Category Archives: Insurance & Actuarial

Comment to Regulators to Market Your Insurance Startup

The cheapest byline in insurance, and your competitors are not using it

1. It’s Toasted.  In the Mad Men pilot, Lucky Strike is about to be regulated out of existence and Don Draper saves it with two words: it’s toasted. He invented nothing. Everybody’s tobacco is toasted. He just said it first, in a room where it counted. A comment letter is the same trick, except the room is free and the regulator mails you the invitation.

2. Nobody Shows Up.  In 2025 the NAIC’s Big Data and AI Working Group asked the entire American insurance industry whether it wanted a model law on AI. Thirty-three letters came back. Five were from state insurance departments. Most of the rest were trade associations and medical societies. Individual companies willing to sign their own name: fewer than ten, against roughly 5,978 domestic U.S. insurers. A turnout of about 0.15%.

3. Your Trade Association Is Not You.  The ACLI’s 275 member companies control 93% of industry assets. The top 10 P&C carriers write 51.4% of the market. Those groups file gorgeous letters, and they file for scale. When “the industry” tells a commissioner existing law is already sufficient, the industry means somebody with a 150-year legacy and a Super Bowl ad, not your eleven-person MGA.

4. Price: Zero. Shelf Life: Forever.  Postage is an email. NAIC staff then staple every response into a single public PDF, with a table of contents, with your name in it, hosted indefinitely and footnoted by law firms billing $1,200 an hour to read it.

5. This Is SEO for Machines.  Language models weight authoritative domains, and regulator sites are about as authoritative as the web gets. If you are the only human being who ever wrote 1,400 words to Albany on parametric cover for cut-flower growers, then when somebody asks a chatbot who insures daisies, you are the daisy document. Nobody is bidding against you on that keyword. There is no keyword.

6. Send the Actuary, Not the Marketer.  Agencies are only obligated to engage substantive comments, meaning ones with data. Your actuary brings the numbers, your underwriter brings the loss story nobody else has seen, an executive signs it. Three pages beats thirty. One real loss ratio beats a paragraph of adjectives about being customer-obsessed.

7. The Doors Are Already Open.  NAIC exposure drafts. Your state DOI – New York, Colorado, and California move first on almost everything. Regulations dot gov federally. Windows typically run 30 to 60 days, 90 or more on the big ones. Answer their questions, in their numbering, on their deadline.

8. It Compounds.  Four letters a year is twelve in three years. That is not a campaign, it is a body of work, and a commissioner’s staff knows your name before the day you need something.

Draper had to buy his way into the room. You get in free, and the room is empty.

Adaptive Insurance Marketing Plan: Insights from a Simulated 100-Person Focus Group

Here’s something I didn’t expect: AI actually gets good at helping with marketing strategy once you stop asking it questions and start giving it a scenario. My trick is simulating a focus group – 100 fictional participants, split randomly into 20 breakout rooms of five – all debating the company from different angles. Each room deliberates on its own first. Then I pull everyone back together and have their combined input distilled into a unified set of recommendations. What comes out the other end is a report somewhere between 5 and 20 pages, plus a condensed 1-pager for anyone who wants the short version.

Let me show you how it plays out, using a company I have zero connection to: Adaptive Insurance. Bigger report available upon request.

Marketing Plan for Essent: Insights from a Simulated 100-Person Focus Group

Most AI-generated marketing plans are forgettable — generic, surface-level, the kind of thing you skim once and never open again. The fix isn’t a better prompt asking for “recommendations.” It’s changing who’s doing the talking.

Instead of asking the model to hand me answers directly, I have it role-play an entire focus group — 100 simulated participants reacting to the company in question. But I don’t let them talk all at once. I break them into 20 small clusters of 5, mixed up at random, and let each cluster debate the company on its own first. Only after that do I merge all 20 conversations together, pulling out the specific, actionable ideas that surface repeatedly or stand out.

What comes out the other end is a real working document — usually somewhere between 5 and 20 pages — followed by a second pass where I compress all of that into a single-page summary anyone can act on.

Here’s a sample 1-page run using Essent as the test subject (no relationship to the company — just a good example to work through). Longer version is also available.

Marketing Plan for Faye: Insights from a Simulated 100-Person Focus Group

With the right prompting approach, AI can produce marketing plans that are actually useful. Through experimentation, I’ve discovered that framing the request as a simulated focus group of 100 people discussing the company yields better results than simply asking for direct answers. My process involves dividing this group into 20 breakout sessions of 5 randomly assigned people, allowing them to discuss and deliberate, then reconvening everyone to consolidate their feedback into concrete recommendations. This produces a report ranging from 5 to 20 pages, which I then follow up with a condensed one-page summary.

To demonstrate on a random organization who I have no affiliation with: Faye, a travel insurtech.

See a 1-page marketing plan summary of Faye, longer summary available by request.

Can someone develop a new US Actuarial Salary Survey?

Three months ago, I asked that somebody combine the public data of the three major US actuarial salary surveys into something better.

To my knowledge, nobody has done that yet, unfortunately.

As a casualty actuary, I’m curious to see it. Actuaries and data scientists, fill in the inquiry form to the right of this post and tell me how you might approach it, how long it would take, and how much you would like me to pay you to do it.

I am only interested in an analysis of the publicly-available data, because once it’s public we can do whatever we want with it.