2026 · What pool staff are saying

The AI and website questions pools are wrestling with

In two interactive conference sessions this fall, public entity risk pool staff worked through real-world AI scenarios and audited their own websites. Here are the questions they answered, how they responded and what it means for pools.

Two national pool conferences Fall 2026 Matt Burleson, OMAG
Pool staff are ready.Most are curious or already using AI, and confidence grows fast once they have a simple framework.
Their instincts are right.AI can draft, but a person verifies, a person decides and the employee stays accountable.
The structure is behind.Policy, data rules, member disclosure and pool websites haven't caught up to staff.
How the room responded: Nearly everyonebroad agreement Mosta clear majority Dividedno clear consensus
01 · Readiness

Are pool staff ready for AI?

The common assumption is that pool employees are wary of AI. The room told a different story.

How are you feeling about AI right now?

Response

Mostwere curious or already using AI at work. Only a few said they were mainly worried. Asked again at the end, the "give me all the AI" group had roughly doubled. Among people who answered both times, 7 moved toward embracing AI and none moved away.

What it means

Fear isn't the barrier. Plan for direction and guardrails, not persuasion.

How prepared do you feel to make AI decisions in your organization?

Response

Nearly everyonefelt at least prepared to ask the right questions by the end of the session, and many said they could help set guardrails or lead the conversation at their pool.

What it means

Staff can be partners in building AI policy, not just audiences for it.

Which AI model would you use for a simple, low-stakes writing task?

Rewriting a short announcement for a member training webinar.

Response

Mostchose a fast, lower-cost model instead of the most powerful option.

What it means

Staff instinctively match the tool to the job, which helps keep AI costs in check as subscriptions spread.

02 · Where we agree

AI drafts. People verify and decide.

Whenever a scenario put AI in front of members or put a decision about a person on the line, pool staff chose the option that kept a human in charge.

Does using AI shift accountability away from the employee who used it?

Response

Nearly everyonesaid no. In fact, every single person did. Whoever uses AI owns the result.

What it means

"The AI wrote it" will never be an acceptable answer to a member, a board or a court, and staff already know it.

AI answered a policy question confidently. Was its answer trustworthy?

The policy had two sections that contradicted each other, and AI quoted one without mentioning the conflict.

Response

Mostcaught it and said AI should have flagged the conflict. Almost no one trusted the answer as given.

What it means

Pool people read the fine print for a living. That healthy skepticism is one of our biggest advantages with AI.

Leadership wants AI to write a press release and plans to approve it with a quick read. How should the team handle it?

Response

Nearly everyonewould let AI draft it, but only after verifying every name, number and fact against the original sources.

What it means

Staff see AI as a fast first draft, not a finished product.

A storm hits member cities overnight and leadership wants a message out in ten minutes. What's the best use of AI?

Response

Nearly everyonechose a quick AI draft with confirmed claim-reporting steps, reviewed before sending. Almost no one would let AI answer members' coverage questions on its own.

What it means

Speed is welcome in a crisis, as long as a person checks what goes to members.

Should AI review an HR complaint, decide whether policy was violated and recommend discipline?

Response

Nearly everyonesaid no. Most would use AI to organize the facts and relevant policy for a human investigator. Some would keep AI out of it entirely.

What it means

Decisions about people stay with people.

A supervisor pastes an employee's medical records into a personal AI account. Green, yellow or red?

Response

Nearly everyonesaid red.

What it means

Some lines are bright, and staff already see them.

03 · Gray zones

Where pool staff need clearer rules

The questions that divided the room are where well-meaning staff at the same pool will make different calls, and where written guidance does the most good.

Can AI help write interview questions from a publicly posted job description?

Response

DividedAbout half treated it as routine work and nearly half wanted to slow down, even though the job posting is public and nothing sensitive is involved.

What it means

We tend to judge risk by the department asking ("HR") instead of the data involved. A simple classification framework fixes that.

