Skip to content

AI products that feel obvious, not magic

Most AI products are a text box and a hope. I design the parts that make AI useful: setting expectations, showing what the model did, letting people correct it, and earning trust one good answer at a time.
Start a project
A traveller pushing his head through the edge of the sky to see what lies beyond

The problems I hear most

  1. .01

    A blank prompt box and nowhere to start

    People don't know what to ask, so they ask something vague, get something vague back and leave.

  2. .02

    Users can't tell when to trust the output

    Without sources, confidence or a way to check, people either over-trust the model or stop using it.

  3. .03

    The demo works, the daily habit doesn't

    A great first answer is easy to design for. A workflow people return to every day needs memory, editing and handover.

  4. .04

    AI bolted onto an existing product

    A sparkle button in the corner rarely changes behaviour. The value comes from putting AI inside the job people already do.

What good looks like in Generative AI

  • Starting points and examples that teach what the product can do
  • Answers that show their sources, confidence and reasoning
  • Easy ways to edit, retry, undo and correct the model
  • Agent flows where people approve the important steps
  • A brand and interface that feel considered, not like every other chat app

Work in Generative AI

All Generative AI work

How I can help

  • Product & UX Design

    Apps, websites and dashboards for products with real complexity underneath. I design the flows first and the pixels second, so the result is calm to use and straightforward to build.

    Discuss this

  • UX Research & Audit

    Evidence before opinion. I find out what users actually do, where the product loses them, and which fixes will move the numbers you care about.

    Discuss this

  • Branding & Marketing

    Identities that are easy to recognise and hard to forget, plus the campaign and AI-assisted graphics that keep them consistent once they are out in the world.

    Discuss this

Generative AI, asked often

  • I need to understand its behaviour: what it is good at, where it fails, how fast it is. I'll spend time with your outputs early, because the design has to fit the real model, not the ideal one.

  • Yes, and that is often where AI earns its keep. I start from the job people already do in your product and find the moment where AI removes the most effort.

  • Short, frequent sessions with real tasks. I test the flow and the trust signals rather than a specific model version, so findings stay useful as the model improves.

Other industries