Services
How we take AI software from idea to production
Sprint 0 prepares the ground once. After that, every stage runs in short iterations and ends with your feedback, because with AI, "it worked in the demo" is where the real work starts.
- Consulting
- Training
- R&D
- Prototyping
- PoC & MVP
- Production systems
- Maintenance
- Reports & documentation
On day one we talk. In Sprint 0, which takes about 5 weeks, we find out what you really need, check what is possible and write it down as a plan a developer can build from: the proposed solution, how it should work and where the risks are.
Then a developer puts the core together as a Proof of Concept, usually in about a month. You test it early and tell us what works, from the layout to whether the AI gives sensible answers.
In the MVP, usually 2–3 months, we build real value in rounds, shaping the product with each round of your feedback. When it works for its users, we take it to production: ready for many users or built into your existing systems.
What happens at each stage
Consulting & Sprint 0
Three steps, each built on the previous one. Starter is a workshop that maps what to prepare. In Preparation we review your data as experts and tell you whether the idea holds up. Ready adds a clickable mockup, so the screens are settled before the PoC starts.
Ends with the feasibility report (and the mockup, if you add it), a walkthrough and a decision on the PoC, made by both sides.
- Plan and data checklist
- Feasibility report
- Optional UX mockup
Proof of Concept
A developer assembles the core of the solution as fast as possible and shows it to the people who will use it. If you took the mockup step, the screens are already agreed, so the PoC focuses on the AI: does it give answers that make sense on your data, and which problems are coming.
Ends with your feedback, which becomes the input for the real build.
- Working prototype
- Feedback summary
- Scope of the MVP
Minimum Viable Product
Here the goal is value for real users. We work in iterations with developers, improve the product after every round of feedback and give it its shape. This stage takes longer than the previous ones, and the result doesn't have to be perfect yet. It has to be used.
Ends when the product works for its users. Source code and IP rights are yours.
- Working application
- Source code and IP rights
- List of next challenges
Full-scale production
When the MVP succeeds, we prepare it for production. Depending on what it is, that means handling a large number of users or implementing it as a component of a real system, with monitoring, performance tests and clear responsibilities.
Each release brings feedback from real use, which starts the next loop.
- Production deployment
- Maintenance agreement
Consulting & Sprint 0
Three steps, three decisions
Sprint 0 runs in three steps. Each one builds on the previous, has its own price and ends with a decision. Yours and ours: if your data can't carry the project, we'll say so before you pay for the next step.
- You can skip Starter if you already know what you need. Send us the materials from our checklist and we start with Preparation.
- The timing also depends on when we get materials from your team. Yes, we'll need your input!
We created Sprint 0 for companies without their own technology team. It describes, in technical terms, what you really need and how it could work.
It is also how we get to know each other: you see how we work and communicate before anything bigger starts. Each step ends with a decision about the next one, so you always know what you're agreeing to.
Consulting on its own is billed by the hour. Sprint 0 is something else: a package we put together from experience, with what teams usually need at the start. It's also a good way to begin working together.
Starter
Map the route · Workshop & plan
- Price
- $2,000
- Time
- about 2 weeks
- Initial meeting
- Workshop preparation (we'll also ask you to prepare your sources)
- 3-hour workshop with your team
- Report
- Workshop summary
- Goals and success criteria
- The data and materials to prepare, with owners and dates
- Indicative costs of the road ahead
Both sides decide on Preparation
Preparation
Check the ground · Expert review of your data
- Price
- $3,000
- Time
- about 3 weeks after we get the data
- Expert review of your data and materials
- Quick technical test
- AI compared with a simple no-AI baseline
- Report
- What we see in your data
- Scenarios and our recommendation
- A verdict: go, go with conditions, or stop
- Architecture, risks, running costs and a PoC estimate
Both sides decide on Ready, or go straight to the PoC
Readyoptional
Ready to climb · UX mockup
- Price
- $1,000
- Time
- about 2 weeks
- Screens and user flow design
- Clickable mockup with sample data
- Click-through with your team and feedback
- A narrower PoC estimate, because the UX is settled
Skip it and we factor the UX work into the PoC estimate instead.
Both sides decide on the PoC
Alongside software
Scientific research
Research follows the same route: first preparation, then the work. Instead of an application, the result is an answer to whether something is possible.
- 1
Application & plan
Like Sprint 0, but for research: we write the grant application with you and plan the stages, from the hypothesis to the methods and milestones.
- 2
Research stages
We carry out the work stage by stage, as planned, testing the hypothesis in experiments and documenting every result.
- 3
Results
A clear answer: feasible or not, and why. With the documentation the grant requires.
You get: Research results · Grant documentation
Talks & training
Teach your team to work with AI (safely)
Talks and hands-on workshops prepared for your team: what AI is good at (and where it makes things up), real use cases from our projects, and how to use it with company data without sending that data where it shouldn't go.
FAQ
Questions we hear before the first call
How long does it take to build an AI MVP?
Sprint 0 takes about 5 weeks (plus about 2 if you add the Ready step), a Proof of Concept about a month and an MVP 2–3 months. How long production takes depends on what the product needs: more users, integrations, monitoring. After Sprint 0 you get a plan for the next steps of your project.
How much does it cost?
Sprint 0 has fixed prices per step, listed above, and each step ends with a decision, so you never commit to more than one step at a time. The PoC, the MVP and production depend on the scope, so we quote them after Sprint 0, when we know what we're building.
Who owns the code?
You do. In the MVP stage the source code and the intellectual property rights are transferred to you. We are also happy to sign an NDA before you share the details of your idea.
Do we need our own tech team?
No. We designed Sprint 0 for companies without one: it describes the solution in a way a developer can build from, and we can build it too.
Can you work with confidential data?
Yes. We have built solutions on on-premise LLMs running on the client's own infrastructure, where the data never leaves the company. One of them ran on 4 GPUs with no internet connection.
Can AI answer questions from our own documents?
Yes. That's RAG: we find the right paragraphs in your documents and hand them to the model together with the question. Only a bit of magic, mostly math. Our legal search engine cites the exact act behind every answer.
Do you build AI agents?
Yes. An agent is a model that takes steps, not just answers: it checks the calendar, pulls the file, fills the form. Our own voice agent answers a real phone line and hands the call to a person, with a summary, when it should.
What happens after launch?
Each release brings feedback from real use and starts the next loop. Production and maintenance are covered by a separate agreement that sets clear responsibilities.
I built a prototype with an AI app builder. Can you take it from there?
Yes. In Sprint 0 we look at what you have and tell you honestly what can stay and what should be rebuilt before real users arrive.
Not sure where to start?
Tell us about your idea. We'll suggest the first step, or the stage you can start from.