AI product consulting, from use case to shipped prototype

Every product team now has a list of AI ideas. A few of them will change how customers work. Most will ship as a chat box that nobody opens twice. The difference shows up long before the build, in which use case you pick and how much care goes into the moments where the model is unsure or wrong.

We help you choose the two ideas worth building and put a working version in front of your users. You see it perform before you commit a build budget, and you keep the reasoning behind every call.

Built for

• A founder with a funded idea and no product team yet. You need to decide what to build first, see it working and brief the people you hire.
• A product team under pressure to "add AI". Leadership wants a launch this year. You want a use case that earns its place in the product.
• A company with a stalled pilot. The demo impressed the board and then nothing shipped. You need a path from pilot to something customers use.

Deliverables

• Use-case scorecard. Every candidate use case scored for user value, risk and feasibility, including data access, cost per task and latency.
• Jobs map. The jobs your users do, drawn from interviews or support logs, with the steps where AI helps and the steps where it gets in the way.
• AI PRD. What the feature does, what good output looks like, the success metrics and the eval plan.
• Designed states. Flows and UI for the full range: loading, confident, uncertain, wrong, corrected and handed to a person.
• Testable prototype. A working prototype built with AI coding tools, tested with real users before anyone writes production code.
• Build plan. Scope for the MVP, the open risks and a 90-day path from prototype to production.

Working together

1. Frame. A clarity session or kickoff on the business goal, the users and the constraints.
2. Choose. We gather evidence and score the use cases with you.
3. Design and prototype. We design the chosen flow with its failure states and build a version people can use.
4. Test. Real users try it. We change what they stumble on and record what we learned.
5. Hand over or carry on. You get the artefacts, or we continue into the MVP.

Most teams start with a clarity session or a sprint. See the engagement formats for scope and pricing. We work remote and async first from Kuala Lumpur (UTC+8), with a weekly live call in the overlap window.

The craft difference

• The failure states come first. An AI feature that looks right in the happy path and breaks in front of a customer loses that customer's trust fast. We design recovery before polish. AI error handling UX shows the patterns.
• Real users before real code. A prototype or a Wizard of Oz test, with a person playing the model, costs a fraction of a build and tells you whether anyone wants the thing.
• Nothing ships straight from the generator. We use AI tools to move fast, and every screen still gets a critique pass and a user check before you see it as finished.

Proof

Marcus teaches PRD-first prototyping as the named instructor of AI Vibe Coding and Rapid Prototyping at AiTraining2U, where participants build working apps in two days. More on his background is on the about page.

TBC: first consented case study for this track.

Related insights

• AI pilot to production: scaling user-centred AI
• AI error handling UX: when the AI gets it wrong
• MVP development: build the right thing first

If the product question turns into "is this feature working?", the product management track picks up evals and roadmaps. If you want to learn to run this process yourself, the founders' coaching path covers it one-on-one.