Chapter six
Then AI offered another set of possibilities.
AI can help with real work. It can help shape a message, organize information, or make a conversation possible across languages. I build with it because those uses can matter to people.
It can also become another layer to manage if the purpose is unclear. A tool that produces more material does not necessarily bring someone closer to the answer they need. The amount of output is not the same as the amount of help.
I wanted to judge an AI feature in the same way I would judge another part of the app. Can the person understand it? Does it produce something useful for their task? Can they tell what to do with the result? Those questions make the discussion more practical than treating the presence of AI as a benefit in itself.
I wanted to use AI with a job in mind. What is the person trying to do, and what would a useful result look like? The question was familiar, even though the technology was new.
The AI layer
I wanted help I could explain.
An app needs to be clear about what it can do, what information it uses, and where its limits are. That matters especially when a fluent answer can sound more certain than the evidence behind it.
The useful work includes setting boundaries, checking results, and leaving appropriate decisions with the person. A tool can help someone prepare a message without deciding to send it for them. It can organize information without turning a guess into a fact.
A boundary can make an app easier to use as well as easier to trust. When the job is clear, the person knows what kind of help to ask for and what to do with the answer. A tool that tries to do everything can leave them with another decision to untangle. I wanted our products to make the next useful action more obvious.
Those choices belong in the product, not just the promise on its website. I wanted the same care behind the screen that a reader would expect from the words in front of it.