Clean and confident
All checks passed and confidence is high. The expert's job becomes a fast confirmation instead of a full manual review.
Kolayz ClaimSense — warranty claim analysis
Every claim analyzed overnight. Your experts decide what matters.
Warranty claims are still reviewed one by one: photos opened, serial numbers compared, service history checked. ClaimSense does the repetitive part of that review every night — so in the morning, every claim arrives pre-assessed, reasoned, and sorted by what actually needs an expert.
The morning after
Every claim lands in one of three trays.
ClaimSense recommends, it never decides. The point is not to replace your experts — it is to concentrate their attention on the claims that genuinely need judgment.
Fixable gaps
A document is missing or a photo is unusable — correctable issues. The expert requests exactly what is missing from the dealer, with the gap already named.
Real judgment calls
The system is unsure, or found something substantive. These are the claims that deserve expertise — and now they get it first, with the evidence laid out.
The nightly run
From raw claim to reasoned recommendation.
Every step is recorded. For any claim, you can see which checks ran, which were skipped for missing input, and how each photo influenced the outcome.
01
Claim intake
New claims are pulled from your CRM or dealer system each night — lines, amounts, fault codes, attachments — with no manual data entry.
02
Evidence gathering
Photos are classified by AI, unit history and maintenance records are retrieved, and prior claims on the same component are surfaced.
03
Policy checks
A configurable battery of checks — 13 in the current deployment — each returning a result, a confidence, and flags. Missing data, an unreadable photo and a genuine mismatch are three different outcomes with three different policies, and a gap on our side never counts against the claim.
04
Reasoned recommendation
A recommendation with a one-line rationale, backed by the full evidence chain — every photo, every check, every flag, one click away.
Before you rely on it
Trust is measured, not promised.
Would this system decide the way your experts do? ClaimSense answers that question with data before it influences a single real claim.
Shadow mode first
ClaimSense starts by analyzing without writing anything back to your systems. Recommendations are produced and compared — operations continue exactly as before.
Agreement, quantified
Already-closed claims are re-analyzed and the system's recommendation is placed next to the expert's real decision. Contradictions are surfaced as their own category — that is the number that matters.
Experts tune it
On every claim, the expert can agree or disagree with a note. That feedback accumulates and calibrates thresholds against your organization's real judgment.
- Every photo is stored with its classification and the check it influenced — a complete evidence chain
- Cautious by design: when unsure, it routes to a human instead of guessing
- Management dashboards show automation rate, top failing criteria, and dealer and product breakdowns
- Four roles, two-factor sign-in for the ones that can change anything, and an audit trail over every login, setting and approval
Built for service organizations
Made for brands that stand behind their products.
ClaimSense is live at a commercial-vehicle OEM: claims are pulled from SAP C4C every night, and the warranty team finds them pre-assessed the next morning. The same engine fits any brand with a dealer or service network and a warranty book to protect.
Plugs into your systems
Integrates with your CRM and dealer systems over standard APIs — SAP C4C in production today. Claims flow in, recommendations flow back.
Data protection built in
KVKK and GDPR shaped the build, not a later review. You choose the region it runs in; claim photos are deleted when their retention window closes while the reasoning behind each decision is kept, and no personal or vehicle data reaches a customer-facing output.
Your cloud or ours
Runs in your own AWS account or on the AWS we operate — Frankfurt today, behind a load balancer and WAF, with encrypted backups and managed secrets. The AI layer runs on Kolayz Anka, our private gateway — prompts are never stored.
Your policy, not ours
Criteria, weights, thresholds and the fault-code catalogue are set in the panel, not in a release. Every change needs a second person to approve it, every change is versioned, and each recommendation carries the configuration version that produced it.
Sees patterns across claims
Dealer and model breakdowns reveal what single-claim review never can: the dealer with recurring missing documents, the component failing more than it should.
Beyond vehicles
Trucks and buses today; the engine generalizes to any serial-numbered product with photos, service history, and a warranty policy — from machinery to appliances.
See it on your claims