Free Governance Framework · Version 1.0
The AI Safety & Use Standard for Insurance Agencies
Your producers are already using AI. Most agencies have nothing in writing about how.
This is a free, adoptable written standard that closes that gap.
An AI use standard is a written agency policy that defines which AI tasks are allowed, which require licensed producer review, and which are prohibited outright, along with the rules for client data, tool approval, training, and record keeping.
Commercial insurance producers are drafting proposals, summarizing coverage, and researching prospects with AI tools right now. In most agencies nobody wrote down what is and is not acceptable. Unwritten rules mean undocumented risk, and undocumented risk lands on the principal.
The ECLIPS AI Safety & Use Standard is published in full on this page. Read it, adapt it, and adopt it. There is no gate and no cost.
1. Purpose and scope
This standard gives an agency written rules for how AI is used in the production of insurance business, so that use is consistent, reviewable, and documented.
It covers any AI tool used in prospecting, servicing, or advising on commercial insurance accounts. It does not cover carrier-side underwriting AI, which the agency does not control.
It applies to commercial insurance producers, account managers, and support staff at adopting agencies.
2. Definitions
| Term | Meaning |
|---|---|
| Public AI tool | A consumer chatbot or assistant where entered data may be retained or used for model training, such as a free public chatbot account. |
| Approved AI tool | A tool on the agency's Approved Tool Register, vetted for data handling and access controls. |
| Client confidential information | Client names tied to account detail, financials, loss runs, policy terms, premiums, claims detail, health information, and anything covered by a client agreement or law. |
| Producer review | A licensed producer reads the full output, corrects it as needed, and accepts professional responsibility for it before it leaves the agency. |
3. The three AI use categories
Every AI use at the agency falls into one of three categories. This is the operational core of the standard.
Approved without extra review
- Researching an industry, exposure class, or prospect using public information
- Drafting first-pass prospecting outreach that contains no client confidential information
- Summarizing your own notes or publicly available material
- Meeting prep, agendas, and internal to-do organization
- Practicing sales conversations and objection handling
Approved with mandatory producer review
- Drafting proposals, coverage summaries, or client-facing explanations
- Comparing policy language or forms
- Drafting responses to client coverage questions
- Any output that will reach a client, carrier, underwriter, or the client file
Nothing in this category leaves the agency without producer review. No exceptions.
Prohibited
- Entering client confidential information into any tool not on the Approved Tool Register
- Sending AI output to a client, carrier, or underwriter without producer review
- Using AI as the deciding authority on coverage adequacy, limits, or recommendations
- Using AI to draft claims coverage positions without producer and, where required, counsel review
- Misrepresenting AI-generated work in any way that violates the agency's disclosure policy
4. Client data rules
Client confidential information may only be entered into tools on the Approved Tool Register. If a producer is unsure whether a tool is approved, the answer is no until the AI Compliance Lead confirms it.
Remove client identifiers before using any AI tool for drafting where the identity is not needed for the task. Default to anonymized detail.
5. Producer review requirements
For every yellow-category use, the reviewing producer confirms four things before output leaves the agency:
- Facts are accurate, including policy language, dates, and figures
- Coverage statements reflect the producer's own professional judgment
- Nothing confidential was exposed in the process
- The final version reads as the agency's work, because it now is
The reviewer is accountable for the output as if they had written it themselves.
6. Documentation and the E&O file
The point of a written standard is that it can be shown. The agency maintains:
- A signed copy of the AI Safety Charter
- The Approved Tool Register
- A training roster showing which producers completed AI-use training and when
- An incident log
Keep these with the agency's other compliance records. If a dispute ever raises the question "did your producers have rules for AI use," this file is the answer.
7. Disclosure
The agency sets its own client disclosure policy and this standard defers to it. At minimum: never state or imply that unreviewed AI output is professional advice, and answer honestly if a client asks how work was produced.
8. Tool approval
The AI Compliance Lead owns the Approved Tool Register. Before approving a tool, they confirm:
- Where entered data goes and whether it is used for training
- Whether the account tier in use has business data protections
- Who at the agency has access
- That the vendor terms have been read
Tools are reviewed on addition and at the annual review. Any producer can request a tool be evaluated.
9. Training
Every producer covered by this standard completes AI-use training before relying on AI in yellow-category work, and refresher training annually.
ECLIPS Academy training for adopting agencies is aligned to this standard, and completion records from ECLIPS can serve as the training roster entries described in section 6.
10. Incidents and annual review
If confidential information enters an unapproved tool, or unreviewed AI output reaches a client or carrier, the producer reports it to the AI Compliance Lead the same day. The lead logs it, addresses it, and notes any rule change it prompts. Reporting early is treated as doing the job right.
The agency reviews this standard, the tool register, and the incident log at least once a year and records the review date.
Train your team on it
A written standard works when producers know how to follow it. ECLIPS trains commercial producers on AI use that stays inside these lines, and a three-seat pilot runs sixty days and credits to your first annual invoice.
Start a 3-seat pilotSee the team trainingCommon questions
Why does an insurance agency need a written AI policy?
Because producers are already using AI whether or not the agency has approved it. Without written rules there is no record of what was permitted, no evidence of training, and no documented review step. In an E&O dispute, the absence of a policy is the exposure.
Is this standard free to use?
Yes. The full text is published on this page and agencies may adapt it. ECLIPS Academy asks only that adopting agencies keep the attribution line and have their own counsel review it before formal adoption.
Does adopting this standard mean banning AI?
No. The standard is built to let producers use AI, with a clear line between work that needs licensed review and work that does not. A blanket ban tends to push AI use underground, which is the situation this standard is designed to replace.
Who should own the standard inside the agency?
A named AI Compliance Lead. That person owns the Approved Tool Register, receives incident reports, and runs the annual review. In smaller agencies this is often the principal or the operations manager.
How does this relate to state insurance regulation?
This standard is an internal governance framework, not a regulatory filing. It does not replace state requirements, carrier guidance, or licensing obligations. Agencies should review it alongside their existing compliance program.
The ECLIPS AI Safety & Use Standard v1.0 is a training and governance framework. It is not legal advice, does not create an attorney-client relationship, and does not replace agency counsel, E&O carrier guidance, or state insurance regulation. Coverage decisions remain the professional judgment of licensed producers. Agencies should have counsel and their E&O carrier review this standard before formal adoption. ECLIPS Academy does not offer state CE credit or designations and makes no income or outcome guarantees. Leo Azar is ECLIPS Academy's AI-built Chief Training Officer, openly disclosed.
ECLIPS Academy · Leo Azar, Chief Training Officer · Version 1.0, August 2026
