How we work. From idea to live software, and we stay on after.
Four phases. We map the work, design the system, build it in weekly cycles and keep it running once it is live.
Map the work
Working sessions with the people who run the process. We find where software and AI will have the most impact and agree on a first release.
- A prioritised roadmap
- A scoped first release
Design the system
Architecture, integrations, data flows and evaluation criteria, agreed with your team before we start building.
- A technical design
- Success criteria and evals
Build in short cycles
We work with coding agents and ship in weekly cycles, with tests and review on every change. You see working software every week.
- Weekly reviews of working software
- Tested, documented releases
Run and improve
We monitor, maintain and extend what we build, or hand it over to your team with everything they need to own it.
- Monitoring and support
- A complete handover when you want one
The models and tools we build with
- Anthropic
- OpenAI
- Gemini
- Mistral AI
- Meta Llama
- DeepSeek
- Ollama
- Hugging Face
- OpenRouter
- LangChain
- LangGraph
- Pydantic
- Qdrant
- OpenTelemetry
- Sentry
- Python
- TypeScript
- Next.js
- FastAPI
- PostgreSQL
- Supabase
- Redis
- Snowflake
- Docker
- Kubernetes
- Google Cloud
- Cloudflare
- Vercel
- Railway
- GitHub
- n8n
- HubSpot
- Stripe
AI in our delivery
The tools we build for clients speed up our own work too, and we answer for every change.
Coding agents
Scaffolding, tests and migrations, delivered in days.
Evals on every release
Answer quality measured against your own cases.
Human review
We review every change before it ships.
Observability
We keep an eye on tracing, costs and errors once it's live.
Ways to work with us
Fixed-scope build
A defined outcome, scope and timeline. The usual way to start.
Ongoing engagement
We keep shipping and running your software, month by month.
AI strategy
A short engagement to map where AI pays off and plan the build.
Common questions
It depends on the scope. We agree the timeline in the written plan after discovery, before any build work starts, and report against it every week.
It depends mainly on the number of systems involved, the state of their APIs, how many exceptions the process has and how many review screens people need. We quote after discovery, in a written plan with scope and timeline.
You meet the people who design, build and run your system during discovery, and you work with them the whole way through.
Usually, yes. If a system has an API, we use it. If it only has a database, a file export or an email interface, we build a connector around that, and we check this in the first weeks.
Yes. Once an invoice is paid, the code and intellectual property in that work transfer to you. The details are in our terms of service.
No. We document the code, prompts, evals and infrastructure as we build, and hand them over with a walkthrough whenever you want your own team or another supplier to take over.
Access on a need-to-know basis, encryption in storage and transit, and data processing agreements with every provider. Where data has to stay in the EU, we use EU-hosted models and regions.
We choose per project, from providers such as Anthropic, OpenAI and Google, and open models where they fit. We evaluate every choice against your requirements and cost.
We keep monitoring and improving the system as an ongoing engagement, or hand it over to your team with documentation and a walkthrough.