In mid-July 2022, Demotech sent private letters to roughly seventeen Florida property carriers warning of impending downgrades from "A" (Exceptional) to "S" (Substantial). Within forty-eight hours those private letters were a public event. The Florida Office of Insurance Regulation wrote to the Treasury Department asking for emergency intervention, and trade press began carrying the names. By the time any of the seventeen carriers had a finalized rating action in front of them, the market had already finished reading.
FedNat was placed into receivership on September 27, 2022. By the end of 2023, Citizens Property Insurance Corporation was carrying more than 1.4 million Florida policies after absorbing displaced business. The single starting event for that cascade was a letter that did not yet announce a downgrade, only that the agency was considering one.
We think that letter is the template for what the first AI-driven rating action against a regional carrier will look like, almost line for line. The next version is likely to come from AM Best rather than Demotech, addressed to a regional commercial carrier and flagging concerns about the Enterprise Risk Management building block as it relates to AI underwriting and claims systems. The mechanics of how that letter moves through the market are already established. The 2022 Florida sequence wrote them.
What the Letter Actually Was
A rating agency letter of the kind Demotech sent in July 2022 functions as an early-warning communication rather than a final action. It is sent privately to the carrier, describing concerns the agency intends to act on absent a satisfactory response. The process is entirely private: no public adjudication, no regulatory filing, no comment period. The carrier receives the letter, has a defined window to respond, and either persuades the agency to hold the rating or watches the rating move at the next published action.
The window is the operative variable. Demotech's window in 2022 was reportedly measured in days, not weeks, and the carriers that received the letters were operating against a clock the agency set, on a timeline most of their reinsurance, lender, and broker relationships were not built to absorb.
Two structural facts shaped the outcome. The first is that the rating agency does not require regulatory authority to act. It is a private commercial body issuing a private commercial opinion. The second is that the opinion sits upstream of institutions that do have hard underwriting requirements, and those institutions move automatically when the rating moves. In Florida in 2022, the institution downstream of Demotech was the federal mortgage system. Fannie Mae's Selling Guide section B7-3-02 accepts Demotech "A" or higher, alongside AM Best "B+" and parallel options from S&P, Moody's, and KBRA. A Demotech "S" does not qualify. The downgrade was a death sentence by structural dependency.
The Leakage Is Part of the Mechanism
A rating agency letter is supposed to be confidential. In practice, a letter to seventeen carriers becomes a public event within forty-eight hours, because brokers, reinsurers, and competing carriers all hear about it through informal channels long before any final action is published. By the time the agency announces, the market has already repriced.
That dynamic is what closed FedNat's placement options. The June 1, 2022 catastrophe excess-of-loss treaty did not come together at the limits and prices the company's exposure required, because the reinsurance market was reading the same Demotech signals the carrier was. We walked the rest of that sequence in the FedNat 89-day death clock piece. The leakage piece deserves naming on its own. The trigger event for an executive team's planning purposes is the market's reading of the letter, which happens days before the carrier has finished drafting a response, not the issuance of the letter itself.
For an executive team, this changes the planning horizon. A response strategy that assumes "we will have thirty days from the final rating action to organize a treaty repair" is operating on the wrong clock. The repair work has to be substantially in place when the early-warning letter arrives, because by the time the letter is final, the relationships that would have supported the repair are already forming a different opinion of the carrier.
The AI Letter, In Advance
Picture the equivalent communication arriving at a regional commercial carrier in 2027. The sender is AM Best rather than Demotech, and the subject has shifted from reserve adequacy to model risk management. The relevant section of Best's Credit Rating Methodology being cited is the Enterprise Risk Management building block, the fourth of the four BCRM components, which sits alongside Balance Sheet Strength, Operating Performance, and Business Profile.
The substance is something close to this. We have concerns about the carrier's model risk management framework as it relates to its AI-driven underwriting and claims systems. Specifically, we have not been able to identify governance documentation supporting the carrier's stated cadence of model retraining, escalation paths when drift is detected, or treaty disclosure practices around material changes in rating philosophy. We are considering a Comprehensive Adjustment to the ERM building block that would result in a one-notch reduction to the financial strength rating. We invite the carrier's response within fourteen days.
