The AI feature, measured
The core LLM, RAG or generation flow, tested against an evaluation set built from real inputs, so you know how often it is right before users rely on it.
Guardrails, privacy and cost limits
Output checks, a fallback when the model is unsure, personal data kept out of prompts that do not need it, and a usage budget so the model bill stays predictable.
The product around the model
Accounts, billing, the web or mobile app and an admin panel, with the code, prompts and evaluation set handed over to you.
Common questions
How much does it cost to build an AI MVP?
With us, an AI MVP starts at €15,000 and takes 2 to 6 weeks: one core AI feature with evaluation and guardrails, accounts, billing if you charge, and an admin panel. Several AI features, training your own model or many integrations cost more. You get one fixed price in writing within 48 hours of a scoping call. Book a free 30-minute call to get yours.
What does an AI app cost to run after launch?
The build is a fixed price, running it is not. Model and API usage is billed by the provider you use, such as OpenAI or Anthropic, and it grows with how much people use the product, so nobody can honestly quote it as one flat number. What we do is work out the cost per user action before launch, set a budget and design against it: caching, batching and the smallest model that passes evaluation. Hosting is billed by your cloud provider, on your own account. More in shipping LLM features that survive real users.
How do you make sure the AI gives correct answers?
We measure it. Before launch we build an evaluation set from real inputs, messy and adversarial ones included, and score the feature against it, so we can say how often it is right and what happens when it is not. Assistants answer from your sources with citations and say "I don't know" when the sources are silent. Every prompt change or model upgrade runs against the same set before it reaches users. How we do it: RAG that cites its sources.
What happens to my users' data in an AI MVP?
It stays under your control. We send a model only the data a task needs, use provider settings that keep your data out of model training where the provider offers them, and build for GDPR. When personal data has to stay in the EU, we choose a model and hosting that keep it there. The accounts, the code and the data are yours.
What is the difference between an AI MVP and AI integration?
An AI MVP is a new product built from the ground up with AI at its core: first version, fixed price, weeks. AI integration adds AI to a product you already run. For image and video generation pipelines or custom models, see generative AI development, for agents that take actions, AI agent development, and for agents that answer the phone, voice AI agents.
Have you built an AI app before?
Yes. TaleTuck is our own AI bedtime story app, live on iOS and Google Play. Two AI pipelines work together: one writes a new, age-appropriate story from a child's name, age and interests, the other clones a parent's voice from a 15-second recording and narrates it. It runs on a subscription with a free tier. Read the TaleTuck case study.
How do I get started?
Book a free 30-minute call. We talk through the product, the AI feature and what a good answer looks like, and within 48 hours you get the scope, a fixed price and a delivery date in writing. If AI is not the right first step, for example because a simpler version would prove the idea faster, we will say so on the call.
How we build.
Unit and feature tests with PHPUnit / Pest. Standard, not an add-on.
Automated tests and deploys on every push. No manual releases.
Every line reviewed by a senior engineer. No juniors on your budget.
Full source code, infrastructure and documentation transfer on handoff.
Book a free 30-minute call. You get a fixed quote and a launch date in writing within 48 hours.