Turn AI ideas into something people can actually use.
TJI Labs helps companies work out what AI is genuinely good for, build a first version, and get it tested in a real workflow.
From idea to something that works.
AI projects usually start with a question: could AI help us do this better?
The exciting part is what AI makes possible. The real work is making it land.
Access to powerful AI is no longer the hard part — and that opens up a huge amount. The interesting questions now are:
That’s the part I love: working in the space between an exciting idea and something people actually reach for every day. Sometimes that means building fast. Sometimes it means reshaping the idea. And yes — sometimes it means learning early that AI isn’t the right tool yet. All of it moves you toward something real.
The goal isn’t to add AI. It’s to find the places where AI makes something genuinely, noticeably better — and then build it well.
Sometimes the answer is no.
A while ago, I worked on a project exploring whether generative AI could replace part of a traditional studio production for a high-end Scandinavian furniture brand. On paper, it looked like an excellent AI use case.
We had high-resolution product photography, colour and material references, and production-grade 3D models. We tested some of the strongest image-generation technology available at the time. The results could look impressive. They just weren’t accurate enough.
As lighting and camera angles changed, materials started to feel different. Colours shifted. Small details could change between generations — a sofa with six folds in the reference might suddenly have seven.
Those details were the product.
We eventually delivered a much more conservative solution, combining the original 3D product imagery with AI-generated environments. But the project left me with something useful: AI being capable of producing something doesn’t mean it can produce the thing your business actually needs. That’s one of the reasons I believe in prototyping early.
Sometimes a prototype’s job is to prove that an idea doesn’t work.
So what can TJI Labs actually help with?
AI Product Discovery
Find out what is worth building.For teams with an opportunity, challenge or early idea but no clear product definition yet. Together we examine the problem, users, workflow, available technology and organisational requirements. The result is a clearly defined use case and a practical blueprint for what should happen next.
Prototype Build & Validation
Make the idea real enough to learn from.A prototype changes the conversation. Instead of discussing what AI might do, users and stakeholders can experience something, respond to it and expose the assumptions that matter. TJI Labs can design and build contained AI prototypes and products, test them with users and iterate toward a validated solution.
AI Product Leadership
Take the project beyond the prototype.Some projects need continued product leadership without requiring a permanent senior hire. TJI Labs can work alongside internal teams, developers and external specialists to translate objectives into product decisions and keep development moving toward something useful. For larger technical projects, this can also include helping identify and assemble the people required for execution.
Built from experience, not theory.
TJI Labs is informed by several years spent building AI products, testing them with real users and discovering what happens when promising technology meets actual organisations.
AI products for film & television marketing.
Co-founded and led the development of an AI platform designed to understand film and television content and turn that understanding into practical tools for creative marketing teams. The journey has included product strategy, team building, funded R&D, AI development, enterprise testing, onboarding and implementation work with organisations including DR and Nordisk Film.
Trailr.ai has received support through InnoFounder, Innobooster and AWS Proof of Concept funding, spanning AI model development, product interfaces, user validation, workflow integration and customer onboarding.
An AI-native life companion built around memory and context.
MountainRiver began as a personal product idea and developed into a full PWA combining journaling, long-term memory, AI reflection, decisions and goal pathways. It has been independently conceived, product-designed and built from the ground up — including its interaction model, AI behaviours, data structure, Supabase backend, Netlify deployment, privacy architecture, onboarding and iterative testing.
The working product uses React/Vite, Supabase and Netlify, with substantial thinking around persistent context, AI interactions, data integrity, deployment and product behaviour.
20 years making things.
Three years making things with AI.
I'm Tommy Duncan, founder of TJI Labs and co-founder of Trailr.ai. Before working in AI, I spent around two decades in media, marketing and creative production in London and Copenhagen, including senior roles at DR and Discovery.
I worked across creative development, campaigns, editing, production and project leadership — inside organisations where ideas ultimately have to survive deadlines, stakeholders, users and real-world constraints.
Since moving into AI product development, my work has expanded across product strategy, prototyping, team building, innovation funding, enterprise implementation and hands-on development. I have led funded development programmes, recruited multidisciplinary teams, worked closely with AI and software engineers and independently built an AI-native software product from concept through working beta.
The combination is what TJI Labs is built around: understanding the problem, communicating across disciplines, and getting far enough into the build to discover what is actually true.
Have something you think AI might solve?
You don’t need a specification. You don’t even need to know exactly what the solution should be. Bring the problem — and let’s work out what’s possible.
Let’s work out what’s possible.