The Persona You're Ignoring Will Break Your Product

Product teams are making a strategic error by treating AI as a feature rather than as a demanding, screenless user, according to a developer's blog post. The post argues that AI agents lack common sense and intuition, so products must be redesigned to support intent-based actions, explicit context, and granular trust levels. The author urges teams to add an 'Agent' persona to their persona walls and measure success by tasks completed without human correction.

Product teams obsess over their persona walls. We meticulously map out the needs of the "Stressed Admin," the "Power User," and the "Busy Executive." But when AI arrived, it didn't earn a persona card. Instead, it became a lazy line item on the product roadmap. That is a massive strategic failure. If you treat AI merely as a feature, you get what everyone is shipping right now: a superficial chat bubble in the corner, demoed at an all-hands, and quietly abandoned by users by month three. One lesson i remember Moe Ali taught in one of his classes, product people's fair full focus should be on contrast, if contrast is what you're looking for; read below carefully. AI is a User, and It Has Zero Common Sense To survive the next era of software, you have to stop treating AI as an add-on and start designing for it as the most demanding, alien user on your platform. Right now, human users are covering up your bad product design for free. A human intuitively knows that a yellow banner means "manager approval needed," or that "Room 4B" shouldn't be booked for a VIP because it's next to the noisy kitchen. The Agent is a terrifyingly literal user with infinite patience but absolutely no intuition. It cannot read visual cues or understand unwritten company rules. If your product relies on human workarounds or visual context to function, this user will fail, and it will confidently report a 100% success rate while booking the wrong room and offending your VIPs. Designing for the Screenless Persona To build for this new user, product managers must radically rethink their acceptance criteria. A screenless user requires a totally different product philosophy: Design for intent, not clicks: Stop forcing agents to mimic human screen flows. If a capability only exists as a tedious sequence of UI clicks, the agent can't use it. They need consolidated, meaningful actions verbs , not simulated button presses. Context over raw information: If the operational context living in your team's head e.g., "this room is loud" isn't explicitly defined in your product's logic, it simply does not exist. Granular trust levels: Autonomy is not a master switch. While finding information should be instant, any action that affects the real world, like canceling a booking; must require human confirmation. The Uncomfortable Reality for Product Teams The hard part of AI isn't the underlying technology; it's the product decisions. Taking this seriously means reopening core features that "work perfectly well for humans" and admitting they are completely illegible to a machine. It means your new primary metric isn't "chat sessions initiated," but rather "tasks completed without human correction." That number is usually humbling. Add the Agent to your persona wall today. Give it needs, constraints, and a North Star metric. Until every new feature spec asks, "Can a screenless agent complete this end-to-end?" your product isn't actually AI-ready. Persona: Agent Needs: meaning over data · verbs over screens · permission with gradations Frustrations: undocumented rules · state that exists only on screen · success responses that aren't true Selects on: how you describe yourself, not how well you work Churns: silently, mid-sentence, without telling you North star: completes the task correctly, without a human checking Put it on the wall. Then go read your own tool descriptions.