Regulatory Capability Is Starting to Move at Software Cadence
Key Takeaways
State insurance regulation remains decentralized by design. Increasingly, the infrastructure regulators use to collect, review, and analyze information is not.
That convergence is not new. The National Association of Insurance Commissioners (NAIC) has spent decades building common filing systems, handbooks, examination procedures, and regulatory standards. What is changing is the mechanism through which common capabilities can spread.
The modernized System for Electronic Rate and Form Filing (SERFF) can already use artificial intelligence to extract and classify information from filings. A separate NAIC application, CLARA, can compare policy language against regulator-defined rules. A common AI evaluation instrument is being tested across twelve states and revised as regulators use it.
None of that eliminates state authority. It does create a new possibility: regulatory capability can begin moving at software cadence even when regulatory authority does not.
Just a Tile
Somewhere inside the NAIC login environment is an instruction that sounds more like a help-desk note than a sign of regulatory change:
“Select the Appian tile on the next screen.”
That’s it. Just a tile.
Behind it is the modernization of SERFF, the shared NAIC platform insurers use to submit rate and form filings to state insurance regulators.
The system itself isn’t new. Texas and Maine can apply different statutes, different regulatory philosophies, and different priorities to filings that arrive through the same platform. Shared plumbing does not mean shared regulation.
But something interesting happens when the plumbing starts reading what flows through it.
That has already begun.
Modernized SERFF uses AI-assisted extraction to identify form numbers and automatically classify forms. The broader roadmap goes further, including filing summaries and eventually helping reviewers compare filings with state-specific compliance requirements.
The shared infrastructure isn’t just moving documents anymore. It is beginning to help regulators read them.
Shared Infrastructure Isn’t New
None of this starts with AI.
The NAIC has been helping states converge around common regulatory practices for decades.
Common handbooks already guide financial examinations, market conduct, and product filing review. Accreditation creates common baseline expectations for financial regulation. SERFF itself has standardized much of the mechanics of rate and form filing without making Texas regulate like Maine.
There is even a much stronger example. Through the Interstate Insurance Compact, 48 member states have delegated a limited product-review function for certain life, annuity, disability income, and long-term care products to a common interstate body.
So regulatory convergence is not new. And shared infrastructure certainly isn’t new.
The question is whether the way that convergence travels is beginning to change.
SERFF provides some scale for why that matters. At Appian World in 2024, NAIC Senior Manager Bridget Kieras described a system containing roughly 7.5 million filings and 48 million attachments, with about 18,000 daily logins and more than 20,000 active users.
Those are 2024 numbers, not current counts. Their value here is scale.
For years, that shared infrastructure primarily moved information between insurers and regulators.
Now the infrastructure is gaining analytical capability.
From Filing System to Analytical System
The NAIC began planning the current SERFF modernization several years ago. A 2022 planning document specifically identified “text processing using artificial intelligence” as one of the capabilities being pursued.
The first modernized SERFF release went live for the Insurance Compact in March 2025.
The interesting part isn’t simply that NAIC replaced old software with newer software. Every organization eventually does that.
It’s how the new system can evolve.
According to the NAIC’s 2025 Annual Report, the modernized platform went through more than 40 releases during its first 115 days. The NAIC describes the process in terms of rapid feedback and continuous improvement.
That does not mean 40 major new regulatory capabilities appeared in four months. A software release can contain anything from a small fix to a meaningful feature.
But compare the development cycle with the traditional regulatory cycle.
A handbook gets revised. A model bulletin gets adopted. States consider it. Some adopt it quickly, others slowly, and others take their own approach.
The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, a model regulatory framework states can choose to adopt, in December 2023. As of August 6, 2026, 25 jurisdictions had adopted it. California, Colorado, New York, and Texas were separately identified by the NAIC as having their own insurance-specific AI rules or guidance.
There is nothing wrong with that pace. State-by-state adoption is the system working as designed.
Software works differently.
A capability can be developed, tested, revised, and made available without waiting for fifty separate legislative or administrative processes.
Available, however, does not mean adopted.
And the SERFF rollout itself demonstrates the difference.
Forty Releases and Ten States
The original modernization plans anticipated phased deployment through 2026.
By July 2026, the NAIC was still working with ten early-adopter states. The goal was to have half of them in production by year-end, with the rest of the early adopters and a second group of states moving during the first half of 2027. One early-adopter state was expected to go live this month.
The reasons are familiar to anyone who has ever tried to roll out a new system across multiple offices: data migration, training, staff readiness, workflow changes.
Hold those two facts together.
