Market selector · Trace
One market selector run's trace on illustrative data: status, level, model, tokens, cost and time, and the three steps
A run, traced. One market selector run, traced (illustrative data): what the engines said, the model step with its model, tokens, cost and seconds, and what was prepared for a person.

The fastest way to fail with a placement workbench is to switch everything on for every team at once. The way that works is smaller: one team, one renewal cycle, agents at draft, and the product’s own record telling you what to switch on next.

The plan in one paragraph

  • Days 1–30: one team’s renewals inside 120 days; readiness, the data-gap chaser and the controls only.
  • Days 31–60: submissions and market selection on the same book; the appetite table maintained by the team.
  • Days 61–90: comparison, proposal and compliance; review the record; decide on the second team.
timeline title Ninety days, one team Days 1 to 30 : Calendar and readiness : Data-gap chaser : Placement controls on, thresholds as shipped Days 31 to 60 : Submission builder : Appetite table owned by the team : Market selector and outreach as dry runs Days 61 to 90 : Quote comparison : Proposal and compliance : Review agreement and controls, plan team two

Days 1–30: see the book

Load the team’s renewals inside 120 days with their exposures, SOVs, loss runs and expiring programs. Turn on the readiness agent, the data-gap chaser and the ten controls as shipped. The first value is visibility: the renewals page sorted by days to inception, readiness on every placement, and the controls’ warnings on Home. Leave every agent at L1.

Days 31–60: build and market

Turn on the submission builder and the market selector. The appetite table is the team’s knowledge — have the placement brokers own it, carrier by carrier. Keep outreach on the dry-run provider until the team trusts the drafts; log approaches and responses on the market rows either way.

Days 61–90: compare, propose, review

Turn on the comparison, the proposal writer and the compliance agent. At day 90, read the record:

QuestionWhere the answer is
Are placements on the calendar earlier?Readiness at the 90- and 60-day marks; placements behind the calendar
Are the agents right?Agreement and overrides per agent and action on the autonomy page
Which controls do real work?30-day hits per rule in the Studio; how often a refer changed the outcome
Where do brokers disagree with the recommendation?Selections that differ from the recommended quote, with reasons
What does it cost?Model calls, tokens and cost per agent run
AI & Agents · Autonomy
The autonomy page: tenant ceiling L2 ask first with L3 and L4 locked, and the data-gap chaser's level per action
Autonomy: the tenant ceiling at L2 (ask first) with L3 and L4 locked, and an agent’s level per action and line — here the data-gap chaser, at L1 Draft on every line.

Governance from day one

  • Model risk reviews the grounding contract (engines compute, models explain, ungrounded numbers rejected), the per-run traces and the versioned prompts.
  • E&O review reads the decision record: controls, prepared decisions, signed answers, the client’s sign-off.
  • Compliance owns the security-floor and surplus lines thresholds in the Studio.
  • IT confirms identity, tenancy, where documents are stored and which model account the agents use — see security & deployment.

What we will not promise

We do not publish customer results, because we will not invent them. The ninety days are for you to measure your own: the product records everything you need to do it.

Keep reading