This post is part of a series based on discussions with senior leaders in insurance, including Arno de Wever, Head of Commercial P&C Insurance at Amazon Web Services (AWS). Drawing on his experience with COOs and underwriting leaders across the London Market and beyond, we explore what comes after Intelligent Document Processing (IDP) — and why multi-agent workflows are becoming the next big step for insurance operations.
Why IDP isn’t the final destination
Over the past few years, IDP has become a proven productivity booster. By automating the classification and extraction of data from PDFs, spreadsheets, and emails, insurers have achieved dramatic efficiency gains. Some report a 75% reduction in processing time and data accuracy above 95%.
But IDP is now the baseline. Most COOs have already seen these gains, and the question has shifted: how can AI go further, turning raw documents into business-ready decisions?
From single-task tools to multi-agent workflows
Multi-agent workflows take AI beyond point automation. Instead of a single tool extracting text, multiple AI “agents” collaborate on the end-to-end underwriting process.
A simple example:
- One agent extracts key data from the submission.
- Another checks the data against underwriting guidelines.
- A third enriches it with external data, such as credit or industry benchmarks.
- A fourth drafts a risk memo or pre-populated quote.
The result is not just faster processing, but a consistent, explainable decision pipeline. Underwriters receive a near-complete package, review it, and make the final call.
How multi-agent underwriting workflows help COOs cut costs and boost capacity
The business impact is more than incremental:
- Capacity without headcount. Underwriters can process significantly more submissions because they no longer start from scratch.
- Consistency and compliance. Every risk is assessed against the same guidelines, with audit trails for compliance teams and boards.
- Speed to quote. Brokers and clients get answers in hours instead of days, improving win rates.
- Portfolio insight. With richer data flowing through each case, COOs get better visibility into risk trends and product demand.
Overcoming barriers to adoption
Moving from IDP to multi-agent workflows raises legitimate concerns. Legacy systems don’t always integrate easily. Governance models are still being built in many firms. And there are cultural questions: Will underwriters trust AI-generated memos? Will workers’ councils see this as a threat?
The way forward is to design adoption carefully. With human-in-the-loop review, underwriters always have the final say. With audit trails, compliance teams can trace how every recommendation was made. And with incremental rollout, COOs can start small — automating submission triage or memo generation — and expand over time.
Beyond efficiency: enabling new growth
The biggest missed point is that multi-agent workflows aren’t just about doing the same work faster. They enable new business opportunities. By lowering expense ratios and increasing throughput, insurers can profitably serve risks that were previously too small or complex. One London Market insurer, after adopting AI, reduced its minimum premium threshold by a third, effectively expanding the market it could serve.
For COOs, this is the real prize: transforming operations not just to save costs, but to grow revenue.
Multi-agent workflows: the COO’s next move
IDP proved that AI can deliver operational efficiency. Multi-agent workflows show that AI can deliver operational excellence. For COOs, the next move is clear: take underwriting automation beyond document processing and into decision-making.
Do you want to see what this looks like in practice? Download our COO guide on Operational Excellence Through Proven AI or book a free consultation.
Take the next step
Discover solutions, customer success stories, and events on our dedicated Insurance page.