More than 40 software releases in 115 days.
Ten early-adopter states moving on individual timelines stretching into 2027.
There is no contradiction there.
Software can accelerate the creation and availability of regulatory capability. It cannot make an insurance department ready to use that capability. It doesn’t migrate the data, train the staff, change the workflow, or decide how the department wants the tool configured.
We’ve seen the same distinction elsewhere in insurance AI.
Access isn’t adoption.
What makes the regulatory story more interesting is what can happen while that slower adoption process continues.
The Regulator’s New Toolbox
SERFF is only one piece.
The NAIC is simultaneously experimenting with different ways of creating common regulatory capability without requiring common regulatory authority.
One is the AI Risk Evaluation Supplement.
The Supplement gives regulators a structured way to examine how insurers are using and governing AI. Twelve states are participating in the current pilot, using the instrument in areas including market conduct examinations, financial analysis, and financial examinations.
The details of the questionnaire are less interesting here than the development process.
The instrument began as a draft in 2025. It has since moved through several numbered versions. Version 4.0 went into the twelve-state pilot. Feedback from those regulators helped produce Version 5.0, which was released for public comment at the end of August.
Participating states can modify the instrument for their own needs. They also coordinate, share experience, and suggest improvements that can flow back into the next common version.
That’s a very different feedback loop from simply publishing a document and waiting to see which states adopt it.
Then there is CLARA, the Compliance Language Assistant for Regulatory Analysis.
CLARA is a stand-alone NAIC application for compliance review. A regulator can define a rule, provide examples of compliant and non-compliant language, and use the system to evaluate filed forms against that rule. The reviewer receives a result and an explanation. The reviewer still owns the decision.
As of June, Arizona, Connecticut, Vermont, and the Insurance Compact were piloting CLARA, with Arkansas and the District of Columbia onboarding.
The more interesting part is where the NAIC says it wants to take it.
CLARA is currently separate from SERFF, which means a state does not have to complete its SERFF migration to use it. Future plans include tighter integration into filing review and the ability for states to share rules with one another.
That last part deserves attention.
A handbook lets regulators share guidance.
A system like CLARA could eventually let them share an encoded check.
Those are not the same thing.
An encoded rule isn’t law. Another state doesn’t have to use it. A machine result doesn’t replace the regulator’s judgment.
But the mechanism by which regulatory practice can travel has changed.
What Actually Moves Faster?
This is where the distinction matters.
Regulatory convergence through the NAIC is old.
What’s changing is the transmission mechanism.
The traditional model still exists. NAIC develops a model law, bulletin, handbook, or standard. States evaluate it through their own legal and administrative processes. Twenty-five jurisdictions adopting the AI Model Bulletin over roughly two and a half years is one contemporary example.
Now place three newer mechanisms beside it.
A shared platform can receive repeated software releases.
A supervisory instrument can be tested by twelve states, modified from their experience, and issued as another version.
A regulatory application can let states define rules in software and, eventually, share those rules with one another.
None of these mechanisms changes state law.
None forces Texas to regulate like Maine.
None eliminates the commissioner, the legislature, administrative procedure, or the regulator sitting at the desk.
But they can change what is available to that regulator much faster than the law itself changes.
Regulatory capability starts moving at software cadence instead of legislative cadence.
That statement needs to be read narrowly.
The capability can move at software cadence. The regulator may not.
SERFF’s own migration schedule proves it.
So What Does This Mean for Insurers?
For carriers, the immediate implication isn’t that fifty regulators are about to become one regulator.
They aren’t.
It’s that the analytical starting point may become increasingly common even while the regulatory endpoint remains local.
Imagine two departments reviewing the same type of filing today. Different statutes and filing requirements may apply. Their reviewers may bring different experience, priorities, checklists, and institutional knowledge to the same document.
Now give both departments access to the same document extraction, the same analytical tools, a common evaluation instrument, and potentially a library of shareable compliance checks.
They can still reach different conclusions.
In fact, common tools may make those differences more visible.
If two regulators begin with the same analytical substrate and still reach different answers, it becomes easier to see where the real difference lies: statute, policy, interpretation, risk tolerance, enforcement priority, or judgment.
That may be one of the more interesting consequences of convergence.
Uniform tools don’t necessarily produce uniform regulation.
They can expose where regulation actually differs.
For insurers operating nationally, that matters. The industry has spent decades dealing with fifty-state variation by building compliance teams, filing processes, state-specific forms, and institutional knowledge around those differences.
AI-assisted regulatory infrastructure probably won’t erase that complexity.
But it may begin moving some of it.
