Services / AI Product Design

AI Product Design

We design interfaces for AI products: how they explain themselves, handle uncertain output and earn the user’s trust.

Start a project

What we deliver

Design for how the model actually behaves.

AI interface design

Screens and flows built around what a model can reliably deliver.

Agentic workflow design

Multi-step, autonomous tasks a user can direct, pause, and trust.

First-run and onboarding for AI

Showing people what the product can do from the first session.

Trust and confidence design

Showing certainty, uncertainty, and everything between honestly.

AI component libraries and design systems

Reusable states for loading, streaming, correcting, and failing.

Prompt and input UX

Interfaces that turn a blank box into a guided, recoverable input.

Usability testing with simulated model output

Real sessions that test how people handle model errors.

Motion for streaming and uncertain states

Motion that shows what the model is doing.

Conversational UI

Dialogue patterns that stay steady when a user goes off-script.

Process

Seven steps,
in this order.

01

Model behaviour audit

We map capabilities, failure modes, latency profile, and output variance before any design work starts. This becomes the constraint set every later decision has to respect.

02

User mental-model research

We find out what users expect the AI to do, and where they over-trust or under-trust it. The design gently corrects for both.

03

Information and intent architecture

We decide what the product surfaces proactively, what stays on demand, and how uncertainty gets communicated. Everything on screen has a reason to be there.

04

Interaction and prompt design

We design how users direct the model, how it handles ambiguity, and how it invites correction. The interface makes mistakes easy to recover from.

05

Visual and motion design

Streaming, progressive disclosure, confidence states, graceful degradation. Every state the model can be in gets a visual language.

06

Prototype and test

Interactive prototypes run against simulated outputs, including edge cases, in moderated sessions before engineering starts. We find the failures while they’re still cheap to fix.

07

Handoff and design system

Specs, AI-specific component states, and documentation that scales as the model changes. Your team can extend it with confidence.

Why V9

Why teams
choose V9.

Experienced team

Every project is led by people who have shipped AI products before.

Embedded with your engineers

We join your sprints and work directly with your engineers.

Measured by outcomes

We are measured by what ships and how it performs.

Designed from model behaviour outward

The interface follows what the model actually does.

States for loading, streaming, uncertain, correcting

Components that show what the model is doing in real time.

Working sessions with your ML team before design locks

We sit with the people who know the model’s limits before we commit to a direction.

Related work

Ways to start

Start with
your numbers.

FREE

Homepage teardown

Send us your website. Within three working days you get a short video walkthrough of the three changes most likely to lift your conversion rate.

ONE WEEK

Funnel audit

We review your analytics, pages and sales feedback, then hand over a prioritised plan with the target metric for each fix. Fixed fee, credited against the project if you go ahead.

  • Analytics and conversion review
  • Page-by-page findings
  • Prioritised roadmap
Book an audit call

Strategy call

Pick a time that suits you.

Calendar not loading? Open it in a new tab.