From an AI-built prototype to a product people can actually use.

A home-repair job passes through many hands. A call-center agent takes the call, a sales rep closes the deal and dispatches a repair crew, and the owner tries to keep track of everything, including the money.

The client had already built the app with an AI tool. It worked, but the user flows, the journey between roles and the mobile experience were weak. They didn't need more features. They needed UX and UI thinking.

• Call-center agent, answer and route — Leads from every channel in one inbox, AI insights from every call, and an AI-recommended sales rep matched by specialty and distance, assigned in one tap.
• Sales rep, close and dispatch — Matched leads arrive with full call context, so reps can close faster and assign the job to the right repair crew. Crews don't use the app; reps dispatch the work directly.
• Owner, the helicopter view — Every job, every role and the flow of money in one place, so owners can see the whole business without chasing updates.

• Each role gets its own home screen — Same product, but every role opens the app to the one question they care about.
• AI recommends, people decide — The match score suggests a sales rep; the agent makes the final call.
• Every task says why — “Overdue by 3 days” or “Previous agent unavailable” under each task, so nobody has to guess.
• Today and carryover, kept apart — Overdue work stays visible without burying today's tasks.

The AI-built version proved the idea, but it was dashboard-first: dense tables, competing panels and no clear order, squeezed onto a small screen. We rebuilt it as a true mobile app: one clear hierarchy per screen, readable type and spacing, and the next action always within thumb reach.

AI got the product to a first version fast. Design decided who sees what, in what order, and why.

The redesign is now in development. Once live, we'll measure time to assign a lead, missed calls, and how fast jobs move from first call to dispatch.

Shadow a few call-center shifts to see how agents juggle calls and assignments in real time, then test whether the AI match score earns their trust.