A Blank Prompt Is a Product Team Avoiding a Decision

The blank prompt box is becoming the default interface for AI products. I think that is often a sign that the product team has stopped one decision too early.

We are seeing prompts positioned as replacements for design tools, assistants offered for complicated household work, agents inserted into enterprise processes and AI connected to sensitive professional systems. At the same time, product teams are being reminded that taste, judgement and practical experience still matter. These developments are connected.

A prompt box gives users freedom, but it also gives them work. They must decide what to ask, how specific to be, which constraints matter and how to recognise a good result. Experts can do this because they already understand the task. Everyone else is being asked to write a miniature product brief before receiving value.

That is not simplicity. It is complexity moved from the interface to the user.

Consider an AI design product. “Create a campaign” looks easier than navigating a traditional editor. But the user still needs to know the audience, format, hierarchy, tone and desired action. If the system does not help make those choices, it may generate something polished without generating something useful.

The same problem becomes more serious in operational products. A family assistant cannot just wait for a clever instruction. It needs to understand whether the user is planning, delegating, reminding or resolving a conflict. A clinical assistant connected to patient information cannot treat every possible request as equally appropriate. An enterprise agent working inside a billing process needs to know which path is normal and when the situation has become an exception.

In each case, the product’s value is not that users can ask anything. The value is that they can reach a sound result without first mastering the system.

This is where product judgement comes back into the work. Teams need to decide what a sensible starting point looks like. They need to expose the few choices that materially change the result and set defaults for the rest. They need to make correction easier than rewriting an increasingly elaborate prompt.

I would rather launch three well-defined starting flows than one universal assistant. For a design product, those flows might begin with the intended channel and audience. For an internal operations tool, they might begin with the job to complete and the relevant policy. For a planning assistant, they might begin with the time horizon, participants and constraints.

The exact structure should come from observing real work, not from debating prompt wording in a meeting. Small experiments are useful here. Give one group an open prompt and another a guided start. Then compare completed outcomes, correction cycles, abandonment and the types of help requests generated. Do not settle for usage numbers that mix curiosity with actual value.

Teams should also review the defaults themselves. A default is a product decision, even when it was generated by a model or inherited from a template. If nobody can explain why it exists, it is an unexamined assumption hiding behind a clean interface.

This requires discipline from product leaders. A broad prompt looks impressive in a demonstration and keeps the roadmap conveniently vague. An opinionated workflow creates harder questions. Which users are we serving first? What does good look like? Which choices can we safely make for them? Where must we ask for confirmation? Those are precisely the questions product teams are supposed to answer.

AI does not remove the need for an interface. It changes the interface from a collection of controls into a sequence of decisions. The product team still owns that sequence.

A blank prompt can be a useful escape hatch for advanced users. It should rarely be the front door. If the main experience starts with “What would you like to do?”, my first question is whether the product team knows what it is for.