What Stays Local
The boundary is becoming easier to see.
The filing system can be common. The extraction can be common. The evaluation questions can be common. The compliance tools can be common. Eventually, some encoded rules may even be shared.
The authority remains local.
States still decide what their laws mean. They decide which filings deserve scrutiny, what constitutes a material issue, how aggressively to enforce a requirement, and whether to use the available tools at all.
The current AI Risk Evaluation Supplement pilot makes that explicit. States are working from a common instrument while retaining the ability to modify it for their own needs.
CLARA works the same way. The regulator defines the rule and owns the decision.
Even SERFF’s rollout is being handled state by state.
There is precedent for going further. The Insurance Compact shows that states can voluntarily delegate a defined regulatory function when they choose to.
That isn’t what is happening here.
This is convergence without consolidation.
And the difference matters.
What to Watch Next
The nice thing about looking at this now is that we don’t have to wait years to find out whether the pattern continues.
The first early-adopter state. In July, NAIC expected one of the ten early-adopter states to move into production on modernized SERFF during September. If that happens, it will be the next tangible test of how quickly shared capability can move from the center into an individual department. If it slips, that tells us something too.
September 29. Comments close on Version 5.0 of the AI Risk Evaluation Supplement. The question is whether another round of feedback produces meaningful changes before the instrument moves forward.
October 1 and 2. The NAIC Insurance Summit includes hands-on CLARA workshops for regulators involved in rate and form review. That is a small event with a potentially larger significance: regulators learning how to define rules, supply examples, and test filings against them.
October 8. The NAIC plans another public discussion of the Supplement.
November. The updated Supplement is expected to be considered for adoption at the NAIC Fall National Meeting. Considered, not necessarily adopted.
Beyond those dates, I would watch two things.
First, whether CLARA’s planned rule-sharing capability becomes real. Shared software is one thing. Shared encoded regulatory checks are another.
Second, watch for the first credible measurement of results. Review time. Backlogs. Consistency. Staffing leverage. Something a regulator is willing to quantify.
Once one department can point to a measurable advantage from these tools, other departments won’t need legislation to notice.
Fifty Regulators, One Platform?
There isn’t one regulator. There isn’t even one AI platform.
SERFF, CLARA, the AI Risk Evaluation Supplement, state law, local procedures, and individual regulatory judgment remain different things.
But increasingly, they sit on top of shared infrastructure and common capabilities.
For most of the history of state-based insurance regulation, convergence meant agreeing on a model, publishing a handbook, adopting a standard, or building a common filing process.
Those mechanisms still matter.
Now there is another one.
Software.
It doesn’t eliminate fifty-state regulation. It doesn’t even guarantee fifty-state adoption.
It simply changes how quickly a regulator in one state can gain access to capabilities developed somewhere else.
And that leaves a harder question than whether AI will make insurance regulation more uniform.
When everyone starts with increasingly similar analytical tools, what remains truly local about insurance regulation?
Sources
National Association of Insurance Commissioners, SERFF Modernization
National Association of Insurance Commissioners, SERFF Modernization RFP #2071 (November 2020)
National Association of Insurance Commissioners, SERFF Modernization Transition Stages Fiscal Impact Statement (January 2022)
National Association of Insurance Commissioners, 2025 Annual Report
National Association of Insurance Commissioners, 2026 Approved Budget
National Association of Insurance Commissioners, 2026 Strategic Priorities
National Association of Insurance Commissioners, 2026 Summer National Meeting Keynote, President Scott White (August 12, 2026)
National Association of Insurance Commissioners, Speed to Market Working Group Materials (July 13, 2026)
National Association of Insurance Commissioners, Big Data and Artificial Intelligence Working Group
National Association of Insurance Commissioners, AI Systems Evaluation Tool Pilot Project Summary
National Association of Insurance Commissioners, AI Risk Evaluation Supplement, Version 5.0 Summary of Changes
National Association of Insurance Commissioners, AI Model Bulletin State Adoption Map (August 6, 2026)
National Association of Insurance Commissioners, Product Filing Review Handbook
National Association of Insurance Commissioners, Accreditation
Interstate Insurance Product Regulation Commission, About the Insurance Compact
National Association of Insurance Commissioners, Insurance Summit 2026 Agenda
Appian World 2024, NAIC & PwC on Modernizing the Systems for Electronic Rates & Forms Filing, Bridget Kieras and Jake Sloan.
AI Disclaimer: This content was created with assistance from artificial intelligence technology. While content is based on factual information from the source material, readers should verify all details directly with the respective sources before making business decisions.
