← All articles
AI · 8 min read

How we use AI in development and QA every day

A practical look inside our studio: which AI tools we use, for what, where they save the most time — and the rules that keep quality and client data safe.

We build AI products for clients, and we use AI ourselves every day. Not because it's fashionable, but because it removes routine and leaves more time for the work that needs people: understanding the problem, making decisions and checking quality.

Where AI helps us most

StageHow we use AIWhat stays with people
DiscoverySummarising interviews and documents, drafting user storiesPriorities and product decisions
Design and prototypingQuick UI variants and copy draftsFinal design and brand voice
DevelopmentCode suggestions, refactoring, explaining unfamiliar codeArchitecture, review of every change
TestingTest ideas, test data, automation drafts, log analysisRisk assessment, exploratory testing, sign-off
DocumentationFirst drafts of guides and release notesAccuracy and final wording
SupportTriage and suggested answersAnything sensitive or unusual

The tools

We use AI coding assistants in the editor and on the command line, chat assistants from several providers for analysis and writing, and AI agents that can run tests and propose fixes in a sandbox. We pick models per task: a fast, inexpensive model for routine work and a stronger one for complex reasoning — the same approach we recommend to clients in our guide to choosing a model.

What it changes for clients

  • Faster first versions: prototypes and MVPs arrive sooner, so feedback starts earlier
  • Better test coverage: generating test ideas and data is cheap, so we check more edge cases
  • More time for quality: less routine means more time for exploratory testing and review
  • Honest estimates: we don't promise “AI will do it all” — we plan with real numbers from our own work

Our rules

  1. A human reviews everything that goes to a client or into production: code, texts, test results.
  2. Client data stays protected: we don't paste confidential code or personal data into tools that aren't approved for it, and we follow each client's policy.
  3. We check facts: AI can be confidently wrong, so every claim and every API detail is verified.
  4. We measure: if a tool doesn't save time or improve quality in practice, we stop using it.

An example from QA

For a new feature, an AI assistant drafts a checklist from the user story in minutes. A tester then removes what's irrelevant, adds the risky cases the model missed — time zones, double submissions, permissions — and turns the stable part into automated tests with AI help. The result: broader coverage in less time, with a person responsible for every decision.

AI makes a good team faster. It doesn't replace the responsibility for what we deliver — that stays with us.