GroveAI
Case Study

Compliance Monitoring Agent for Specialist Insurance

Balfour Underwriting needed continuous compliance oversight across thousands of policy documents. We built an AI agent that monitors, audits, and flags regulatory issues in real time — before they become problems.

Client: Balfour UnderwritingIndustry: InsuranceDuration: 10 weeks

Results

Impact delivered

100%

Policy coverage

Every policy now reviewed, up from 15% sampling

81%

Compliance review time reduction

Time per policy review reduced from 45 minutes to 8.5 minutes

340+

Issues caught pre-binding

Compliance issues identified and resolved before policy issuance in the first year

Zero

Regulatory audit findings

Clean audit results in the first annual regulatory review post-deployment

The Challenge

What they faced

Balfour Underwriting, a Lloyd's of London syndicate specialising in professional indemnity and directors' & officers' insurance, processes over 6,000 policies per year. Each policy must comply with an evolving landscape of FCA consumer duty requirements, Lloyd's minimum standards, and market-specific regulations. Their compliance team of six was conducting manual reviews on a sampling basis — checking roughly 15% of policies each quarter. This left significant blind spots. Two near-miss regulatory findings in successive audits made it clear that a sample-based approach was no longer tenable. The board demanded continuous, comprehensive compliance monitoring without tripling the team's headcount.

Our Solution

How we solved it

We developed a custom AI compliance agent that continuously monitors every policy document against a structured regulatory knowledge base. The agent ingests new and amended policies as they're created in the underwriting platform, checks them against current regulatory requirements, Lloyd's standards, and Balfour's own internal guidelines, and generates detailed compliance reports. Issues are categorised by severity and routed to the appropriate compliance officer. The agent also monitors regulatory updates from the FCA and Lloyd's, automatically flagging policies that may be affected by new requirements. It operates as a persistent background process, providing the compliance team with a live dashboard of their regulatory posture.

Approach

Step by step

01

Regulatory knowledge base construction

Built a comprehensive, structured knowledge base covering FCA Consumer Duty, Lloyd's minimum standards, market bulletins, and Balfour's internal compliance policies.

02

Agent architecture and workflow design

Designed a multi-stage agent pipeline: document ingestion, clause-level analysis, regulatory mapping, issue detection, severity classification, and human routing.

03

Underwriting platform integration

Integrated with Balfour's Eclipse underwriting system to automatically ingest new policies, endorsements, and amendments as they are created or modified.

04

Regulatory update monitoring

Built an automated feed that monitors FCA and Lloyd's publications, parses new requirements, and updates the knowledge base — triggering re-analysis of affected policies.

05

Compliance dashboard development

Created a real-time dashboard showing compliance status across the entire book, drill-down issue views, trend analysis, and audit-ready reporting.

Before Grove's agent, we were essentially hoping that our 15% sample was representative. Now we have eyes on every single policy, every single day. The agent caught 340 issues in its first year that we would have missed entirely. It's not an exaggeration to say it's transformed our regulatory posture.

Katherine Ellwood

Head of Compliance, Balfour Underwriting

Technology

Stack used

Claude APILangGraphPythonFastAPIPostgreSQLElasticsearchReactDocker

FAQ

Frequently asked questions

The agent monitors FCA and Lloyd's publication feeds automatically. When new guidance or requirements are published, it parses the changes, updates its knowledge base, and re-analyses any policies that may be affected — alerting the compliance team to potential issues.

No. The agent identifies and flags potential issues but never makes binding compliance determinations. All flagged items are reviewed by a qualified compliance officer. The agent is a tool that augments human judgement, not a replacement for it.

We conducted a three-month parallel run where the agent's findings were compared against manual compliance reviews. The agent achieved a 97% agreement rate with senior compliance officers, with the 3% variance predominantly consisting of edge cases that the agent flagged conservatively.

Want similar results?

Book a free strategy call and we'll assess how AI can deliver impact for your business.