Test what your AI agent assumes before it acts
When an agent proposes something big, find out whether the assumption behind it holds on a small group before it reaches everyone.
See the record →Use cases · Operators running AI agents
Keep AI agents from re-proposing ideas you already dropped
Your agents can read past Decision Records before they propose, so a dropped idea stays dropped unless new evidence says otherwise.
The situation
Every few weeks an agent suggests a free tier, as if for the first time. Your team tested it in the spring and dropped it. Each time, someone has to dig up the reasons, or the agent researches it again from zero.
What you find out
- The next time an agent raises a free tier, it can read this record first and say what was already tested.
- If something real has changed, SwiftCNS proposes a new experiment instead of repeating the old idea.
- A dropped idea stays dropped, with the reason attached.
Decision Recordworked example
- assumption
A free tier will bring in more paying teams than a 14-day trial.
- experiment
Offer a free tier to half of sign-ups from organic search for six weeks. The other half gets the trial.
- success criteria · locked
The free-tier half produces at least as many paying teams by week six. Locked before the free tier goes live.
- evidence
Sign-ups rose, but the free-tier half produced 31% fewer paying teams.not met
decision
Drop it
More sign-ups, fewer paying teams.
Team decides; agents can read past records through the SwiftCNS MCP before they propose something new.