The Insurance Agent's Guide to AI
AI for Insurance Agents: How Working Agents Actually Use ChatGPT and Claude
AI for insurance agents comes down to one thing: you hand the tool your knowledge of the client and the coverage, and it hands back a working draft in seconds. This guide covers the six insurance workflows I teach first, with exact prompts, and the lines a prudent agent never crosses.
1-on-1 with Pete Enestrom, on your actual work

Written by Pete Enestrom
Yale & Columbia, ex-Microsoft & Intel — 1-on-1 AI coaching for executives

Your coach’s background
ChatGPT and Claude are general-purpose AI assistants, and for an insurance agent the honest description is a fast junior copywriter that lives in a chat window. You give the tool the context about your client, the coverage, and the tone you want, and the tool gives back a working draft in seconds. There is nothing to install and nothing to code. The writing-heavy, repeatable parts of an agent's week—renewal letters, policy explanations, prospecting emails, claims updates—are exactly the work these tools take off your desk.
One clarification up front, because the phrase trips people up. If you searched “ai insurance agent,” you may have seen two very different things under those words: software vendors selling autonomous “AI agents” that quote or answer customers on their own, and working insurance agents learning to use AI tools. This page is the second one—written for human agents and brokers who want to use ChatGPT or Claude well. Nothing here is about replacing the agent; everything here is about giving the agent better drafts and more hours.
This guide is written for the independent agent or broker principal running a small book of business, not for a carrier's innovation team. Every workflow below runs on the free plans, and every prompt is the material I cover in a first session with the insurance professionals I coach.
What can AI actually do for an insurance agent in week one?
In week one, AI gives an insurance agent first drafts of the words the job demands all week long. The renewal letter for the commercial account whose premium jumped, the plain-English email explaining what a deductible actually means, the follow-up after a claims call, the first-touch note to a prospect—each of these starts as a blank page, and ChatGPT or Claude removes the blank page. The draft is never the finished product, because the agent still supplies what only the agent has: knowledge of the actual client, the actual policy language, and judgment about what can and cannot be said. What the tool hands back is time: the writing that used to eat the evening is done before dinner, and the saved hours go back into clients and prospecting.
Week one is also the right time to learn what these AI tools for insurance agents are not. ChatGPT and Claude are writing and thinking partners, not coverage authorities. Neither tool has your carriers' current forms, neither tool knows a client's declarations page unless you paste the relevant lines, and neither tool can tell you whether a specific policy covers a specific loss. An agent who asks an AI tool for a coverage opinion will get a confident, well-written answer that may be wrong—and in this business, a wrong coverage statement is an errors-and-omissions problem, not a typo. The agents who get the most from AI use it the way they would use a bright new assistant: drafts, structure, and wording come from the assistant; coverage facts, figures, and final judgment stay with the licensed professional.
Six insurance prompts worth saving
These are the six prompts I teach first to every agent and broker I coach. Copy them verbatim, replace the brackets with your own material, and adjust the last line to taste. Notice how often these prompts keep client names and figures out—the checklist later in this guide shows where that line sits.
I run an independent insurance agency. A commercial client's [policy type—e.g., general liability] policy renews on [date], and the premium is increasing mainly because [the carrier's stated reason—e.g., higher construction costs in the rating area]. Draft a renewal letter that reaches them well before the renewal date, explains the increase in plain English, notes that I reviewed their options, and invites a call this week. Under 200 words, warm and direct, no insurance jargon. Leave placeholders for the name and figures—I'll add those myself.
Sending the renewal letter early is the retention move; the prompt just makes sure the letter gets written on a Tuesday instead of never.
A client asked whether their [homeowners] policy covers [scenario—e.g., water damage from a failed sump pump]. Explain in plain English how this type of coverage typically works, which exclusions and optional endorsements commonly apply, and give me three questions to check in the client's actual policy documents before I answer them. Do not tell me whether their policy covers this loss—only their policy language can do that.
That last sentence is the whole trick. The AI explains how coverage generally works; the policy documents decide the claim. You stay out of the coverage-opinion business.
I'm an independent insurance agent reaching out to [type of business—e.g., HVAC contractors with 5–20 employees] about a [coverage—e.g., commercial package] review. Draft a short first-touch email: one sentence on who I am, one sentence naming a coverage problem common in their trade [e.g., certificates from uninsured subcontractors], a low-pressure ask for a 15-minute review, and an easy way to say no. Under 120 words, no flattery, no “just checking in.”
The specific trade problem is what makes this a letter from an agent rather than spam. AI tools for insurance agents write the words; your knowledge of the trade supplies the substance.
A client's [auto / commercial property] claim has been in [stage—e.g., adjuster review] for [timeframe]. Draft a status-update email that explains in plain English what this stage means, what happens next, what the client does and does not need to do right now, and when they will hear from me again. Calm and factual, under 150 words. No promises about the outcome or the timeline—I control neither.
Most claim complaints are really silence complaints. A clear status note, sent before the client calls you, is the cheapest service win in the business.
I'm training a new producer at my independent agency on [topic—e.g., how to read a commercial general liability declarations page]. Create a one-page teaching outline: the five things a new producer must be able to find and explain, one realistic client scenario to practice on, and three questions I'll use afterward to check their understanding. Assume no prior insurance experience.
