Let agents settle claims. Never above anyone’s authority.

Agents can read the file, propose the settlement and pay the small ones in minutes. TraceMem holds every payment to the authority your handlers already work to, and sends anything above it to the one person who can sign it off.

Or call us on +1 650 550 1418

CLM-26-38802Motor, bodily injury

claims-agent, for Priya Shah, proposes a settlement of

€18,500

What the model read, after TraceMem

  • GP letter2 diagnoses tokenized
  • Physiotherapy notes1 mental health note obfuscated
  • Dashcam still1 bystander's face painted over

On the record for this claim

Who the agent worked for, what it read, what the model saw, the policy version that applied, and who signed the payment off.

Who it reaches, under claims_settlement_authority, version 3

  1. The agent paysWithin its own authorityto €2,500
  2. Claims handlerPriya Shah, in Microsoft Teamsto €10,000
  3. Team leadMarta Kowalski, in Microsoft Teamsto €25,000
  4. Claims managerDaniel Osei, in Slackto €75,000
  5. Agents don't settleBack to the handlerover €75,000
Settlement paymentCLM-26-38802
Pay
€18,500
To
The claimant, bank details tokenized

Held for Marta Kowalski

Team lead, asked in Microsoft Teams. Nothing is paid until Marta answers.

Illustrative claims. Names, amounts and limits are made up; the limits are yours to set.


Agents are moving into every stage of a claim.

Each stage hands an agent somebody else’s documents. Pick a stage to see what goes wrong without a governed path, and what TraceMem does instead.

First notice of loss

Taking a claim from an email, a web form or a transcribed call, photographs and documents attached.

Without a governed path

A bystander's face in a dashcam still and a full driving licence go to the model provider.

With TraceMem

Faces and identity documents are painted over before any model sees the images, and a retried write cannot open the same claim twice.

PII Protection

Regulators are already asking how claims get decided.

Some of these already apply and some are still coming. Between them they ask how automated decisions were made, who was accountable and where the data went, and only a record kept at the time can answer.

  1. 31 Jul 2023 · UK

    The Consumer Duty applies

    Insurers must show their customers get good outcomes, claims handling included, whoever or whatever handled the claim.

    Source (opens in a new tab)
  2. 4 Dec 2023 · US

    NAIC Model Bulletin on AI

    A written programme governing every AI system an insurer uses, third parties' included. 24 states and DC had adopted it by April 2026.

    Source (opens in a new tab)
  3. 1 Jan 2025 · US

    California SB 1120

    A licensed clinician, not an algorithm, must decide a health plan's medical-necessity denials.

    Source (opens in a new tab)
  4. 17 Jan 2025 · EU

    DORA applies to insurers

    ICT risk, incident reporting and third-party oversight. A model provider your agents call is one more dependency to account for.

    Source (opens in a new tab)
  5. 22 Jul 2025 · UK

    FCA review of claims handling

    Home and travel insurers were found with weak oversight of outsourced claims handling and no documented criteria for cash settlements.

    Source (opens in a new tab)
  6. 6 Aug 2025 · EU

    EIOPA Opinion on AI governance

    Record-keeping, explainability and human oversight, in proportion to risk, for the AI systems insurers use.

    Source (opens in a new tab)
  7. 1 Sep 2025 · US

    Texas SB 815

    A utilization review agent may not use an automated decision system to make an adverse determination, wholly or partly. It reaches health plans issued or renewed from 1 Jan 2026.

    Source (opens in a new tab)
  8. 15 Oct 2025 · US

    Colorado extends AI governance

    The Division of Insurance's governance rules for algorithms and external data now reach auto and health insurers, not only life.

    Source (opens in a new tab)
  9. 5 Feb 2026 · UK

    UK rules on automated decisions apply

    Under the Data (Use and Access) Act, significant automated decisions taken from this date need safeguards: telling the person, human review on request, a right to contest.

    Source (opens in a new tab)
  10. 1 Jul 2026 · US

    Colorado compliance deadline

    Auto and health insurers must comply, and file their first annual compliance report on algorithms and external data.

    Source (opens in a new tab)
  11. 6 Jul 2026 · UK

    FCA Mills Review published

    It asks how firms oversee autonomous and agentic models, and how senior managers stay accountable for them.

    Source (opens in a new tab)
  12. Fall 2026, expected · US

    NAIC AI evaluation tool

    Piloted by 12 state regulators since March 2026, and expected to be adopted at the NAIC's Fall 2026 meeting.

    Source (opens in a new tab)
  13. 1 Jan 2027 · US

    Colorado automated decisions law

    Consumer notices and rights around automated decisions in insurance and lending, under SB 26-189.

    Source (opens in a new tab)
  14. 2 Dec 2027 · EU

    Life and health pricing becomes high-risk AI

    Risk assessment and pricing for individuals in life and health insurance: human oversight and logging expected of the insurer.

    Source (opens in a new tab)

Dates checked against each regulator’s own publication, September 2026. Not legal advice.


Automated decisions at insurers already have a file with regulators.

None of these involved an AI agent. Each was a system acting for an insurer, and each came down to decisions nobody could account for, or data that went where it should not.

Liberty Mutual Insurance Europe

2010 to 2015 · FCA, 2018

£5.28 million

FCA fine (opens in a new tab)

Mobile phone claims and complaints were handled on its behalf by a third-party administrator, and some claims were unfairly declined or not investigated adequately.

What it came down to

Claims decided in its name, without the oversight to see how they were decided.

For an agent, with TraceMem

Every action an agent takes on your behalf names who it acted for, the rule that applied and who decided.

Decision Trace

GEICO and Travelers

2020 to 2021 · NYDFS, 2024

$11.3 million

NYDFS penalty (opens in a new tab)

Attackers pulled driver's licence numbers and dates of birth out of online quoting tools and an agent portal, for about 120,000 New Yorkers.

What it came down to

Automated tools that exposed identifiers they should have kept back, and a portal without multi-factor authentication.

For an agent, with TraceMem

Identifiers are tokenized before a prompt leaves your network, and every caller is verified against your own directory.

PII Protection

UnitedHealthcare

2019 to 2022 · US Senate staff report, 2024

8.7% to 22.7% denied

Senate report (opens in a new tab)

Its denial rate on prior authorization requests for post-acute care rose from 8.7% in 2019 to 22.7% in 2022, as it used an algorithm-driven review tool. Lawsuits over the tool are pending, and their claims are allegations.

What it came down to

Coverage decisions shaped by an algorithm, with a person's part in each one hard to show.

For an agent, with TraceMem

Above the limits you set, an agent cannot decide. One named person does, and the record says who.

Policy and Exceptions

Every month you wait, agents handle claims nobody can account for.

Put in your own numbers. What your agents read and paid before a governed path is in place can never be recorded afterwards.

100
6 months

21 working days a month. Your numbers stay in your browser.

In 6 months of waiting

100,800

Agent actions with no record of what they read, what the model saw or who signed them off.

Next 18 monthsNo recordOn file

Settle the first claim on the record this month.

Pick whichever is quickest for you. However you get in touch, a person replies. With our forward-deployed engineers, TraceMem can be running inside your network in days.

What would you like?
Where are your agents working?

Pick any that apply.

When do your agents act on production systems?

Optional. A sentence is enough.

45days, free

Free for 45 days on our Docker-based trial image, inside your own network. When you are ready, it upgrades to the full enterprise product.

Forward-deployed engineers, ours or our partners', can set it up in days and help you define policies and integrations. This may be charged separately.