In our pipeline, AI drafts a lot: flows, components, tests, translations. None of it ships without a person who signs it. This is how we split the work between machines and people, stage by stage.
Speed alone is not the point
AI makes the first draft cheap. That changes how a project feels: more options early, faster feedback, less time on repetitive work. It does not change who is responsible. Clients do not hire a model; they hire a team that answers for the result.
So we set one rule before choosing any tool: every change that reaches a client has a named engineer behind it.
Where AI sits in each stage
- Discover. AI summarises research and drafts questions. The team runs the workshop and writes the scope.
- Design. AI drafts many flows quickly. Designers keep the few that work and test them with users.
- Build. AI pairs on code and tests. An engineer reviews every diff before it merges.
- Verify. AI runs regression and accessibility scans. A person checks the delivery against what was agreed.
The speed comes from the machine. The signature comes from a person.
Measured, not assumed
Every time we bring AI into a task, we record how long it took before and after, and what the review found. Some tasks get much faster. Some do not, and we stop using AI there. The numbers decide, not the hype.
task: component + tests before: [hours] after: [hours] review: 2 changes requested ✓ signed off by a named engineer
What this means for clients
Projects start faster and reach working software sooner. Decisions stay with people you can talk to. And at the end of each cycle you see, in numbers, what AI changed in the work.
Machine speed. Human signature. That is the line we work by.
Co-founder of Coopers. Leads operations, R&D and the adoption of AI across the company.
LinkedIn