An AI feature only works when people trust it

AI product design means designing the interface of features built on artificial intelligence so that the user knows when a result can be trusted. It covers how the result, the uncertainty and the error are shown. Squares establishes which decision or task the AI should take over, designs the interaction, builds a prototype and tests it with users.

AI product design is priced per project, for an agreed scope. The first stage covers defining what the AI feature should do and a prototype, and we quote it after a conversation about the feature and the data. Later stages are priced separately, and after each one the team decides whether to order the next.

In a regular interface the same action gives the same result, whereas an AI result can be uncertain. That is why the interface of an AI feature shows the level of confidence, the source of the answer and how to correct a mistake. A model can be accurate and the feature still goes unused when nobody knows when to believe the result.

The outcome is an AI feature interface that users understand and trust, plus a prototype tested with users before any production code is written. This gives the Product Owner evidence that people will use the feature before the budget goes into the build. Changes after testing are made while the work is still in design.

Design before the build makes sense when an existing B2B or SaaS product is getting an AI feature, or when an entire new product is based on AI. It is especially needed in medical and health products, where the user must understand where a result comes from and someone is accountable for an error.

The time depends on the scope of the feature and on the data the AI works with. The work has four stages: understanding the problem and the data, interaction design, prototype and validation with users. We agree the date for the first stage, which ends with a prototype, together with its quote, after a conversation about the feature and the data.

Yes, we design to WCAG and the European Accessibility Act, enforced since June 2025. For medical products we document how the interface shows the uncertainty of a result. In healthcare there is also accountability for errors, so the user sees where an answer comes from and how to correct it.

Squares designed and built the ProseCaptain AI assistant, which runs for two clients: on the Tropokolagen.pl website and in the Daily Effect app. In the first it answers product questions based on the brand's materials; in the second it took over users' repetitive questions. ProseCaptain is also a content editor in which AI writes from the company's knowledge and a person edits the text.

AI product design happens before the code, whereas AI prototype to production starts from code that already exists. The first service produces an interaction design and a prototype for user testing. In the second we review a prototype built with AI tools and bring it up to production standard.

We need to know what the AI should take over, what data it works with and what happens when it gets things wrong. The conversation concerns data and product plans, so a light NDA is signed before the paid stage begins. We run the project remotely, and later stages can be ordered under a framework agreement.