TPA and claims modernisation for insurers and health payers
We help insurers, health payers and their third-party administrators (TPAs) run member support, claims and risk review with AI agents, while your people keep every decision about cover, claims and medical necessity.
The agents work inside the claims and policy systems you already run, through their APIs and the Model Context Protocol (MCP). We bring them in stage by stage, alongside the way you work today, and nothing about an individual is decided by a machine.
Last updated
What does a modern TPA look like?
Members reach you in the way and language they choose, claims arrive complete and go to the right person, and every referral says why. Agents do the gathering and checking; your people make the decisions.
See all agents for insurers-
Member support by voice, chat and email
Members get answers on cover, claim status and documents in their own language. Calls run on Prime Assist technology, our voice AI, and our policyholder service agent answers chat and email. Each one says it is an AI and passes complex cases to your team with the conversation so far.
Who decides: Anything about cover, a claim decision, a complaint or a member who may be vulnerable goes to a person on your team.
-
Digital first notice of loss and claims intake
A claim can start by phone, web form or email. The intake agent gathers the details and documents, checks what is missing and opens the claim in your claims system, ready for a handler.
Who decides: A handler decides cover, liability and settlement. The agent never accepts or declines a claim.
-
Claims triage
New claims are sorted by complexity and urgency and routed to the right handler or team, with the reason for each routing written on the file.
Who decides: Handlers can change any routing, and members who may be vulnerable go straight to a person.
-
Prior authorisation packs
For health payers, the agent gathers the clinical documents your rules ask for, assembles each prior authorisation request and tracks it to a decision.
Who decides: A licensed clinician makes every medical necessity decision. The agent never refuses a request.
-
Claims integrity review for fraud, waste and abuse
Claims whose patterns may point to fraud, waste or overcharging are referred to your counter-fraud team with a reason code and the evidence behind each referral, built on our risk intelligence work.
Who decides: An investigator reviews every referral and decides what happens next. No claim is refused on a referral alone.
-
Bordereaux and delegated authority checks
Bordereaux from coverholders, MGAs and TPAs are checked against the binding authority and the market data standards, and every exception is listed with where it came from.
Who decides: Your delegated authority manager clears every exception.
-
Agents inside your core systems
Agents read and write through the APIs of the claims, policy and administration systems you already run, and through MCP where the platform offers it, so work stays in one system of record and every action is logged.
Who decides: Your system owner approves what each agent can read and change, and a named business owner signs off each agent before it goes live.
How do you handle risk assessment?
We build risk review that helps your people decide, never systems that decide about people. Agents rank and explain; a named person at your organisation makes each decision.
- Risk stratification for human review: members or claims ranked for a case manager or clinician to look at first, with the reasons shown for each one.
- Claims review: files checked for missing documents, coding questions and patterns worth a second look, then passed to a handler or investigator with the evidence.
- Reason-coded referrals: every flag carries a code and the evidence behind it, so the person reviewing it can see why, agree or overrule it.
- No automated decisions about individuals: the agents never accept, decline or reduce a claim, refuse a prior authorisation or cancel cover.
- No pricing: we do not build systems that set prices or premiums for anyone.
Who decides: Your handlers decide cover and claims, your investigators decide on referrals, and licensed clinicians decide medical necessity.
Which terms and technologies are payers asking for?
Agents inside core claims platforms, shared data standards, FHIR-based prior authorisation, regional claims exchanges, and evidence of good outcomes, with the rules and dates behind each one.
-
Agentic claims inside core platforms
The main claims platforms now ship their own agent frameworks. Duck Creek launched its Agentic AI Platform on 28 April 2026, with an AI Gateway that supports MCP and agent-to-agent (A2A) connections. Guidewire released Qusar on 3 August 2026, with claims and underwriting agents and an MCP server. Naming a product here is not a partnership: we build agents that connect to whichever platform you run.
