Airclerk helps financial and professional services firms in New Zealand and Australia put AI to work. We deploy it into a workflow and embed it in how the work is done, we keep it working once it is live, and where a decision has to come first, we advise on it. Claude, Microsoft Copilot or ChatGPT, whichever fits your environment.
This is where almost every engagement starts. You bring a piece of work: the renewal note, the year-end workpaper file, the accounts inbox, the matter chronology. We design the workflow against how that work is really done, connect it to your documents and systems, and build it into production with the review controls and audit trail it needs. It ends with the owner running it, not with a demo.
Design, build, test, acceptance and handover for one workflow at a time. Your workflow owner gives time in design and testing; you arrange access and sign off the controls. Fixed fee and timeline quoted in writing after a free introductory conversation.
Workflow implementation → Inside every build · also standaloneThe controls model a workflow runs under: what the AI can access and recommend, where approval is required, what is logged, how outputs are reviewed and what it costs to run. Written down, not assumed.
AI governance →A workflow is only useful if the AI can reach the documents and systems the work lives in. Claude and Copilot both reach Microsoft 365, SharePoint, Outlook and Teams through connectors the platforms provide. For everything else we build MCP servers and connectors: CRMs and broker portals, policy and claims systems, Xero, document repositories, and browser automation where an API does not exist. Permission-aware and logged, scoped to the workflow.
Talk to us →For firms with their own developers: a structured rollout of Claude Code with the repository boundaries, review gates and adoption measurement a regulated team needs.
Claude Code → Separate engagement · customer-facingA different shape of work: your products answering your customers inside their AI assistant, with the disclosure, audit trail and complaints handling a regulator expects. For insurers, MGAs and brokers.
Apps inside ChatGPT and Claude → Optional · platform subscriptionDefine the path the work should follow, record the path it took, show the difference. Our software, sold on subscription and implemented alongside the workflows we build. The AI Workpaper is what you keep, and what a reviewer reads. It covers the workflows and tools connected to it.
AI Process Assurance →A separate agreement, scoped to the workflows you have in production. It is never switched on by default, and the build does not depend on it.
A workflow that nobody looks after drifts. Documents change shape, a supplier renames itself, a connector's permissions get tightened by someone in IT, the model version moves. Ongoing operations is the arrangement under which the people who built your workflow keep it working and make it better.
Watching the workflows in production, fixing what breaks, refining prompts and tools as the work shows where they fall short, keeping connectors current, updating the evaluation sets, and reporting usage and cost back to you.
Output quality against the evaluation set agreed at handover. Exceptions raised by the controls, and whether they are being cleared. What each workflow costs to run. Connector health and permissions. Whether the people who own the workflow are still using it.
Anything that stops the workflow or produces a wrong output comes first. Cost and exceptions next. Improvements are agreed with the workflow owner on a regular review, at a cadence set in the agreement, rather than arriving unannounced.
Fixing, tuning and extending an existing workflow within its agreed scope is maintenance. A new workflow, a new system connection, or a change to the controls model is additional implementation, and is scoped and quoted on its own.
Optional. Most firms do not need it before an implementation. If you already know the workflow, start with that.
Some questions are bigger than a workflow. What does AI do to our margins and staffing? Do we build, buy, partner or wait? Claude, Copilot, our existing platform, or a vertical vendor? Those are board and owner decisions, and they get a different service: founder-led advisory, fixed-price and scoped to the decision, delivered as a briefing, a two-week decision sprint, a retained arrangement or a diligence memo.
We work out where to start in the free conversation, then agree the detailed design with you before building. Advisory is for the decision above that.
A platform or operating-model decision, made before budget is committed. Details →
One workflow designed, built and handed over under one fixed fee, with its controls model and run record. Details →
Kept working and improved under a separate agreement; the next workflow scoped and quoted on its own. Details →
Add the platform to keep a business-readable record of what the AI did, the evidence it used and who signed it off, retained as AI Workpapers. Subscription plus implementation, and not required for anything above. Details →
Most insurance and financial-services work clusters around document-heavy review and preparation. These are the shapes we know best; the industry pages show one illustrative example each.
The same four stages for every implementation. Fixed fee in NZD plus GST, quoted after the first conversation; platform licences and usage are yours, paid to the provider.
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.
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.
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.
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.
A free introductory conversation is enough to know whether we can help and what it would cost.
With one workflow. A free introductory conversation covers the work, the systems it touches and who owns it, which is enough for a written scope with a fixed fee. The engagement opens with the detailed design, then builds that workflow into production. Ongoing operations and further workflows follow only if you want them.
The detailed design, the build, the connections to the documents and systems the workflow needs, the governance and controls model for that workflow, the run record that model calls for, training for the people who own it, and handover. Claude Code enablement, customer-facing apps and strategic advisory are separate engagements.
It is a separate agreement, not part of the build. It covers watching the workflows already in production: output quality against evaluation sets, exceptions raised by the controls, run cost, connector health, and the prompt and tool refinements that follow. A new workflow or a new system connection is additional implementation and is quoted on its own.
We implement Claude, Microsoft Copilot and ChatGPT. We recommend one in the first conversation based on what you already license, where your documents live and what the workflow needs. Licences are your own, under your agreement with the provider, and sit outside our fee.
No. AI Platform Advisory is for boards, owners and executive teams with a platform or operating-model decision to make first. If you already know the workflow, start there.
No. Most of our work is implementing AI directly into our clients’ existing platforms and systems, and those engagements stand on their own. AI Process Assurance is an optional subscription for firms that want a retained, exportable record of what the AI did, compared against the process it was meant to follow. Every Assurance deployment includes an implementation project, because the platform is not self-service.