Airclerk NZ · AU Talk to us →
Industry · Insurers & MGAs · NZ & AU

The submission read, summarised and checked before the underwriter picks it up.

Airclerk puts AI to work on the document-heavy side of underwriting and claims for insurers, MGAs and underwriting agencies in New Zealand and Australia. The workflow reads the file, prepares the summary and lists what is missing. Appetite, pricing and the decision stay with the underwriter.

We implement Claude, Microsoft Copilot and ChatGPT; the platform is chosen around your systems and the work.

Talk to us → See the underwriting example ↓
Underwriter or claims handler workstation with a policy file open alongside a carrier policy admin interface.
01 / An example · a commercial submission

Forty pages in. Three things missing.

Illustrative example

Underwriting document review is the workflow we would start with at most insurers and MGAs, so it is the one drawn here. The risk, the figures and the gaps are made up. The shape of the work is not.

Source documents
Broker submission packExpiring scheduleFive-year claims historySurvey report

Read as they arrive from the broker, in the formats agreed during design: PDFs, spreadsheets and the covering email. The policy admin system and the document store stay the systems of record.

What the workflow prepares
Underwriting summaryMissing-information listChanges since expiryDraft broker query

A summary of the risk with each statement cited to the page it came from. What the submission does not answer, listed rather than assumed. What moved since the expiring terms. A draft email back to the broker asking for the gaps, never sent by the workflow.

The underwriter's review
Appetite and pricing stay humanAuthority limits applied

The underwriter reads the gaps first, then the summary, and decides appetite, terms and price. Where the risk sits above an authority limit, the workflow marks it for referral instead of letting it through. At an MGA that limit is the binding authority, and the referral goes to the capacity provider rather than an internal line manager. Same workflow, different gate.

AI Workpaper · illustrative commercial submission Illustrative · not a client record · Assurance enabled
  • Day 1 Workflow read the submission pack, claims history and survey. Summary prepared with page citations; sum insured up on expiry, sourced to the broker's covering email. · workflow
  • Day 1 Missing information listed. Summary held for the underwriter. · control ↳ no sprinkler certificate · survey older than the insurer's threshold · one claim in the history not explained in the submission
  • Day 2 Underwriter reviewed the summary, approved and sent the broker query, set appetite as "quote subject to survey". · underwriter
  • Day 6 Documents received, summary updated, terms issued by the underwriter. Sources, gaps and the approval retained on the workpaper. · underwriter
The underwriter still made every decision. What changed is that the reading and the checking were done before the file was opened, and the record of both was kept. Keeping that record is what our optional AI Process Assurance platform does.
The short answer

How do insurers and MGAs put AI to work?

Insurers, MGAs and underwriting teams use AI models such as Claude, Microsoft Copilot or ChatGPT to support insurance workflows such as underwriting review, claims file summarisation, policy wording comparison, broker communication, renewal review and compliance evidence preparation. In regulated insurance environments, the assistant needs to be connected to documents, systems, permissions, human approval workflows and audit trails before it can be used safely in production.

02 / Further opportunities

Other insurance workflows with the same shape.

Each of these is a candidate first workflow, and we have accelerators for several of them. The common thread is a document-heavy job with a named reviewer who already knows what a good output looks like.

WF-01

Claims file summarisation

Chronologies, key issues, outstanding actions, citations across large claims files.

WF-02

Renewal pack comparison

Year-on-year changes, missing data, draft broker follow-up, renewal state.

WF-03

Broker inbox triage

Shared-mailbox classification, urgency, missing documents, draft responses.

WF-04

Policy wording comparison

Exclusions, changes, coverage differences with evidence and citations.

WF-05

Complaints handling

Triage, customer history, regulatory timeframes, draft responses, evidence packs.

WF-06

Compliance evidence review

Control evidence, policy compliance, audit prep, evidence gaps.

WF-07

Internal knowledge assistant

Permission-aware search across policies, procedures and internal knowledge.

WF-08

Board & executive reporting

Summaries of operational data, executive packs, evidence-backed narratives.

Commercial renewals are the pattern we have taken furthest, because Airclerk began by building renewal software for insurance: the renewal runs as a governed, evidenced workflow, with optional AI Process Assurance recording each outcome. Direct-to-consumer insurers and MGAs also ask about putting quoting and claim service inside ChatGPT and Claude; that is a different build, described on its own page.

03 / How an engagement runs

One conversation, then one workflow into production.

The same four stages whether you are a general insurer, a life office or an MGA writing under delegated authority. Only the scope, and the gates inside it, change.

01 · Discuss
Discuss the opportunityFree · introductory

Where AI could make the biggest difference, how that work runs today, the systems it touches and who owns it. We also work out together whether Airclerk is the right partner for it. If we are, the conversation gives us enough to quote.

