Sampling was a compute limit. It isn’t one any more.

Your warranty team audits a sample of dealer claims and pays the rest on trust. Actuora reads every claim, including the technician’s narrative, and joins them across dealers, VINs, technicians, parts and payments.

Run the Coverage Check ↗ What it finds ↗

Dealer claim population — one bar per thousand claims Inspected today 4%

POPULATION / WARRANTY / DEALER CLAIMS Sampled · 4% audited

Tracks

Reading Full population Six build lots · failures reported as they arrive

> coverage --all-claims

FIG. 01 — FOUR TRACKS, ONE CLAIM RECORD DEFECT, WARRANTY, FRAUD, RECOVERY — SELECT ONE ON THE LEFT

01 The sample

The cause of a failure is written in the one field nobody reads.

Complaint, cause, correction — three free-text boxes where the technician says what actually happened, and a labour code that says what fit the dropdown.

Warranty systems adjudicate on coded fields: labour operation, causal part, failure code. Those are what a dealer had time to select. The narrative is where the truth sits — the handwritten annotation, the note that the customer had been back twice, the mention of a part that was never billed.

And a perfect reading of one claim still tells you nothing about collusion. A dealer inflating labour hours looks unremarkable alone. Four dealers in one region using the same unusual repair sequence, on the same model, in the same month, is not a claim-level pattern at all. It exists only between claims — which is why no sample, however carefully drawn, can contain it.

By the numbers

Of the claim population read, narrative included.

Vehicle manufacturers running Actuora agents in production today.

0 hr

To produce an adjudication report. Measured before-state: one week.

Claim records sent to a public model API.

02 What full coverage produces

One capability. Three consequences.

Reading the whole population gives you throughput. Joining the whole population gives you findings.

FIG. 03 — many observations, written many ways, one cause

01Defect

Defect Intelligence

Solves: the same defect is written up a dozen different ways across FIR, PCR and PIC reports, so nobody can see that the observations share one root cause.

Every observation in every report is read and classified OK or NOT OK, with the reasoning retained. The NOT OK population is then analysed across the fleet to derive root cause — not case by case, but all of it at once.

Every observation classified OK / NOT OK, with the basis recorded
FIR, PCR and PIC read as one population, not three separate systems
NOT OK observations grouped to candidate root causes
Fleet-wide RCA rather than one investigation at a time

Owner — Quality & Engineering

FIG. 04 — four independent checks, one of them catches it

02Warranty

Warranty Intelligence

Solves: altered and duplicated invoices clear warranty audit because no team can forensically examine every supporting document.

Every invoice and supporting document is passed through a bank of independent detectors, each looking for a different species of document fraud. A document that trips any of them is flagged with the detector that caught it.

Duplicate and near-duplicate invoice detection
Alteration and tampering checks on the document itself
Price and quantity anomalies against the claim
Vendor and reference validation

Owner — Warranty & After-Sales

FIG. 05 — the pattern is not in any claim, it is between them

03Fraud

Fraud Intelligence

Solves: collusion is invisible to claim-by-claim review, because every individual claim is unremarkable. Adding reviewers does not help.

Claims are joined across dealers, VINs, technicians, parts and payments, and the population is tested for co-occurrence rather than for outliers.

Cross-dealer collusion and coordinated repair sequences
Duplicate claims under altered part numbers or VINs
Repairs billed but not evidenced anywhere in the narrative
Systematic labour inflation, sublet abuse and policy leakage

Owner — CFO & Internal Audit

03 Between the claims

Some of what we do saves effort. Some of it finds what no one could.

Adjudicating faster is a cost story with a headcount answer. Surfacing a collusion ring no reviewer could have seen is a different line on the P&L.

Dealers × repair operations

FIG. 02 — one cluster, four dealers Cluster surfaced

A single inflated labour line is invisible. The same inflation, on the same operation, at four dealers who share a parts supplier, is a finding — and it exists only once every claim is in the same place.

Claims a warranty audit team can reviewA sample

Claims an Actuora agent readsAll of them

Where a collusion pattern livesBetween claims

Who keeps the scoring harnessYou do

04 How we engage

Three steps. You can stop at any of them.

The first needs no contract, no NDA and no vendor onboarding, because it never leaves your building and does not need us to run it.

