Tau
AI for insurance · 11 lessons

AI in insurance for claims and underwriting

Use AI to summarise claim files, explain policy wording in plain English, draft letters and sort new business, treating every customer fairly. For claims handlers, underwriters and brokers.

Who it’s for: Claims handlers, underwriters, brokers, actuaries and insurance ops teams.

Insurance runs on documents. Claim files run to hundreds of pages, policy wordings are dense and submissions arrive in every format. AI can read and sort that material quickly. This course teaches you to use it while treating every customer fairly.

You'll summarise a long claim file into a timeline, a list of the parties, the coverage questions and any missing documents, all with page references. You'll explain an exclusion or endorsement in plain English and check it line by line against the wording. You'll draft acknowledgement, information-request and decline letters that cite the policy basis. For underwriting, you'll pull exposure data from broker submissions, check it against your appetite, meaning the risks you want to write, and compare a broker's wording with your standard form.

The later lessons deal with fairness. You'll treat fraud scores as a reason to look closer, never as proof. You'll decide when a photo damage estimate can stand and when a person should inspect. You'll also learn to test pricing models for unfair bias, and see what regulators expect of AI in insurance. The course is for claims handlers, underwriters, brokers, actuaries and operations teams.

By the end you’ll be able to…

  • Summarise a claim file into facts, coverage questions and next actions
  • Write a plain-English policy wording explainer that stays accurate
  • Triage underwriting submissions with AI and record the decision
  • Test a pricing or claims model for unfair discrimination

Short on time? Start with the essentials

These 4 lessons cover the most useful ideas in this course. Each one takes about a minute.

Essential 1

Where AI fits from quote to claim

Map AI uses across distribution, underwriting, pricing, policy servicing, claims and fraud, and mark where a human must decide.

Essential 2

Summarise a 200-page claim file in minutes

Use AI to pull a timeline, parties, coverage questions and missing documents from a claim file, with page references.

Essential 3

Write claims letters that are clear and fair

Use AI to draft acknowledgement, information-request and decline letters that cite the policy basis and meet your fair-handling rules (e.g. US state unfair claims practices laws, the UK FCA Consumer Duty).

Essential 4

Triage underwriting submissions with AI

Extract exposure data from broker submissions, check appetite fit and flag missing information before underwriter review.

Every lesson in this course

This course has 11 video lessons of about a minute each, from the basics to expert habits. Each one ends with a question, three takeaways and its sources.

Foundations 2 lessons

  1. Where AI fits from quote to claimEssential

    Map AI uses across distribution, underwriting, pricing, policy servicing, claims and fraud, and mark where a human must decide.

    “Quote, bind, claim, pay: AI already touches every step. Where should it stop?”

  2. Summarise a 200-page claim file in minutesEssential

    Use AI to pull a timeline, parties, coverage questions and missing documents from a claim file, with page references.

    “200 pages of adjuster notes, reports and emails become a one-page claim summary.”

Building skills 4 lessons

  1. Explain policy wording in plain English

    Draft a customer explainer for an exclusion or endorsement that is readable and checked line by line against the wording.

    “'Notwithstanding the foregoing…' Can your customer understand their own policy?”

  2. Write claims letters that are clear and fairEssential

    Use AI to draft acknowledgement, information-request and decline letters that cite the policy basis and meet your fair-handling rules (e.g. US state unfair claims practices laws, the UK FCA Consumer Duty).

    “A decline letter is where complaints start. AI can draft it; you must own it.”

  3. Triage underwriting submissions with AIEssential

    Extract exposure data from broker submissions, check appetite fit and flag missing information before underwriter review.

    “What if every broker submission arrived already read, keyed and gap-checked?”

  4. Compare a broker's wording against your standard form

    Use AI to produce a clause-by-clause comparison of a manuscript wording against your standard form, flagging every broadened cover or deleted exclusion with clause references.

    “The broker says it's "market standard". Clause 14 quietly drops the flood exclusion.”

Advanced 4 lessons

  1. Use AI fraud scores without accusing the innocent

    Treat fraud scores as referral prompts, investigate with evidence and avoid biased or unexplained adverse decisions.

    “A fraud score is a reason to look closer, never a reason to deny.”

  2. Decide when a photo claim needs a human inspector

    Set rules for when an AI photo estimate can stand and when to send an inspector, based on damage type, image quality, claim value and where computer vision struggles.

    “From a phone photo to a repair estimate, and when to send a human anyway.”

