How to Write the Rate-Increase Email Without Making It Worse

Every agent has a version of this in their drafts folder. The renewal came back higher, the client has not seen it yet, and the message that explains it has been sitting half-written since Tuesday.

The delay is the part worth looking at. A client who learns about an increase from their agent is having a conversation. A client who finds it on the bill is starting a different one.

So the question is not whether to send it. It is what goes in it.

The reason you give is the whole message

When one of these emails goes wrong, it is usually a single line that did it, and usually the explanation. Something like “rates are going up across the industry” or “this reflects current market conditions.”

Both sound reasonable. Neither is something you can support unless it comes from a documented source your carrier or agency has authorized you to use, and both are the kind of statement that gets read back to you later.

Here is the discipline that fixes it: you only explain what you were authorized to explain. If you have an approved reason, from the notice, from carrier talking points, from an agency bulletin, you use it and attribute it. If you do not, the message says the premium changed, says you are looking at options, and says nothing about why.

One more thing this message is not. A service email does not replace, modify or satisfy any carrier notice, prescribed wording or state-required disclosure. Insurance regulation is state-based and requirements differ by jurisdiction, product and transaction, so what your carrier and your state require sits on top of anything here.

That feels thin the first time you write it. It is also the version you can stand behind if the message is ever read back to you.

The shape that works

Four moves, in this order.

Acknowledge it before you explain it. Nobody enjoys a rate increase, including you. One sentence that says so buys you the next paragraph.

State the change plainly. The amount, the term, the effective date. No softening language around the number, because softening reads as hiding.

Give the approved reason, or give none. “Per the carrier’s notice, this reflects [reason]” or nothing at all. Never a reason you inferred from the file.

Move to a next step. Not a promise about the outcome, a specific offer: a call this week, a review of the coverage, a look at what options exist. The client needs to know something happens next.

That is the entire message. Under two hundred words, most of the time.

What that looks like assembled

Note what it does not do. It makes no claim about the market. It promises effort, not an outcome. And the reason in the third paragraph is quoted, not constructed.

If you were not given an approved reason, delete that paragraph. Do not replace it with a softer version of one. The message gets shorter and stays true, which is the trade you want.

Three phrases to cut before you send

“Everyone’s rates are going up.” You are making a claim about the market. Unless it came from an authorized source, you cannot support it, and a client who has just been quoted less by a competitor will tell you so.

“I’ll get this fixed for you.” This is a promise about something you do not control. It also sets up a second conversation where you have to explain that you could not fix it. “I will look at every option we have” carries the same warmth and commits to effort rather than outcome.

“This is probably because of [reason].” The word “probably” does not save it. If the reason is not from an authorized source, it is not yours to offer. This is a common way a helpful agent creates a problem.

Where AI helps, and where it does not

A lot of agents have started drafting these with ChatGPT or Claude, and the results are usually one of two things. Either a bland paragraph that could belong to any agency, or something worse: a fluent, confident explanation of why the rate went up that the AI simply made up.

The second one is the real risk. The draft reads well. It sounds like you. And it contains a reason nobody at the carrier ever stated.

The fix is structural rather than a matter of better wording. You do not ask the AI to explain the increase. You paste the approved explanation into the prompt and ask it to package what you gave it. Something like:

Leave that bracket empty and the model is far more likely to fill the gap itself, because filling gaps is what it does. Fill the bracket and it has less room to speculate. It does not have none. A model can still alter, omit, overgeneralize or contradict material you supplied, so the completed draft still gets read line by line against the approved source.

The same input-first approach helps with a lot of client writing. It does not replace approved forms, carrier procedures or line-specific requirements, and some communications need exact language, specific disclosures or handling by someone with particular authority.

The part that is not optional

Whatever you use to draft it, you read it before it goes. Every factual and coverage-related statement, every figure, every disclosure your agency requires. That review is not a formality bolted onto the end of an AI workflow. In this business it is the job, and the tool exists to give you more time to do it rather than less.

Draft, verify, personalize, approve, send. The middle three are where the agent still lives.

The Claude Workflow System for Insurance Agents is a 32-page system covering 66 of these conversations, from the first lead reply through renewals, claims support, nonrenewals and reviews. Every workflow is built the same way this one is: the facts come from you, in brackets, so the model has less room to invent them.