Cheng
Shape what gets built and the value it creates.

How do you know what to build when your users are constantly adapting your product? When AI agents enable power users to "abuse" features for unintended purposes, that behavior isn't noise -it's signal. Latent demand is the idea that you can discover real product needs by building extensible, hackable systems and watching what people actually do.

Beyond User Interviews​

Traditional product discovery relies on interviews, surveys, and hypotheses. Users tell you what they want. You build it. Sometimes it works; often it doesn't. Stated preferences are unreliable. People don't always know what they need until they're doing it.

Latent demand flips the script: build extensible systems, then observe how power users repurpose them. The "abuse" is the feature. Once you see enough people doing the same hack, you productize it.

The Meta Playbook​

Facebook Dating didn't come from a roadmap. Analytics showed that 60% of profile views were between opposite-gender non-friends. Users were already using Facebook for dating; the product team noticed and built a dedicated surface. Facebook Marketplace emerged the same way -40% of group posts were buy-sell transactions. Users had turned groups into marketplaces. Meta built the real thing.

The pattern: make the platform flexible enough that users can jury-rig solutions. Watch the data. When a behavior recurs, build the proper feature.

Agents Change the Calculus​

With AI agents, users can customize and extend products in ways that used to require engineering. Slash commands, hooks, configuration files, plugins -if you expose them, power users will bend them to new use cases. An agent can wire up integrations that your team never imagined.

Boris Cherny (Anthropic) put it simply: "You build a product in a way that is hackable, that is kind of open-ended enough that people can abuse it for other use cases. Then you see how people abuse it and then you build for that."

Design Principles for Latent Demand​

1. Make extension points first-class. Hooks, plugins, and configuration aren't "advanced" features -they're discovery mechanisms. The easier it is to customize, the more creative usage you'll see.

2. Instrument the right behaviors. You need analytics that capture how users combine and misuse features. Not just "feature X was used" but "feature X was used in context Y with integration Z."

3. Quantify before you build. Multiple independent users discovering the same hack is stronger signal than one vocal power user. Prioritize productization when the behavior reaches meaningful scale.

4. Accept that extensibility is the product. This conflicts with tightly controlled, opinionated UX. You're trading some coherence for discovery. The best agent products will be those that embrace the trade-off.

Connection to Systems Thinking​

This echoes how systems optimize for control. Systems that optimize for control tend to close off extensibility. Systems that optimize for latent demand do the opposite -they create space for emergence. The tension is real: too much control, no discovery; too little, chaos. The art is designing guardrails that allow creativity without collapsing into unmaintainable complexity.