Airclerk charters the renewal workflow, gates the critical decisions, and produces the evidence file behind every approved outcome - pulling policy data, documents, communications and claims history into one governed path. Claude handles the analysis, drafting and explanation. Your team reviews and decides.
Because the process is chartered, gated and evidenced, you can scale with renewal volume instead of headcount - and your team gets back the work that matters: expertise, judgement, client relationships, decisions.
It is AI Process Assurance applied to commercial insurance renewals. Airclerk charters the expected path of a renewal - gathering context, enriching the file, surfacing what changed, preparing outputs, routing approvals - records what the AI assistant actually does at each step, and produces an audit-ready workpaper behind the outcome. Claude is plugged in for analysis, drafting and explanation; Airclerk holds the workflow state, the integrations and the evidence; humans review and decide.
For brokers, MGAs, insurers and underwriting teams managing high-volume commercial renewal books across email, documents, spreadsheets and legacy systems - where state, integration and audit have become the limiting factor.
Not for: personal lines, retail consumer insurance, or replacing your broking platform, policy admin system or document store. Airclerk sits above them as the assurance and action layer.
Claude is becoming the AI platform businesses trust. It is a brilliant reasoner inside a conversation - but a commercial renewal lives across weeks, people and systems. Airclerk is the layer that gives Claude everything it needs to be useful in that workflow.
Commercial insurance renewals run across inboxes, documents, spreadsheets, conversations and disconnected systems. The Airclerk app pulls that scattered work onto a single stateful workflow. Claude supports the reading, comparing, drafting and explaining at each step. Your team makes the decisions.
Every stage carries four things: the expected path the charter requires, what actually happened, the evidence produced, and any approval gate. That is AI Process Assurance, stage by stage.
120, 90, 60, 30 days out - the app opens the renewal file and kicks off the preparation workflow on its own. No human has to remember.
Policy details, prior documents, claims context, communications history, structured renewal inputs - assembled into one review-ready file.
Gaps identified. Follow-ups drafted. Information requests coordinated. Review-ready outputs prepared in the format your team and your market expect.
Your team reviews, edits, approves or escalates. The app tracks follow-ups, routing and outstanding work so renewals keep moving - in parallel, at the volume your book actually needs.
A commercial renewal isn't a chat. It's weeks of work, multiple people, hundreds of artefacts, dozens of decisions. The Airclerk app is what ties it together.
An underwriter's cowork session on Monday, the broker's chat on Thursday and the reviewer's questions on Friday all read from - and write to - the same renewal state. No re-explaining context.
Which documents have come back. Which questions are outstanding. Who's reviewing what. What the market is waiting for. Visible to your team and to Claude - long after any individual chat ends.
Hundreds of renewals at different stages, all running at once. Capacity scales with volume, not headcount.
Claude can read what's put in front of it. It cannot reach into your broking platform, policy admin or claims system. The Airclerk app does - so Claude can act, not just advise.
The app sits above your systems of record as an assurance and action layer. It connects data, documents and workflow to the conversations your team has with Claude - without replacing the core systems your business already runs on.
Where vendors expose a proper API, we integrate. Where they don't - and a lot of insurance software does not - browser-driving and web-agent techniques are a workable bridge until they catch up.
The work product looks like the work product your team already produces - just prepared faster, with the evidence attached.
A standalone chat is not enough of an audit trail for a regulated renewal workflow. Airclerk's Semantic Audit records the work as the assistant does it: what it read, what it drafted, what was approved, who approved it, and which source documents supported each output - so the same workflow that produces a renewal also produces the AI Workpaper a regulator expects.
Each tool call, document read, system write, draft and approval the assistant makes is recorded as it works - with timestamp, identity and source provenance. Outputs cite the documents they came from.
We shape what Claude can read, write or trigger per workflow stage - through tool availability, stage-specific instructions and approval gates, rather than trusting a prompt alone. Consequential actions like binding cover or sending to market sit behind explicit human approval, so the model's behaviour is influenced where it can be and gated where it must be.
