Coding
Use AI to accelerate development while keeping review, testing and security controls in place.
What good looks like
A good tool improves the outcome or reduces the time required without creating unacceptable privacy, cost, quality or dependency risks.
A practical test
- Use a disposable branch
- Run tests before merge
- Review dependencies and secrets
- Measure time saved, not code produced
Common trap
Testing on a polished demo instead of your real work. Use the same task, source material and success criteria for every candidate.
FAST questions for this category
- Fit: Does it improve the category-specific outcome you actually need?
- Access: Can the intended user learn and operate it within the available workflow?
- Scale: What happens to cost when usage, exports, collaborators or automation volume increases?
- Trust: Are data handling, rights, reliability and review controls acceptable?