Add AI to a Sprint

AI doesn’t replace sprint work — it removes friction around the edges. The highest-leverage moments are repeatable, text-heavy tasks where the output goes through a human anyway.

Good places to start

Ticket drafting — paste a brief description into Synapse Chat and ask it to write a Jira story with acceptance criteria. Edit the result; don’t use it verbatim.

Test generation — open a file in Claude Code and ask it to write unit tests for a specific function. Review carefully; models miss edge cases.

Documentation — ask Claude Code to write a README section based on the actual code. Faster than writing from scratch, easier to edit than a blank page.

Summarising PRs or discussions — paste a long thread or diff into Synapse Chat and ask for a one-paragraph summary before a review.

How to introduce it without disrupting the team

• Start with tasks that only affect your own output, not shared artifacts.
• Don’t commit AI-generated code without reading it line by line.
• If you use AI on a ticket, note it briefly in the PR description — no need for a disclaimer, just transparency.

What doesn’t work well (yet)

• Complex multi-file refactors without careful supervision
• Tasks where correctness is hard to verify quickly
• Anything where the prompt takes longer to write than just doing the task