Airclerk helps insurers, brokers, wealth managers, accountants and other regulated firms in New Zealand and Australia move from AI experimentation to production. Claude is our default agent for new builds; we also implement Microsoft Copilot and ChatGPT where they fit your existing stack - workflows, integrations, governance and controls designed for regulated operating environments.
A fixed-price, two-week engagement that turns scattered AI experiments into a board-ready Claude implementation plan.
Most financial and professional services firms have already tried AI. The boards have heard the demos, the teams have run the workshops, the slides have been printed. Many pilots stall before production.
The blockers are not usually model capability. The blockers are governance, data access, integration, permissions, auditability, workflow ownership and change management - exactly the problems AI vendors won't solve for you.
Models like Claude and Microsoft Copilot are absorbing many of the standalone AI tool use cases that defined the last two years. The work that's actually hard is now down the stack - integration with policy systems and claims platforms, governance, and an audit trail of the AI-assisted work itself.
Airclerk doesn't push users into yet another portal. People work inside the model they already use. Airclerk builds the rails underneath: connectors into your existing systems, orchestration, human approval, audit trails and governance. And it runs where your risk team needs it - hosted by Airclerk, or deployed single-tenant into your own Azure tenant on request.
We built that layer, and named it. Boards are under pressure to put AI to work while carrying the accountability if it goes wrong, so the brief lands on someone: make it safe enough to say yes to. AI Process Assurance is how they answer it - charter the path the work should follow, record the path it actually took, and show the difference.
Plan, build and operate AI in regulated environments - picked up at the depth that fits where you are today.
Diagnose existing AI activity. Identify the first workflow worth taking to production. Map the path, governance and architecture.
Learn more → 02 — BuildBuild governed Claude workflows connected to real systems — via MCPs, connectors, and Claude Code in engineering teams.
Learn more → 03 — OperateDesign controls, approvals and audit trails for production AI. Monitor, improve and expand workflows over time.
Learn more →The industries we like working with share a shape. The work runs on documents and judgement, and the decisions may have to be explained long after they're made. That shape shows up everywhere from underwriting to engineering sign-off.
Airclerk's strongest practice. Submission to market, renewals, broker inbox triage, FAP file completeness, claims advocacy and complaints handling.
Explore brokers →For insurers and MGAs: underwriting review, claims summarisation, renewal packs, policy wording comparison and compliance evidence. Significant MGA experience.
Explore insurers & MGAs →Governed Claude workflows for advisers, paraplanners and compliance - file review, advice preparation and ongoing service evidence.
Explore wealth & advice →Claude for NZ law firms - matter chronologies, discovery review, contract review and briefs of evidence, built on the open-source Claude for Legal suite localised for New Zealand.
Explore legal →Claude for NZ accounting firms - year-end working papers, provisional tax, FBT and IRD correspondence, with the review and evidence discipline the work demands.
Explore accounting →Credit memo drafting, responsible-lending file completeness and covenant monitoring - the document work behind every decision a lender may later have to defend.
Explore lending →Design report checking, peer review preparation and producer statement readiness - is the file behind the signature actually complete?
Explore engineering →Proposal and SOW drafting, deliverable QA against the brief, and engagement files that show who reviewed what before it shipped.
Explore professional services →These are examples, not a boundary. We're drawn to complex processes and firms that like doing things properly - if someone later asks to see your working, we should talk.
Talk to us →Document-heavy, regulated, deeply unforgiving of demoware. That's the work we know.
We've spent careers building services around transformational platforms - and surviving the long tail of integration, compliance and change that follows.
Local insurers need a partner with regional context, systems literacy and people on the ground. We also work remotely with select international clients - particularly US carriers and brokers - who value implementation-led delivery over another AI wrapper. Two decades of cross-border consulting tells us three hours of daily overlap with New Zealand is enough - that's the US west of the Rockies, and Asia from Singapore east.
ISO 27001 and SOC 2 Type 2 in progress. Governance isn't a phase - it's part of the product.
Claude is our preferred platform for regulated workflows - chosen for its safety posture, structured reasoning and audit-friendly outputs. Where Microsoft Copilot, ChatGPT or other models fit a client's existing landscape better, we implement those too.
Airclerk is an AI implementation consultancy for regulated financial and professional services in New Zealand and Australia. We help insurers, brokers, wealth managers and other regulated firms move AI from experiment to production — and prove what it did. Claude is our preferred platform; we also implement Microsoft Copilot and ChatGPT where they fit a client's landscape.
It is the category Airclerk pioneers. A Workflow Charter defines the path a repeatable piece of AI-assisted work is meant to follow; a Semantic Audit records the path it actually took; comparing the two surfaces missing steps, missing approvals and bypassed gates as exceptions. The retained result is an AI Workpaper — an audit-ready record of what the AI did, per run.
Claude (Anthropic) is our default for regulated work, chosen for its safety posture, structured reasoning and audit-friendly outputs. We also implement Microsoft Copilot, ChatGPT and other models where they fit a client's existing landscape. We are not a model provider and we do not resell licences.
Airclerk is based in Christchurch, New Zealand, and works across Australia and remotely with select international clients, particularly US carriers and brokers.
Most engagements begin with a two-week Claude Production Readiness Sprint that turns scattered AI activity into a board-ready implementation plan. Ask a question or request the sprint through our contact page — we respond within one business day.