The Coverage Check

Free · offline · your IT runs it

One signed file your own team runs offline against a single export of closed claims. It reports what share of your claim population is actually inspected today, and where your own warranty engineers disagree with each other on failure cause.

Current inspected share of the claim population
Agreement rate between your own reviewers on the same claim
Agreement of our read against your historical failure coding
No network calls, no telemetry, nothing transmitted

No feeRuns on CPU

The Full Read

Fixed fee · fixed clock

Four weeks across your real claim corpus. The deliverable is a measured number on a jointly sampled, jointly signed held-out set. If the numbers do not justify a deployment, we say so.

Field-level accuracy on causal part and failure cause, with methodology
The failure envelope — claim types we refuse to adjudicate, and their frequency
Value of suspect claims your own investigators confirm
The versioned eval set and scoring harness, handed over as your property

FixedCredited against build

Standing Coverage

In production · inside your network

Every claim read as it lands, with an accuracy floor agreed in writing before we start — expressed as a fraction of the measured human ceiling, never as an absolute. If it is not met, we keep working at our cost until it is.

Deployed in your tenant or on-premises alongside the warranty system
Quarterly re-scoring against the frozen eval set
Drift monitoring and a model-change register
A named threshold at which we tell you to switch it off

AnnualVolume floor applies

05 How we're measured

We publish the ceiling before we quote the number.

No system can be more accurate than the coding it is graded against. So the first thing we measure is not our model — it is how often two of your own warranty engineers, shown the same claim, assign the same failure cause.

The human ceiling, measured first

Two of your engineers independently code the same subsample. That agreement rate is the ceiling, and every figure we quote afterwards is expressed against it.

You own the measuring instrument

The held-out claim set and scoring harness are handed over as your property, with a written right to re-run them against any vendor — including us.

The failure envelope is published

We name the claim types we will not adjudicate — photo-only evidence, code-mixed narratives, contradictory DTC and text — and how often each occurs in your data.

A named switch-off threshold

Written into the contract: the metric at which we tell you to turn the agent off, and the reversion runbook for doing it.

Escrow from signature

Weights, adapters, scorer, container and runbook in escrow, released on insolvency, two consecutive quarters below floor, or exit for convenience.

The clock starts on access

Timelines begin at claim data access, environment sign-off and a scoped service account — not at signature. DMS integration is our scope, not your homework.

06 Plain speech

Things you will not read on this website.

Every one of these is said by nearly every vendor in this category, which is precisely why none of them tells you anything.

~~“Sovereign AI.”~~

That phrase belongs to national programmes and cloud regions, not to an application vendor. We will show you an egress diagram and the steps your security team runs to verify it.

~~“Live in 4–6 weeks.”~~

Deployment was never the bottleneck. Claim data access, DMS integration and security review are. We put a start gate in the contract rather than a range in a headline.

~~“Reduce warranty spend by 30%.”~~

A savings figure with no stated baseline is arithmetic, not evidence. We measure your current inspected share and cost per claim first.

~~“99% accurate.”~~

On whose claims, on which field, at what sample size, against what agreement rate between your own engineers? Without those five things the percentage is decoration.

~~“Not a chatbot — real enterprise workflows.”~~

This is the opening line of almost every deck in the category. If our work is real, the Coverage Check shows it and the sentence is unnecessary.

~~“Nobody else can do this.”~~

Someone else usually can. We would rather be measured than be the only one making a claim that is checkable in a single search.

07 Who we work with

Vehicle manufacturers and Tier-1 suppliers carrying a warranty reserve.

Sponsor

Head of Warranty · VP After-Sales Quality

Owns adjudication, dealer relationships and the audit sample. Has to defend both the payout and the coverage.

Economic buyer

CFO · Financial Controller

Owns the warranty reserve — an audited line whose accuracy currently rests on extrapolation from a sample.

Beneficiary

Head of Quality · Product Engineering

Needs failure signatures early enough to contain a build lot rather than announce a campaign.

Beneficiary

Supplier Recovery · Internal Audit

Needs evidenced, claim-level support for recovery against suppliers, and for anything escalated as fraud.

08 Start here

Your closed claims are an answer key. Let us be graded against them.

Send us nothing. We send you one file, your own IT runs it offline against a single export of closed claims, and you keep what it finds either way.

hello@actuora.ai ↗ How we’re measured ↗