Case study · Regulated & Compliance-Heavy Workflows
Insurance: Agentic travel claims processing, with control
Led and delivered by founder Kannan Durairaj, prior role. Client identity withheld.Challenge
Travel claims arrive in bursts, driven by disruptions, cancellations, lost baggage and overseas medical events. Manual document checks, data entry and follow-ups for missing evidence put pressure on claims teams. The insurer needed more capacity without proportionate hiring, while maintaining consistent assessments and human accountability for consequential decisions, particularly declines.
Approach
Redesign the workflow, not just the technology. Cross-functional teams redesigned the claims journey from lodgement to payment or decline, defining how claims officers, agents and automation work together.
Match the tool to the task. Agents extract information, identify missing documents and prepare evidence-based recommendations. Deterministic rules check coverage and calculate benefits, while claims officers approve payments and decline letters.
Build in control and oversight. Behavioural controls, evaluations and monitoring support a traceable workflow. Risk-based routing directs complex claims and declines to claims officers for review and approval.
Outcome
Missing information is identified at lodgement, complex claims and declines are routed to claims officers, and a 32% increase in claims capacity helps absorb demand spikes without proportionate hiring, with claims officers working alongside governed agents.
- Customer experience: Missing information is identified at lodgement, reducing avoidable back-and-forth. Complex claims and declines are routed to claims officers for review and approval.
- Frontline capacity: A 32% increase in claims capacity helps absorb demand spikes without proportionate hiring, with claims officers working alongside governed agents.
- Operational value: Agents prepare evidence and recommendations, while deterministic rules check coverage and calculate benefits. Claims officers retain approval of payments and decline letters.
Lodgement (FNOL)
- Before: paper and process
- Customer submits a form with scanned receipts and documents, and an officer rekeys the details
- After: governed Human-AI workflow
- An agent reads the claim and documents, extracts the details and classifies the claim type
Completeness
- Before: paper and process
- Missing documents found late, followed by emails back and forth
- After: governed Human-AI workflow
- Gaps identified at lodgement and requested straight away
Coverage check
- Before: paper and process
- Officer manually checks policy dates, destination, limits and excess
- After: governed Human-AI workflow
- Deterministic rules check cover, limits and excess, consistently every time
Assessment
- Before: paper and process
- Officer reads every document against the policy wording
- After: governed Human-AI workflow
- An agent assembles the evidence and prepares a recommendation with its reasoning
Decision
- Before: paper and process
- Varies by officer and workload
- After: governed Human-AI workflow
- Routed by risk and value; declines and complex claims go to a claims officer for approval
Payment or decline
- Before: paper and process
- Manual calculation and letters
- After: governed Human-AI workflow
- Benefit calculated by rules; the officer approves the payment or decline letter
| Stage | Before: paper and process | After: governed Human-AI workflow |
|---|---|---|
| Lodgement (FNOL) | Customer submits a form with scanned receipts and documents, and an officer rekeys the details | An agent reads the claim and documents, extracts the details and classifies the claim type |
| Completeness | Missing documents found late, followed by emails back and forth | Gaps identified at lodgement and requested straight away |
| Coverage check | Officer manually checks policy dates, destination, limits and excess | Deterministic rules check cover, limits and excess, consistently every time |
| Assessment | Officer reads every document against the policy wording | An agent assembles the evidence and prepares a recommendation with its reasoning |
| Decision | Varies by officer and workload | Routed by risk and value; declines and complex claims go to a claims officer for approval |
| Payment or decline | Manual calculation and letters | Benefit calculated by rules; the officer approves the payment or decline letter |