Routed approval queues per workflow stage. The model proposes, humans dispose - and both sides of that exchange are recorded against the renewal file.
Per-workflow budgets, token and tool-call limits, model-tier routing and usage alerts. Finance sees what AI is costing per renewal before the invoice arrives.
The app passes the user's identity and permissions through to every backend call. No bypassing the access model you already enforce.
Outputs that can't be traced back to the source documents and the reasoning path do not leave the review queue. Operating discipline, not a model feature.
The Airclerk MCP, and the audit trail it writes, can be hosted by Airclerk with per-tenant isolation, or deployed single-tenant into your own Azure tenant on request. Your evidence and controls live inside the cloud boundary your risk and compliance teams already govern - not a third party's.
Airclerk is deployed as a governed renewal workflow for teams with meaningful commercial renewal volume.
A 30-minute conversation to confirm fit - your renewal volume, the systems you run on, your Claude position and where state and audit have become the limiting factor.
Two weeks. We confirm scope, map your systems, define the controls and produce a board-ready implementation plan for the first production workflow.
The Airclerk app is implemented as the governed renewal workflow - connected to your systems, your Claude environment and your governance model. Run alongside your team while they grow into ownership.
For teams that want us to run the workflow day-to-day while their operations team scales into it. See Managed Operations.
AI Process Assurance applied to commercial insurance renewals. It charters the renewal workflow and runs the operational work - state, integration, triggers, approvals - records what actually happened at each step, and produces an audit-ready workpaper, with Claude plugged in for analysis, drafting and explanation.
Your team works inside Claude (or Claude Enterprise). The Airclerk MCP exposes the renewal file, backend systems and workflow state to Claude as tools. Every Claude action that touches a renewal goes through Airclerk - so state, evidence and approvals are captured by the platform rather than left in a chat history.
Two layers, two answers. The Claude conversations themselves run inside your Claude environment (Claude Enterprise or Team) under Anthropic's enterprise terms - your prompts and outputs are not used to train Claude. The Airclerk MCP - which holds renewal workflow state, brokers calls to your backends and writes the audit trail - can be hosted by Airclerk on our multitenant infrastructure (Vercel and Supabase), with per-tenant logical isolation, or deployed single-tenant into your own cloud - including your Azure tenant - on request. Backend integrations are called using credentials you provide, with the end user's identity passed through. Full security posture, sub-processors and ISO 27001 / SOC 2 Type 2 progress on the Trust page.
No. Airclerk sits above your systems of record as an assurance and action layer. Your BMS, PAS, claims system and document store stay where they are. Airclerk connects to them, holds the renewal workflow state, and produces the outputs and evidence your team needs.
Where vendors expose a proper API, we integrate directly via MCP servers and connectors. Where they don't - and a lot of insurance software does not - browser-driving and web-agent techniques are a workable bridge. We've connected to broking platforms, policy admin systems, claims and case-management systems, document stores, shared mailboxes and submission inboxes.
No. Consequential actions - binding cover, sending to market, releasing funds - sit behind explicit human approval. Airclerk prepares the work, surfaces the evidence and routes the decision; your team reviews, edits, approves or escalates.
Every tool call, document read, system write, draft, approval and escalation is logged against the renewal file - with timestamp, identity and source provenance. Outputs cite the documents they came from. The same workflow that produces a renewal produces the evidence a regulator expects.
Brokers, MGAs, insurers and underwriting teams managing high-volume commercial renewal books. It is not designed for personal lines or retail consumer insurance, and it does not replace your broker management system, policy admin system or document store.
Claude alone is a brilliant reasoner inside a single conversation. It does not hold renewal state across weeks, trigger work on a 120/90/60/30-day schedule, reach into insurance backend systems, or keep an audit trail outside the chat. Airclerk does those things - so the same conversation can be useful to a regulated renewal workflow rather than just a smart assistant.
If you're running commercial renewals at the volume where state, integration and audit have become the limiting factor, this is the conversation.