Sources: Duck Creek: Agentic AI Platform launch, Guidewire: Qusar release
-
ACORD standards and the London market Core Data Record
ACORD standards carry policy, claims and delegated authority data between carriers, brokers and coverholders, now with JSON standards for APIs. In the London market, the Core Data Record is the shared data standard: version 3.3 for treaty reinsurance was launched on 18 May 2026, and delegated authority and claims versions were put out for consultation in 2026.
Sources: ACORD: digital standards for APIs, ACORD: Delegated Authority Standards, LMG: Core Data Record v3.3, LMG: Delegated Authority CDR consultation
-
HL7 FHIR, the Da Vinci guides and US prior authorisation APIs
Under the US federal rule CMS-0057-F, affected health payers must give specific reasons when they deny a prior authorisation from 1 January 2026, and must offer prior authorisation and other data exchange APIs built on HL7 FHIR from 1 January 2027. For the prior authorisation API, CMS recommends the HL7 Da Vinci implementation guides.
Sources: CMS: Interoperability and Prior Authorization Final Rule (CMS-0057-F), CMS: CMS-0057-F fact sheet
-
UAE claims exchanges: DHA eClaimLink and DoH Shafafiya
In the UAE, health claims are exchanged through the Dubai Health Authority's eClaimLink in Dubai and the Department of Health's Shafafiya in Abu Dhabi. Each exchange sets its own onboarding and technical rules, and they change, so we confirm the current rules with the exchange and with you before any work on it starts.
-
Outcomes monitoring under the FCA Consumer Duty
In its December 2025 review of outcomes monitoring, the FCA asked insurers to monitor settlement values, benchmark them and analyse why claims are declined. Completing a process is not, by itself, evidence of a good outcome. Agents that log every step and reason make that evidence easier to gather.
Source: FCA: insurance multi-firm review of outcomes monitoring
-
EU AI Act high-risk rules for life and health insurance
AI used for risk assessment and pricing of individuals in life and health insurance is high-risk under the EU AI Act from 2 December 2027, the date set by Regulation (EU) 2026/1744. Insurers that use it will need human oversight, an assessment of its impact on people's fundamental rights and explanations for the people affected.
Sources: EUR-Lex: Regulation (EU) 2026/1744, European Commission: AI Omnibus enters into force
As they stood on 26 September 2026. Rules and dates change, so every project starts by checking where each one stands, and your legal advisers confirm how each rule applies to you. For readiness under the EU AI Act, see AI assurance.
How do you modernise in stages?
One queue at a time, with the old and new ways of working running side by side until your team is ready to change over.
-
Map
Walk through your member support, claims and referral work with your team, and agree where an agent would help and who decides at each step.
-
Connect
Connect to your claims, policy and administration systems through their APIs or MCP, with the access your system owner approves.
-
Pilot one queue
Start with one queue, such as email intake or bordereaux checks, with pass marks agreed with you before anything goes live.
-
Run side by side
Run the agent alongside your current process and compare the two, so your team can see every difference before changing over.
-
Extend and review
Add the next queue once your business owner signs off, and review each agent again after every change of model, instructions or tools.
Who decides: Your business owner decides when each queue moves from side by side to live, and can move it back at any time.
Questions people ask
What is a third-party administrator (TPA)?
A TPA runs claims, member support or other administration on behalf of an insurer or health payer. It works under the insurer's authority and rules, so the insurer stays accountable for the service its members receive.
Do the agents make claims or medical necessity decisions?
No. Agents gather information, prepare files, route work and flag cases. A handler decides cover, liability and settlement, an investigator decides what happens after a fraud referral, and a licensed clinician makes every medical necessity decision.
Do you build pricing or premium setting systems?
No. We do not build systems that set prices or premiums for anyone. Our risk work supports review by your people, such as referrals to a counter-fraud team or cases ranked for a clinician to look at.
Do we have to replace our claims system?
No. The agents work through the APIs of the systems you already run, and through MCP where your platform offers it. We run the old and new ways of working side by side before anything changes over.
Are you a partner of Guidewire, Duck Creek or the other platforms named here?
No. We name these products because insurers ask about them. Naming a product is not a partnership, and we build to whichever platform you run.