Run this for each onboarding topic and you have a training binder that used to take years to assemble; your corrections to the drafts are where your agency's way of doing things gets baked in.
Below are my own notes from three carrier quotes for the same [restaurant] client: [paste your notes—limits, deductibles, key exclusions, premium, service reputation]. Turn my notes into a one-page comparison I can walk the client through: the differences in coverage, not just price; what each option gives up; and three questions the client should answer before choosing. Use only what is in my notes, and flag anything missing instead of filling it in.
“Use only what is in my notes” is the sentence that keeps the summary honest. The comparison stays yours—the AI just makes it legible for the client.
What AI can't do in an insurance practice
ChatGPT and Claude cannot see your carrier portals, do not know your carriers' current forms or rates, and do not know your state's insurance regulations. Both tools also write with total confidence whether they are right or not, which is the dangerous combination in a licensed profession: an agent who pastes an AI draft into a client letter without checking will eventually send a wrong coverage statement, an invented endorsement name, or an exclusion nobody verified—and the agent, not the tool, owns that error. The working rule I give agents and insurance brokers is simple. Facts go into the prompt from you; facts in the output get verified by you. Every coverage statement in anything a client receives gets checked against the actual policy documents, every time.
One more boundary deserves plain words: nothing an AI tool produces is legal, compliance, or E&O advice, and nothing in this guide is either. When a client asks a genuine coverage question, the answer lives in the policy language and, where needed, with the carrier or a coverage attorney—not in a chat window. Where AI genuinely helps is everything around that judgment: the plain-English explanation of how a coverage type generally works, the letter that delivers your verified answer clearly, and the checklist of what to confirm in the documents. Agents who hold that line get the time savings without the exposure; agents who blur it are how E&O claims get born.
What to keep out of AI tools
One short list to print and keep near your desk. Treat anything you paste as shared with an outside vendor—because that is what it is. The privacy guide and the AI usage policy template on this site go deeper.
- Client names, contact details, and anything that identifies a policyholder
- Policy numbers, claim numbers, and loss details from an open claim
- Health, financial, or driving-record information from applications
- Carrier portal logins, rating-system access, or documents from a shared drive
- Anything your agency's written policy or your carriers' rules have not cleared
The mistakes I see agents make with AI
After coaching insurance professionals through their first months with ChatGPT and Claude, the failure modes are remarkably consistent. None of them are about the technology—they are habits, and every one of them is fixable in a week.
- One-line prompts. “Write a renewal letter” gets generic copy. The client context, the reason for the increase, and the tone are the brief—and the brief is the work.
- Sending the first draft. The edit conversation—“shorter,” “less formal,” “the second paragraph overstates the endorsement”—is where the quality lives.
- Pasting client data. Names, policy numbers, and claim details do not belong in a chat window. Describe the situation, never the identity; the checklist above is the line.
- Trusting its coverage answers. AI explains how coverage generally works; it cannot read your client's policy. Verify every coverage statement against the documents before a client sees it.
- Expecting carrier knowledge. ChatGPT and Claude do not know your carriers' current forms, appetites, or rates—your notes and carrier relationships are the expertise; the tools only shape the words around them.
Where does 1-on-1 coaching fit?
Everything in this guide, an agent can do alone this week, for free. Self-teaching absolutely works, and most agents should spend a month doing exactly that. The ceiling appears around month two: the generic prompts are running, and the real question becomes which of your workflows—your renewal process, your prospecting cadence, the way your producers write to clients—should be rebuilt around these tools, and in what order. That question is specific to your book of business and your carriers, and generic answers stop helping.
That is the point where working 1-on-1 with a human teacher—not an AI coach bot, not another course—earns its keep. In a coaching session we open your actual renewal list and your real pipeline, and we build the prompts and habits around your clients, your carriers, and your agency's written policies—not a generic playbook borrowed from the internet. This guide is the map. Coaching is walking it together, on your book, with your next renewal letter as the homework.
1-on-1 coaching
See it on your actual workload.
Bring the work you did this week. We'll build the workflows on it together, live.

Your coach
Coached by Pete Enestrom
Yale and Columbia grad, former Microsoft and Intel, and a venture-backed exited founder. Pete has spent the last four and a half years going deep on every major AI tool, and he teaches the way operators learn: on your real work, at your pace, with nothing assumed.

Common questions
Straight answers, the way I'd give them across a table.
Keep reading
For Your Role
ChatGPT for Real Estate Agents: A Practical Guide
ChatGPT for real estate agents, in plain English: six copy-paste prompts for listings, follow-ups, and client emails—plus what to never paste into it.
ReadResource
ChatGPT Privacy, in Plain English: What Really Happens to What You Paste
ChatGPT privacy in plain English for business owners: what happens to what you paste, the real risks, the settings to check today, and rules to hand your team.
ReadResource
AI Use Policy: A Practical Template Your Team Will Actually Follow
A plain-English AI usage policy template for business owners: what a policy should cover, ten copy-ready clauses, and how to roll it out in one team meeting.
ReadBeyond coaching
When the workflow needs to be encoded, not just learned — that's what our team at Zaigo builds.