That letter is the same instrument Demotech used in 2022, retargeted at AI governance instead of reserve adequacy, and it does not arrive labeled as a downgrade. The Comprehensive Adjustment behind it is well-established methodology, not exotic invention. We covered it separately in our analysis of the AM Best Comprehensive Adjustment as the hidden AI downgrade lever. The instrument is publicly documented in BCRM, the trigger is the building block assessment, and the carrier finds out by mail.
The downstream consequences track 2022 with one variation. The institution downstream of an AM Best rating is rarely the federal mortgage system, because most commercial lines are not mortgage-collateralized. The forcing function is the reinsurance treaty's rating-trigger clause, with eligibility lists from producing brokers and large commercial counterparties as a secondary channel. We mapped the broader treaty exposure in our analysis of why the reinsurance treaty is where AI risk becomes existential. When the rating moves below the contractual floor, the treaty's right to non-renew, cancel, or reprice activates, and the cedent does not control the timing.
The leakage dynamic mirrors 2022 closely. By the time the Comprehensive Adjustment is final, the cedent's reinsurance broker has already had three quiet conversations with the lead market about whether the renewal is going to be a problem, and the market has already started forming its own opinion of the carrier's governance in the absence of cedent-led disclosure.
Three Things the 2022 Letter Should Have Taught Carriers
The 2022 letter recipients had useful, if uncomfortable, lessons available for any carrier facing a future rating action driven by governance posture. Three are worth carrying into the AI conversation.
Waiting for the final action is too late. Carriers that received the July 2022 letters and treated them as a notification rather than a starting gun lost their renewal options inside of weeks. Reinsurance underwriting decisions that normally take ninety days can compress to a single email exchange under rating pressure. The repair work has to be staged before the letter arrives, which means the carrier needs to assume an early-warning communication is one possible quarter-end outcome and have the response materials current.
A second lesson is harder to swallow: the rating agency reads governance more strictly than the regulator does. Where state insurance departments operate on filing deadlines, statutory standards, and a defined process for engagement, a rating analyst operates on professional judgment, with discretion to weigh the absence of documentation against the carrier directly. A carrier that has cleared a recent state market conduct exam may still be exposed to a rating action, because the analyst's standard for "evidence of effective governance" sits well above the bulletin's standard for "satisfactory compliance." The two reading frames produce different conclusions from the same set of facts.
The third lesson is the most operational. The response template has to exist before it is needed, and Demotech's 2022 letters caught most recipients without a pre-drafted reply. The carriers that fared best had something close to a runbook: a designated executive owner for rating agency communications, a current set of governance artifacts that could be attached, a legal review path, and a board notification protocol. The ones that fared worst spent the first week of the response window deciding who was supposed to draft the reply.
For an AI-era equivalent, the three lessons translate cleanly. A regional carrier running a material AI system without governance disclosure should assume an AM Best ERM concern letter is a possible outcome of any annual review cycle, maintain a model change log written in language a rating analyst can read through every retraining cycle, and keep a pre-drafted response runbook ready to move from receipt to delivered reply inside the agency's stated window. None of that work can be done from scratch under pressure.
The CEO Question for the Next Board Agenda
The single useful question belongs on the next board agenda for any regional or specialty carrier with a meaningful AI footprint. What is the equivalent letter we would receive if AM Best decided our AI ERM was Weak, and what is in our drawer ready to send back?
If the answer is some combination of a "responsible AI" slide deck, a vendor risk assessment, and a verbal commitment that the data science team has things under control, the carrier sits roughly where the seventeen Demotech recipients sat on July 18, 2022. The materials in the drawer were not built for the question the agency was about to ask, and the window between letter and final action was too short to build them.
The carriers that produced credible answers in 2022 had been doing the documentary work for years before the letter arrived. Those that survived the structural transitions did so through some combination of state-engineered policy migration into Citizens, holding company support arrangements, and pre-existing reinsurance relationships that could be repositioned. Most of those mitigants are unavailable to a regional carrier outside Florida facing an AI-driven ERM action. The runway between letter and structural consequence is the runway the carrier built in advance.
The 2022 Demotech letter is the cleanest precedent in the recent record for what a private rating agency communication does to a public market. It is worth treating as a template, not a Florida story. The next version of that letter is going to be about AI governance, and the carriers that have already drafted their reply will be reading it from a different chair than the carriers that haven't.
