From $3,000
Your renewal questionnaire has a new section. We build the answers.
This is for the medical practice, the accounting or law firm, the financial-services shop — any insured business whose cyber renewal lands in the next twelve months and whose staff already use AI agents for real work. Last year the questionnaire asked whether your people were trained. This year it asks what controls govern your agents. Most firms answer with a paragraph of good intentions. The carrier reads it as "no controls." We build insurability: four artifacts, written the way an underwriter reads them, kept current after the policy binds.
What you get
The deliverables, stated plainly.
Agent inventory
Every agent in use, what systems and data it can reach, what actions it is allowed to take, and the named owner responsible for it.
Access-control map
Least-privilege scopes per agent, the approval gates in front of consequential actions, and the transaction limits above which a human signs.
Evaluation evidence
What was tested, how, the agreement scores and error rates, and the date of the last run.
Incident-response plan
How you detect a wrong action, stop the agent, roll it back, and notify whoever must be told — with owners and time targets.
A one-page summary sheet
Formatted to drop into the renewal application.
Optional monitoring attach
$550–750 per month to keep all four artifacts current as agents, scopes and models change.
How it runs
Week one is discovery. We get system access within 5 days of signing and walk every place an agent acts: email, documents, billing, scheduling, code, the tools your staff adopted without telling anyone. Week two is the inventory and the access map. Week three is evaluation and the incident plan; where you have no evaluation evidence yet, we run a bounded test and record the result honestly. Week four is review against the questionnaire language your carrier or broker actually sent. Drafts come back within 3 days of any question, and we hold a weekly check-in throughout. With the monitoring attach, the artifacts are rechecked monthly and reissued when anything changes.
What it costs
Payment is 65% at signing and 35% on acceptance. The founder confirms the number on the call. The founder spent years on the insurance side of the table, which is why the artifacts are written for the person who prices your risk rather than for the person who built your agents.
Why now
As reported on October 4–5, 2026, a major model provider notified more than 100 organizations of unauthorized agent activity on their accounts. A public breach at a major model hub is the precedent carriers now cite. Litigation over agent actions is live. Cyber-insurance renewal questionnaires now ask what controls govern autonomous agents — inventory, allowed actions, approvals, transaction limits, error rates, evaluation evidence and incident response. Staff training alone no longer answers the question. The firm that can produce these four documents on request will renew on better terms than the firm that cannot, and the gap widens each cycle.
Questions people ask
We only use a couple of AI tools. Is this overkill?
The inventory is usually longer than the owner expects, because staff adopt tools on their own. The pack is sized to what we actually find.
Will this lower our premium?
We do not promise premiums. We promise you can answer the section with evidence instead of a paragraph, which is what decides whether the carrier treats you as controlled or uncontrolled. Your broker will tell you what that is worth.
Can our IT provider do this?
They can and should help. What they usually lack is the underwriting frame: which facts the carrier weighs, in what form. We work alongside your IT people and your broker, not around them.
What does the monitoring attach do?
It re-walks the inventory, rechecks scopes and limits, reruns the evaluation on any model or tool change, and reissues the four artifacts, dated. At the next renewal the pack is current rather than a year old.
Book 15 minutes
The founder confirms the number on the call. Fifteen minutes, no deck, no pitch — we talk about your situation and tell you if there is nothing to build.