A pilot rollout of Breeze in a 200-person sales org stalls in week three. Reps are not using the agent, leadership cannot tell why, and the explanation surfaces in a hallway conversation: “I don’t know what it’s doing to my deals.” Adoption was not a feature problem, it was a visibility problem. Audit Cards are the fix HubSpot shipped, and using them as an enablement tool changes the rollout entirely.
What users see in an Audit Card
After every Breeze agent interaction, an Audit Card appears in context — on the contact record, deal record, ticket, or wherever the action took effect:
Audit Card: Prospecting Agent
Run at: 2026-04-28 09:14 UTC
Trigger: Daily ICP prospecting
Action: Created 3 contacts, scored 12, recommended 4 for outreach
Contacts created:
Rachel Liu - Acme Corp - VP Eng - score 78
Marco Vega - BetaCo - Director - score 65
Hana Tanaka - GammaInc - VP Ops - score 71
Reasoning:
"Matches ICP signals: company size 500-2000,
industry SaaS, recent funding event detected via
Crunchbase signal."
Users see what changed and why, not “the AI did something.”
The adoption mechanism
Trust gates adoption. The pattern in every previous AI rollout I have watched:
Day 1: "Cool, let me try it"
Day 7: "Wait, did it just update my deal stage?"
Day 14: "I'm not sure what it's doing"
Day 21: rep stops using the tool
Audit Cards interrupt the cycle at day 7. The rep sees the decision trail, validates the action, and either accepts or corrects it. Trust accumulates because evidence accumulates.
Training playbook
Make Audit Card review a daily ritual during the first 30 days:
Week 1: Review every Audit Card. Flag wrong changes.
Week 2: Review Audit Cards on owned records. Spot-check others.
Week 3: Review weekly summary. Track personal correction rate.
Week 4: Review only flagged exceptions.
Pair each rep with a buddy to compare flags. Patterns emerge that no individual would catch alone.
Feedback loop into agent prompts
Corrected changes are gold for prompt tuning. Build a workflow that captures every flagged correction with context:
// Workflow on Audit Card flagged
const flag = event.inputFields;
await postToOpsBoard({
agent: flag.agent_name,
card_id: flag.card_id,
expected: flag.expected_outcome,
actual: flag.actual_outcome,
reason: flag.user_explanation,
record_url: flag.record_url
});
return { outputFields: { logged: true } };
RevOps reviews the board weekly, batches corrections by pattern, and ships prompt updates. Users see their feedback shape the agent — adoption compounds.
Operational metrics
- Audit Card review rate (% of cards reviewed by owner)
- Correction rate (flagged / reviewed)
- Time-to-flag (when issues are caught)
- Repeat-correction patterns (the same agent error twice)
- Agent confidence vs correction rate (calibration signal)
A low correction rate without a corresponding low review rate is the signal of healthy adoption. A low correction rate with a low review rate is a signal that nobody is paying attention.
Org-wide rollout sequence
Phase 1: 5-rep pilot, daily Audit Card review, weekly retrospective
Phase 2: 1 team, weekly review, prompt tuning cadence
Phase 3: Department, monthly review, exception-driven tuning
Phase 4: Org, dashboard surveillance, quarterly governance review
Each phase only moves forward when the previous phase shows correction rate trending down and review rate staying high.
What to do this week
Make Audit Card review part of your daily team standup for the agent rollout, build the feedback workflow above, and measure correction rate per agent before scaling beyond the pilot team.