Insurance · General Lines · East Africa

Motor-claim turnaround from 11 days to 36 hours.

A regional general insurer processing 4,000+ motor claims per month on a manual workflow. Adjusters were spending 60% of their time on data entry and document chasing, not on adjudication. We built a document AI and workflow orchestration layer that changed that ratio permanently.

Document AI / OCR Workflow orchestration Auto-adjudication engine IRA compliance
Insurance claims processing — workflow automation team
Client
Regional General Insurer (Confidential)
Sector
Insurance
Duration
8 months
Year
2025

The challenge

The insurer was handling 4,200 motor claims per month. Average turnaround time was 11 days from first notice of loss to payment authorisation. The IRA's target is 5 days. Claimant CSAT sat at 38 — below the industry median. The core problem wasn't adjudicator skill; it was pipeline. Claims came in via WhatsApp, email, walk-in, and an agent portal, each feeding a different inbox. Documents — police abstract, repair estimate, photos — were chased manually, sometimes across three follow-up calls. By the time a claim reached an adjudicator, 70% of the work was already done; the adjudicator just couldn't see it.

4,200
Motor claims/month
11 days
Average claim turnaround
38
Claimant CSAT (out of 100)

What we built

We started with a six-week process audit — sitting with adjusters, timing every step, and mapping every exception. The audit surfaced that 41% of delay came from a single source: the police abstract. Claimants had to physically collect it, scan it, and email it. We worked with a legal partner to get the insurer recognised as an authorised third-party requester — the abstract is now fetched electronically in under 4 hours for 78% of claims.

Document AI handles the rest of the intake pipeline. Claims submitted via any channel are normalised into a single queue. The AI extracts key fields from repair estimates, vehicle photos, and accident reports, cross-referencing against the policy record and flagging anomalies (mileage inconsistencies, repair costs above book value for vehicle age, duplicate claim patterns). By the time a claim reaches an adjudicator, the jacket is pre-populated and the AI has already scored the claim for fraud risk and complexity.

Simple, low-risk claims — 62% of volume — are now auto-adjudicated within the engine's rule set, with a human review only on the system-generated payment notice. An adjudicator sees them for 90 seconds, not 90 minutes.

Insurance workflow automation — document processing

IRA compliance and explainability

The Insurance Regulatory Authority requires that every claim decision be explainable to the claimant on request. The system logs the specific rule path and evidence set for every auto-adjudicated decision. Adjusters can pull a plain-language explanation in one click — critical for customer-service calls and formal complaints.

"The claims team now spends their time on the hard cases, not on chasing paperwork. The pipeline that used to clog at document intake now flows. And our IRA turnaround stats have gone from a liability to a selling point."

Approach

Process audit first, automation second.

01

Process audit

6-week deep dive: every step timed, every exception mapped, every delay traced to a root cause.

02

Document AI

Multi-channel intake normalised into one queue; extraction of all key fields across claim types.

03

Auto-adjudication

Rule engine for straightforward claims; AI fraud scoring and complexity triage for all others.

04

Compliance layer

IRA-compliant explainability logging for every automated decision; full audit trail.

Outcomes

11 days to 36 hours. 62% auto-adjudicated.

11d → 36h
Motor-claim turnaround
62%
Auto-adjudicated
+19
CSAT points
8 mo
Audit to full deployment

Claims taking too long?

We start with a process audit — not a demo. We find out where the time actually goes before we propose a solution.