Airclerk NZ · AU Talk to us →
Service · Customer-facing AI apps · NZ & AU

Your insurance, inside the chat your customers already use.

A customer who has connected your app asks ChatGPT or Claude what cover would cost. Your product answers in that same conversation: a real quote with the agreed disclosures, and a way to buy the policy. Airclerk builds and launches those apps for direct-to-consumer insurers and MGAs in New Zealand and Australia.

Talk to us →
An insurance quote returned inside an AI chat conversation, with price, product summary and disclosure block.

Illustrative example A mocked-up quote conversation, showing the shape of the response rather than a live product.

01 / At a glance
Who it's for
Direct-to-consumer insurers and MGAs who quote and service customers online today and want the same product available where customers now ask. Also insurers who want a broker submission channel that is not a forty-message email chain.
What improves
Distribution moves to where the question is asked. A customer gets a quote or a claim update without leaving the chat, and every one of those conversations carries your disclosures, an audit trail and a complaints route from the first day.
What you get
An app callable inside ChatGPT and Claude, built once for both, connected to your quoting and policy systems, with disclosure blocks, evidence capture for every quote, complaints routing and regulator-readable logs. Submitted to the app directories and launched.
What you bring
A product owner, someone from compliance to sign off the disclosure model, and access to your quoting and policy APIs (or your submission case-management system for the broker version). Your web channel's disclosures and product summaries are the starting point.
Timing and cost
A fixed fee, quoted in writing after a free introductory conversation. The integrations behind the app set the timeline, so a quote-only app and one with claim service are different builds. Platform accounts, directory listings and usage costs sit with you, under your agreements with OpenAI and Anthropic.
Afterwards
Launch, handover and the support that follows are written into the scope. Watching the conversations, updating disclosures as products change and extending the app to new products are ongoing operations, agreed separately.
02 / Two different things

Connecting your AI to your systems is not this.

Both use the same plumbing, which is why they get confused. They serve opposite directions.

Internal · your staff

Your AI reaching your systems

Your underwriters' Claude or Copilot reading the policy system, the claims platform, SharePoint and the shared mailbox. Connectors and MCP servers scoped to a workflow, inside your tenancy. That is part of every workflow build, described under connectors and integrations.

Customer-facing · this page

Your product reaching your customers' AI

A member of the public, in their own ChatGPT or Claude account, asks about cover. Your app is what answers. It is a distribution channel with the same disclosure and audit obligations as your website, and it is what the rest of this page is about.

The short answer

What is an MCP app for insurance?

An MCP (Model Context Protocol) app exposes an insurer or MGA's products, quoting and claim service, as a callable app inside ChatGPT, Claude and other AI assistants. OpenAI ships these as ChatGPT Apps (October 2025); Anthropic ships the equivalent as MCP Apps in Claude.ai (January 2026). Both build on the Model Context Protocol, an open specification; the MCP Apps extension is what lets an app show its interface inside the assistant. The same surface can also serve insurer-to-broker workflows: replacing email as the channel through which a broker submits a quote request and an underwriter responds, without connecting to a rating engine.

03 / The shift

Customers are moving from the search bar to chat.

  • For a growing share of customers, the first place they ask "how much would insurance cost for…" is ChatGPT or Claude, not Google.
  • OpenAI shipped ChatGPT Apps at DevDay in October 2025. Anthropic's equivalent, MCP Apps on Claude.ai, shipped in January 2026. Both surfaces run on the same Model Context Protocol: build once, ship to both.
  • In February 2026, Insurify launched the first insurance comparison app inside ChatGPT and Tuio launched the first direct-to-consumer insurance quoting app. Major global broker stocks fell sharply on the news. In March 2026, Aviva announced its own ChatGPT home-insurance quoting app. The channel is not theoretical.
  • The first NZ and AU insurers and MGAs to ship stand to earn distribution inside the chat before the channel is contested locally.
  • Done well, this is your product, callable as an app, with the same disclosures and audit obligations as any other channel. It is not a chatbot bolted onto your website.
04 / Use cases

Two distinct shapes.

Primary: direct-to-consumer in the chat, with real quotes and real claim service. Secondary: a broker submission channel that replaces email between brokers and underwriters.

Primary · Direct-to-consumer

Customers describe what they need in ChatGPT or Claude. Your product responds in the same conversation, bound by the same disclosure and audit obligations as your direct web channel.

APP-01

Quote

Customer describes their need in ChatGPT or Claude. Your app returns a real quote with the disclosures and product summary your regulator requires, and shows the customer how to buy the policy.

APP-02

Claim status

Identity-verified policyholder asks for a claim update. Your app returns sanitised status, next steps and expected timeframes.

Secondary · Insurer to broker

Submissions in the chat, not in email.

A different shape, and a different audience. Brokers raise submission requests through the same conversational surface to your underwriting team: structured fact-find capture, clarification back and forth, and the resulting indication or quote, all in one thread instead of a forty-message email chain.

Not connected to your rating engine. Pricing stays with your underwriters. The chat replaces email as the channel for the request, the back and forth and the outcome, captured, structured and audit-ready.

05 / How we ship it

From the first conversation to the app directory.

01 · Discuss and scope
Free conversation · about 30 minWritten scope · fixed fee

Which product, which systems sit behind it, who signs off disclosures. Enough to quote. The engagement then opens with the detailed design: use case, MCP surface, disclosure model and target architecture, agreed before anything is built.

02 · Build
MCP serverIntegrationsConversation design

The app itself, the integrations (to your quoting and policy systems for direct-to-consumer; to your submission and underwriting case management for broker use, with no rating engine required), and the conversation patterns customers or brokers actually use. Tested against your real products before launch.

03 · Govern
DisclosureAudit trailComplaints

Disclosure blocks, evidence capture for every quote, FSLAA / CoFI-aligned complaints routing, regulator-readable logs. Built into the app, not retrofitted.

04 · Launch and hand over
ChatGPTClaudeOthers

Submission to the ChatGPT apps directory, the Claude apps surface and any other MCP-compatible assistant your customers use. Each platform reviews and approves apps on its own timetable; the written scope sets out those dependencies. Handover and post-launch support as written in the scope; ongoing operation by separate agreement.

06 / Why Airclerk

Insurance depth. MCP fluency.

i.

Insurance from the inside.

Our strongest practice is brokers, and we have done significant work with insurers and MGAs. We know what a quote and a claim conversation actually has to do.

ii.

Apps and MCP from the inside.

ChatGPT Apps (October 2025) and Claude MCP Apps (January 2026) both run on the same MCP protocol. We build MCP servers and connectors as a regular part of our practice; customer-facing apps use the same surface.

iii.

Governance built in, not bolted on.

Disclosure, audit trail, identity and complaints belong in the app itself. Done at launch, not retrofitted.

iv.

NZ and AU regulatory fluency.

FSLAA, FAP, CoFI and the FMA's conduct expectations in New Zealand, and ASIC's conduct and complaints-handling expectations in Australia, shape how an insurance app should behave inside an AI chat. We agree the disclosure and complaints requirements with your compliance team for each market before building.

Start here

Discuss an app inside ChatGPT or Claude.

A free introductory conversation about the product, the systems behind it and who signs off disclosures. If it is a fit, a written scope with a fixed fee follows.

Talk to us →