How much does an AI feature cost in a product, from idea to launch

The cost of calling a language model is rarely the bulk of the budget. What costs is finding the right use case, designing screens users trust, connecting the feature to your data and checking that it complies with regulation. Here are the four steps of a typical project, with our own prices, to give realistic orders of magnitude.

1. Scope: find the use case worth building

Before pricing any development, you need to know what problem the feature solves, for whom, and how you will measure that it works. It is the shortest step and the one that avoids the most wasted spending: an idea that does not hold up with users costs a few days of interviews, not three months of code.

• What we do: interviews with your users and teams, analysis of the current journey, choice of one or two use cases.
• Duration: a few days to two weeks.
• At Kapti: included in the AI Sprint, or as standalone UX research , on quote.

2. Prototype with a real model and test

A static mockup is not enough for an AI feature: the model is sometimes wrong, answers slowly, hesitates. You need to test the real experience, flaws included, with five users. That is where we decide how to show uncertainty, sources and the control left to the user (we detail these choices in this article on trust).

• What we do: screens, a prototype connected to a model, user tests, recommendations.
• Duration: three weeks.
• At Kapti: the AI Product Sprint, from €16,500 excl. VAT, with a designer, an AI developer and a PM.

3. Build and connect it to your data

This is the most variable step. A feature that summarises a document already in your tool has nothing in common with an assistant that queries several internal databases. What drives the budget:

• data access (existing or new APIs, permissions, data quality);
• security and hosting (model hosted in Europe, data that never leaves your information system);
• monitoring answer quality once live;
• integration into your design system and existing screens.

As an order of magnitude, count from a few weeks for a targeted feature in an existing product to several months for an assistant connected to many sources. After the prototype, we price this step at a fixed fee: you know what you pay before you start.

4. Check compliance

The AI Act imposes transparency obligations on interfaces that use AI: telling users they are interacting with an AI system, flagging generated content. If your product also serves consumers, the European Accessibility Act has applied to its screens since June 2025. Better to plan for it from the design stage than to fix it afterwards.

• At Kapti: theAI screens audit, two weeks, from €8,000 excl. VAT, and accessibility compliance, from €8,000 excl. VAT.

What to remember for your budget

• Do not start with development. A few weeks of scoping and prototyping cost far less than a feature nobody uses.
• The cost of the model matters little at first. Data, screens and quality are what make the budget.
• Ask for a fixed price per step. Each step delivers something usable, and you can stop if the idea does not hold up.

Frequently asked questions

How much does an AI feature prototype cost?

At Kapti, the AI Product Sprint starts at €16,500 excl. VAT for three weeks: scoping, screens, a prototype connected to a real model and tests with five users.

What drives the budget of an AI feature?

Mainly data access, security and hosting, monitoring answer quality, and integration into existing screens. The cost of calling the model weighs little at first.

Should the AI Act be in the budget?

Yes, from the design stage: transparency obligations directly affect screens. An audit of AI screens takes about two weeks.