Every product team we talk to is asking the same question: how much of our workflow should AI take over? After shipping AI-assisted features across several client products this year, our answer is more nuanced than the hype suggests.
AI has genuinely transformed three parts of our process: research synthesis, first-draft interfaces, and test coverage. What it hasn't changed — and won't any time soon — is the judgment about what to build and why.
Where AI actually accelerates delivery
The biggest wins come from compressing the boring middle: turning research notes into themes, generating variations of a settled design direction, and writing the tests nobody wants to write. On a recent SaaS engagement this cut our discovery-to-prototype time roughly in half.
Where craft still wins
Positioning, information architecture, and the hundred small judgment calls that make a product feel coherent — these remain stubbornly human. The teams that win pair AI acceleration with strong product taste, not one or the other.