Selling insurance agency data for AI

An insurance agency runs on a handful of workflows repeated thousands of times: quoting, binding, endorsements, renewals, certificates, claims advocacy and service requests. Each one involves gathering information, applying underwriting appetite and coverage judgment, and coordinating between a client and one or more carriers. That's exactly the kind of multi-party, rules-heavy work AI labs want examples of.

Agencies also hold sensitive personal and financial information and operate under carrier contracts and state insurance regulation. This guide explains where the realistic opportunity is and what should stay out.

This guide is general information, not legal advice. Insurance privacy and data security rules vary by state. Talk to your counsel before licensing any agency data.

The rules that shape scope

GLBA and state insurance privacy rules

The Gramm-Leach-Bliley Act applies to insurance agencies as financial institutions. For insurance, state insurance regulators generally enforce the privacy requirements, often through state laws and regulations based on NAIC model acts. These limit how nonpublic personal information about consumers can be disclosed to third parties and require privacy notices. Your notice and your state's rules should be part of the review.

State insurance data security laws

Many states have adopted insurance data security laws based on the NAIC Insurance Data Security Model Law, which require licensees to maintain an information security program and oversee third-party service providers. A data licensing project should fit inside that program.

Carrier and agency agreements

Agency agreements with carriers often address ownership of expirations, confidentiality of carrier information, and use of carrier systems and data. Commission schedules, underwriting guidelines and rating information shared by carriers are commonly confidential. Read your agreements before including anything carrier-related.

Health and benefits information

If you place employee benefits, you may handle health information, and you may be a business associate of group health plans under HIPAA. Treat benefits files as a separate, high-risk category. See our healthcare data guide.

What is a realistic starting point

For most agencies, the best opportunity is workflow and process data with consumer and client identities removed:

• Service workflows: how an endorsement request, certificate request, or cancellation moves through the agency, from inbound request to carrier submission to confirmation
• Renewal and remarketing processes: timelines, checklists, decisions about when to remarket, and how options are presented
• Commercial submission preparation: how account managers gather exposures, complete applications and package submissions (with insured details removed)
• Internal procedures: agency operations manuals, workflow standards for your agency management system, E&O-driven documentation standards, training material for new CSRs and producers
• Internal communication in Slack, Teams or email about process questions
• Activity and task histories from your agency management system, with insureds and contacts replaced by codes

Commercial lines agencies tend to have more varied, judgment-heavy workflows. Personal lines agencies tend to have higher volumes of repeatable ones. Both can be interesting.

What to exclude by default

• Nonpublic personal information about consumers: names, addresses, dates of birth, SSNs, driver's license numbers, driving records, claims histories, financial details
• Health information from benefits, life, disability or long-term care business
• Claims files, especially those involving injuries
• Carrier confidential material: commission agreements, underwriting guidelines, rating manuals, carrier portal content
• Consumer credit-based information
• Producer compensation and agency HR material

De-identifying agency data

Insurance records carry identifiers in many places beyond the obvious fields:

1. Policy, claim and account numbers, which are direct identifiers
2. Vehicle VINs and property addresses, which can be looked up
3. Business names and FEINs for commercial insureds
4. Loss descriptions and notes, which can describe distinctive events
5. Attachments such as applications, loss runs, ID cards and inspection photos

Replace insureds with consistent codes ("Insured 318"), generalize locations to a state or region, generalize amounts where they aren't essential, and exclude attachments unless reviewed. Our de-identification guide explains the techniques.

The goal is to keep the workflow readable ("CSR received a request to add a vehicle, confirmed garaging location, submitted the endorsement to Carrier B, and sent the updated ID card two days later") without revealing who anyone is.

Who is a good fit

• Independent agencies and brokerages that have used one agency management system consistently for several years
• Agencies with documented workflows and procedures
• Wholesale brokers and MGAs with distinctive submission and placement workflows, subject to carrier and program agreements
• Agencies that are comfortable keeping consumer data and carrier confidential information out of scope

What affects the value

• Years of activity history in a consistent system
• Line-of-business mix: commercial, personal, specialty and benefits each bring different workflows
• Procedural documentation that explains the why behind the activity
• Connected systems: agency management history plus email or chat plus procedures shows complete workflows
• Exclusions: process-focused scoping reduces volume but makes the dataset far easier to approve

Offers vary with these factors. See how much is my company's data worth?

Questions to settle before you start

• What does our privacy notice say about sharing information with third parties?
• Do our carrier agreements restrict use of carrier information or data in carrier systems?
• Does our information security program cover a project like this, and who oversees it?
• Are we a business associate for any benefits clients?
• Who has authority to approve: the principal, the partners, or a parent company?

Getting started

To get an offer, share rough estimates: lines of business, which agency management system and other tools you use (email, chat, document storage), roughly how many years of history, headcount and years in business.

DataOffer reviews insurance-sector data case by case, helps you scope around consumer information and carrier obligations, handles packaging and de-identification (or supports your team), and brings you competing offers. There's no upfront cost, and nothing is shared until you approve the buyer, price and terms.

This guide is general information, not legal, tax or financial advice. Figures and ranges are illustrative; talk to qualified advisors about your situation.