Every product in the Insight suite has the same autonomy dial: L1 draft, L2 ask first, L3 act and tell, L4 act. What differs is where each product puts the ceiling. In InsightProofing every action is capped at L2, whatever the dial says, because everything an agent could do here ends with something reaching a carrier or a client.

The level for an action

flowchart LR A["Agent's level for
this action and line"] --> MIN{"lowest of"} B["The action's hard cap
(L2 for every action)"] --> MIN C["The tenant ceiling
(PROOF_AGENTS_MAX_LEVEL)"] --> MIN MIN --> L["Effective level"]

The six governed actions: ask for missing documents, confirm label mappings, raise failed checks as corrections, send the correction request, deliver the policy to the client, and the practice leader’s sign-off. Each starts at L1 and can move to L2; L3 and L4 are locked on the autonomy page.

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.

A prepared decision

What an agent prepares is a card: the agent, its level, whether the engines or the model wrote it, a title, the lines that matter, The facts behind it, and three answers — Accept (carry it out as the person), Reject (with a reason), Ask (question the agent from the facts). One live ask per agent and action: a newer one supersedes an older one still open.

stateDiagram-v2 [*] --> Open: agent prepares Open --> Superseded: a newer ask for the same action Open --> Accepted: person accepts, carried out as them Open --> Rejected: person rejects, with a reason Accepted --> [*] Rejected --> [*] Superseded --> [*]
Findings · prepared for you
Decisions prepared by the agents on an illustrative general liability renewal
Prepared for you: a carrier label to confirm, three failed checks to raise with the carrier, and a practice leader sign-off — each with the facts behind it and Accept, Reject or Ask.

Agreement is still measured

Each answer is evidence. The autonomy page shows agreement and overrides over the last 30 days; the rule that recommends a notch up (90% agreement over 50 answers) and down (below 70% over 20) runs as in the rest of the suite, within the cap. In InsightProofing that measures how well the agents prepare; it does not let them send.

For the account team

Nothing leaves without you. The agents get the work to the point where your answer is one click.

For broker IT

The shared AutonomyService with a per-action hard cap of L2 and an environment ceiling.

For practice leadership

Agreement rates per agent tell you which parts of checking the team trusts — and which it doesn’t yet.

Questions we get asked

Could a brokerage raise the cap?

Not from the dial: the cap is in the product. Sending to a carrier or a client stays a person’s decision.

Keep reading

How it works

The decision record is the E&O file

Every control, every prepared decision, every answer — with who and on what.

October 29, 2026 · 9 min read Read article
Use case

A property renewal with a halved flood sublimit

Not a proofing failure — a renewal change the client must hear about in plain words.

October 29, 2026 · 15 min read Read article
Use case

A missing waiver and an exclusion nobody agreed

The binder lists the waiver the client's contracts require. The policy doesn't.

October 29, 2026 · 15 min read Read article