02 · Agree
Agree the scopeWritten proposal · fixed fee

One document: the outcome we are aiming for, what we deliver, what your team contributes, the fixed fee and timeline, and what acceptance looks like. Nothing starts until it is agreed on both sides.

03 · Build
Design and buildDetailed design first

The engagement opens with the detailed design, agreed with you before anything is built. Then we configure and connect the workflow to your documents and systems, build in the agreed review controls and audit trail, test it on real work, and prepare the people who will run it.

04 · Support
Support and improveHandover · ongoing by agreement

Handover, and the support that follows go-live, are written into the scope so you know what is covered before you sign. Ongoing operations and further workflows are a separate agreement, made when you want them rather than switched on by default.

Read how an engagement runs in detail →

Hosted by Airclerk, or deployed single-tenant into your own Azure tenant on request, so the controls, evidence and audit trail sit where your risk and compliance teams need them.

04 / Controls and local context

Document-heavy. Judgement-driven. Audit-aware.

  • Insurance is document-heavy at every stage of the lifecycle.
  • Workflows depend on judgement, context and evidence, not single-shot answers. The workflow prepares; a named person decides.
  • Existing systems are fragmented across policy admin, claims, broker portals and legacy platforms. They stay where they are; the workflow reads from them.
  • Teams operate across email, documents, spreadsheets and shared mailboxes rather than a single application.
  • Authority limits are a control, not a footnote. An insurer's internal referral thresholds and an MGA's binding authority both become gates the workflow enforces before anything is issued.
  • Regulatory, conduct and audit expectations matter, and they vary by line and jurisdiction.
i.

Global AI vendors are likely to prioritise US insurance first.

Roadmaps follow market size. ANZ insurers are unlikely to be the first served.

ii.

Local workflows, systems and market structures.

NZ and Australian insurers have local platforms, brokerage structures, regulators and conduct expectations.

iii.

A local partner with regional fluency.

Airclerk operates across New Zealand and Australia, with active relationships across insurers, MGAs and brokers.

iv.

NZ insurance contract law is mid-transition.

The Contracts of Insurance Act 2024 reshapes proposal questions, policy wordings and claims decision processes, with a backstop commencement of 15 November 2027. AI is a natural fit on the document side: plain-language wording reviews, proposal-question audits and consistency checks across a book.

05 / Common questions

Common questions.

01How do insurers and MGAs put AI to work, and how does Airclerk implement it?

Insurers, MGAs and underwriting teams use AI across document-heavy work such as underwriting document review, claims file summarisation, renewal pack comparison, policy wording comparison, complaints handling, compliance evidence review and board reporting. Airclerk designs and implements these as governed workflows, connected to your documents, systems, permissions, human approval steps and audit trails, so the work can move from pilot into production safely. We implement Claude, Microsoft Copilot and ChatGPT, choosing the platform around your systems and the work.

02What is AI Process Assurance and how does it apply to regulated insurance work?

In plain terms: we agree how a piece of underwriting or claims work should run before it is built, and keep a readable record of what happened on each file. Airclerk calls this AI Process Assurance: a Workflow Charter sets out the path the work should follow, a Semantic Audit records what actually happened, and the result is a retained AI Workpaper. It is optional platform software, sold on subscription and implemented alongside the workflow. For underwriting, claims and renewals, that means the reasoning, evidence and human approvals behind an outcome are captured rather than lost, giving your risk and compliance teams something to stand behind.

03Does Airclerk capture a complete record of everything our AI does?

Airclerk records the work that runs through the instrumented workflows and tools it implements: the inputs, the steps, the evidence and the human approvals along that path. It does not claim to capture every interaction or guarantee total capture; what is evidenced is the designated work that flows through the governed rails Airclerk builds with you.

04Can Airclerk help us evidence our readiness for the Contracts of Insurance Act 2024?

The Contracts of Insurance Act 2024 reshapes proposal questions, policy wordings and claims decision processes for New Zealand insurers, with a backstop commencement of 15 November 2027. Airclerk helps you evidence your own readiness on the document side, through plain-language wording reviews, proposal-question audits and consistency checks across a book, with the reasoning and evidence retained as an AI Workpaper. This is implementation and evidence support, not legal advice.

05Where does our data and audit trail sit when Airclerk implements a workflow?

Workflows can be hosted by Airclerk or deployed single-tenant into your own Azure tenant on request, so the controls, evidence and audit trail sit where your risk and compliance teams need them. Workflows connect to your existing documents and systems with permission-aware access, rather than requiring everything to move into a single new application.

Talk to us

Discuss your underwriting or claims workflow.

Bring the file your underwriters or claims handlers spend the most time reading. A free introductory conversation is enough to tell you whether it is a good first workflow and what it would cost, with the fee fixed in writing.

Talk to us →