The fastest way to adopt policy checking is not to switch everything on at once. Start with one team and one renewal month, let the agents prepare and the people decide, and measure the result from the product’s own record. This is a suggested plan, not a promise of outcomes.

gantt title A suggested first ninety days dateFormat X axisFormat Day %s section Set up Taxonomy review with the team :a1, 0, 14 Controls set to your standards :a2, 7, 21 section Run One team, one month of renewals :b1, 14, 60 Teach labels as they appear :b2, 14, 60 section Measure Read the record and decide next step :c1, 60, 90

Days 0–30: set it up

  • Pick the lines. Property and GL renewals give the most varied forms; a mix teaches the taxonomy fastest.
  • Review the taxonomy with two account managers: the synonyms your carriers use, the severities you care about.
  • Set the controls to your standards — the large-account threshold, the premium-jump percentage, the late-delivery days.

Days 15–60: run one team

  • Every policy through Check; every prepared decision answered. Labels confirmed as they appear.
  • Corrections sent from the product so every one has a due date. (In this build, sending is a dry run until you connect mail.)
  • Every delivery signed in the product, so the record is complete.

Days 60–90: measure

MeasureFrom
Days from policy received to deliveredStage history on each check
Policies delivered with no unresolved failureControls at delivery
Failures found per policy, by categoryProofing results
Carrier days to correct; overdue requestsCorrections
Undelivered 30+ days after inceptionThe late-delivery control
Agreement with each agentThe autonomy page
Labels taught; labels still unmappedThe taxonomy

Write the definitions down before day 60, so the numbers mean the same thing in month four.

AI & Agents · Activity
Agent activity on the illustrative book
Agent activity: every agent step on the book, what it prepared and what is waiting on someone.
AI & Agents · Autonomy
The autonomy page with every action capped at ask first
Autonomy: the tenant ceiling at L2, agreement over the answers so far, and each agent’s level per action with L3 and L4 locked.

For the account team

One month of renewals is enough to feel the difference and teach most labels.

For broker IT

Every measure above comes from the product’s own tables; no separate tracking.

For practice leadership

Decide the next team on evidence, not on a demo.

Questions we get asked

What does a pilot need from IT?

A database, a model account if you want the agents’ reasoning on, and sign-in. Documents can be uploaded by hand.

Keep reading