Can an approved enterprise AI tool look for trends across thousands of anonymized incident records?

Response

Mostsaid yes, with safeguards and a human reviewing the findings. Some treated it as routine work.

What it means

Big-data projects can be a great use of AI, but they deserve a defined review step.

04 · AI policy

The building blocks of a pool AI policy

When pool staff work through the core policy questions together, a fairly clear consensus emerges, with two questions every pool should settle on purpose.

Should employees be limited to organization-approved AI tools?

Response

Mostsaid yes. Others would allow any tool for brainstorming but keep work data in approved tools. Almost no one wanted no restrictions.

What it means

Give staff an approved tool. Without one, they'll use whatever they have.

Which AI outputs need a human review before they're used?

Response

Mostsaid everything, every time. A sizable group would require review only for higher-risk work involving data, HR, legal matters or major decisions.

What it means

Both are reasonable. Pick one and write it down.

Where's the line for sensitive organizational information?

Response

Mostsaid it should never leave an approved, organization-controlled environment. A meaningful minority felt anonymized or aggregated data is fine anywhere.

What it means

Define "anonymized" for your pool. Claims and loss data from small members can identify people even without names.

Should members be told when AI helped create something they receive from the pool?

Response

Mostsaid no, as long as a person reviewed it. Some would disclose only for significant content like coverage letters, claims communications or legal guidance. Few would always disclose.

What it means

That may be the right answer, but few pools have decided it deliberately. In a business built on trust, make this call on purpose.

05 · What staff ask

The questions staff bring to work

When attendees shared the real AI questions they face, the same themes came up. Most touch personal information, employment decisions, claims or rating, which are the areas where pools carry the most risk.

Personal AI accountsIs it OK to use my own AI account for work tasks?
Data with names in itCan I upload loss data that includes employee names into a tool that's supposed to be private?
Anonymizing claimsHow do I strip identifying details from a long, complicated claim before AI summarizes it?
HR documentsShould AI help write disciplinary or probation memos?
Rating and underwritingShould AI play a role in decisions like experience modifiers?
Everyday productivityCan AI help plan a year of newsletter topics or similar routine work?
Shadow AI is already here. Several of these questions suggest staff are experimenting before their organization has given them an approved tool or a clear answer. That's a signal to build the official path, not a discipline problem.
06 · Pool websites

Does your website get members to the next step?

In a session on improving pool websites, attendees audited pages from their own pool's site using four FIND questions. The results were consistent: 8 in 10 pages fell short on at least one, and many fell short on most.

Directed: Does the page point members to a clear next step?

Response

Mostpages did not. This was the most common gap: no button, phone number or "here's what to do now."

What it means

Every member-facing page should end with an obvious action.

Findable: Can members find this page quickly?

Response

DividedRoughly half the pages were hard to find, often buried under confusing navigation or too many competing links.

What it means

Organize around what members are trying to do, not around the org chart.

Intuitive: Is the page easy to use once members get there?

Response

DividedMany pages hid important information or relied on internal language and acronyms that members don't use.

What it means

Write for the member who's stressed and in a hurry, like someone reporting a claim.

Necessary: Is the content still current and useful?

Response

DividedMany pages were flagged for outdated content, sometimes years old.

What it means

Retire what's stale. Old pages crowd out the ones members need.

Why this matters for AI. Members increasingly ask AI tools their questions instead of browsing pool websites, and those tools read our pages to answer. Clear, current pages with an obvious next step get summarized correctly. Cluttered ones get misread or skipped.

About these results

These summaries come from anonymous, live audience participation in two interactive sessions for public entity risk pool staff at national conferences in fall 2026: one on leveraging AI in pool work and one on using AI to improve pool websites. No names, emails or organizations were collected.

Most responses are described in general terms. Where a figure is shown, it counts one response per participant. Attendees chose to be in the room, so read this as a snapshot of engaged pool professionals, not a statistical survey of the industry.