  3. Test pricing and underwriting AI for unfair bias

    Explain proxy discrimination and run basic outcome-disparity tests on a pricing or underwriting model, as some regulators now require (e.g. Colorado's testing rules for life insurers in the US), and check what applies in your market.

    “Postcode isn't a protected trait. But in a pricing model it can act like one.”

  4. Know what insurance regulators expect of your AI

    Summarise the governance, documentation and testing expected for AI in underwriting, pricing and claims, using examples such as the NAIC Model Bulletin (adopted by many US states), the EU AI Act and EIOPA's AI opinion (EU), and check what applies in your market.

    “Life and health insurance pricing AI is 'high-risk' under the EU AI Act. Now what?”

Expert 1 lessons

  1. Set the rules for fast-tracking claims with AI

    Define which claims AI may route to straight-through settlement by type, value, coverage certainty and fraud signals, and set the sample audit that catches leakage and unfair outcomes.

    “Same-day payouts are possible. Which claims should get one, and who checks?”

Apps we recommend

These are the tools this course uses, and what each one is best for. We have no partnership with any of them.

Sixfold

Best for: AI underwriting assistant for carriers and MGAs

Reads broker submissions against your underwriting guidelines and summarises risk, gaps and appetite fit.

Paid

Shift Technology

Best for: Claims automation and fraud detection

Insurance-specific AI that flags suspicious claims and automates claim decisions for insurers worldwide.

Paid

Tractable

Best for: Photo-based motor and property damage assessment

Computer vision estimates vehicle and property damage from photos to speed up claims and repair decisions.

Paid
Hands-on project

Build an AI-assisted claim or submission pack

Take a fictional or fully anonymised claim file or broker submission from your line. With approved tools, produce a page-referenced summary, coverage or appetite questions and a customer or broker letter. Then map your process, marking what AI automates, assists or never touches, with KPIs.

What you’ll make: A summary with page references, a checked draft letter and a one-page process map showing the AI/human split and four KPIs.

Check your own work

  • Every fact in the summary cites a page or document in the file
  • The letter quotes the policy clause it relies on, and you checked it line by line
  • Every decline, adverse or fraud-related decision in the map is owned by a named human role
  • The KPIs include cycle time plus at least one fairness or complaints measure
  • No real policyholder or broker data went into an unapproved tool

Try a ready-made prompt

Copy a better prompt for a real task, and see why it works better than the version most people type.

Further reading and viewing

We’ve chosen these articles, research papers, videos and guides from other trusted sources.

Guide · NAIC · 25 min

Model Bulletin: Use of Artificial Intelligence Systems by Insurers

The US regulators' model for AI governance, testing and third-party oversight, adopted in December 2023.

Guide · EIOPA · 40 min

Opinion on Artificial Intelligence Governance and Risk Management

How EU supervisors expect insurers to govern AI, from data quality to human oversight (August 2025).

Research · The Geneva Association · 45 min

Gen AI Risks for Businesses: Exploring the role for insurance

Survey-based report on how insurers can underwrite and cover the new risks businesses face from generative AI.

Questions about this course

How is AI used in insurance claims?

AI can summarise long claim files, draft letters and flag claims that might need a closer look. Tools such as Shift Technology help with claims automation and fraud detection, and Tractable estimates damage from photos. In the course, you learn to keep a person in charge of decisions. You'll set rules for when a photo estimate can stand and when an inspector should visit.

Can AI help underwriters?

Yes. AI can pull exposure data out of broker submissions, check it against the risks you want to write and flag missing information. AI underwriting assistants such as Sixfold do this for insurers and MGAs, the agencies that write business on an insurer's behalf. The underwriter still makes the decision. The course also shows you how to compare a broker's wording with your standard form, clause by clause.

What do insurance regulators expect when you use AI?

Expectations are growing, and they vary by market. In the US, many states have adopted the NAIC Model Bulletin on AI. In the EU, the AI Act and EIOPA's opinion on AI set out what's expected. Fair-handling rules, such as the UK FCA's Consumer Duty, still apply to AI-drafted letters. The course uses these as examples. Check what applies in your market, and keep customer data in approved tools.

Do I need to know how AI works to take this course?

No. Tau is for people who use AI at work, so you don't need to know how it works inside. Each lesson is a short video of about a minute, followed by one question, three takeaways and the sources. A three-question AI skills check sets your level. Your first 5 lessons are free, and Tau Pro gives you every course and lesson.

Get early access to this course

Tau opens soon. Join the waitlist and we’ll email you as soon as you can start this course.