Which activities can benefit from AI
AI can prepare implementation options, tests, documentation and repetitive changes. Requirements, architecture and compatibility with existing systems need to be verified by the responsible team. The speed of code generation is not the same as the speed of delivering a product.
What needs to be checked before use
Proposed code is reviewed against the requirement, data access, dependencies and error handling. Tests should check the outcome the user needs, rather than simply repeat the implementation. Releases follow the project's acceptance criteria.
Project data and the tools we use
We establish which information can be used with AI tools and what restrictions apply to providers, accounts and environments. Confidential documents and data do not automatically become permitted inputs to an external service.
See project security and data management.
How we assess whether AI helps
Measurement needs to define the activity, period, complexity and accepted outcome. We also track review, rework, defects and tool costs. Comparisons between different projects require an explanation of differences in scope and team.
Savings observed in one activity are not a guaranteed reduction in the budget of the whole project. The proposal and timeline are based on your requirements and dependencies.
What to ask in a project discussion
- Where are AI tools used, and with what data?
- Who checks the outcome, and against which criteria?
- What deliverable can be demonstrated and accepted?
- How are costs, errors and changes tracked?








