Read the repair order, the technician's story, the parts used and the coverage in force, then assemble a claim file — the operation stays the manufacturer's and an administrator submits it.
Read the repair order as written — the concern, the technician's story, the parts, the sublet and the time punches.
02
Pull the vehicle's own record — in-service date, odometer at write-up, terms and campaigns open against the VIN.
Reason
03
Match the story to the manufacturer's published labour operation and the allowance already set against it.
04
Read the narrative for the concern, the cause and the correction the manufacturer's procedure asks for.
05
Check parts, sublet invoices and prior-approval numbers against what that operation is meant to carry.
Decide
06
Hold a file where the documentation, the coverage or the authorisation would not stand up in an audit.
07
Route added operations, straight time and repeat visits to the service manager who has to approve them.
Out
08
Present a claim file with the gaps named, for a warranty administrator to correct and submit.
09
Keep the repair order as it arrived, the gaps raised and every correction the administrator made.
→Product statement
The agent assembles and checks. The labour operation and its allowance belong to the manufacturer, the submission carries a named administrator, and coverage is the manufacturer's to decide.
Example workflow
One repair order, end to end
AgentHuman
1Repair order closedA warranty-flagged repair order arrives from the DMS with its lines, parts, sublet and time punches
2Vehicle and coverage readVIN, in-service date, odometer at write-up, the terms that apply and the campaigns open on that vehicle
3Story matched to an operationThe technician's words matched to a published labour operation and the allowance it already carries
4Documentation checkedConcern, cause and correction, the signed authorisation, parts retention, sublet and any prior approval
No human action required
Stages 1 to 4 run without a person in the loop — the coverage read, the operation match and the documentation check all finish before anyone is asked to look. Nothing has left the dealership at the end of them.
5DecisionSplits on whether the file would hold up if it were audited
File would hold up
Reaches the administrator ready to check and send.
File would not hold up
Held with the gap named, for a person to close.
Warranty administrator
Reads the repair order against the operation, the allowance and the coverage, closes the gaps and submits the claim under their own name.
Submit · Correct · Return to the technician
Submitted — handed back▼
6Claim file assembledWritten to the claim draft in the DMS; the submit action and the administrator's name are not the agent's
7Outcome evaluatedAdministrator edits, manufacturer rejections, resubmissions and what came back as a debit months later
Debits
A chargeback landing months later is scored against what the agent said at the time.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required8 items
Submitting a claim to the manufacturer.
Deciding whether a repair is covered.
Changing a labour operation or its time allowance.
Certifying that the repair was carried out.
Automation boundaryAgent acts unaided
✓Read the repair order, the parts and the time punches as recorded.
✓Look up terms, in-service date and campaigns open against the VIN.
✓Suggest which published operation the technician's story matches.
✓Name the documentation gaps and hold the file for a person to close.
The agent writes the claim draft and the gaps it found. The submission, and the allowance behind it, are not its to write.
Authorising a repair or an added operation.
Signing off straight time or extra hours.
Deciding a vehicle is eligible for a campaign.
Accepting or appealing an audit debit.
Example output
One warranty line, annotated
Everything the agent puts forward is attached to the repair order it was written on and the terms in force at write-up.
Claim file · single warranty lineIllustrative example
Repair order
Technician's story
Coverage
Claim file
Confidence
Labour operation
One warranty line
Rough running, coil replaced
In term
Held — cause not written
88%
As published, unaltered
As receivedThe repair order as the advisor and the technician wrote it, and the terms read from the vehicle's own record.
Evidence usedPublished operation matchOdometer at write-upOpen campaign on the VIN
Why it was heldThe story names what was replaced, not what was found. An auditor reads no cause.
ActionSubmitCorrectReturn to the technician
What the score decidesConfidence ranks what the administrator checks first. It never decides submission.
Value
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value landsValue 01 – 04
Warranty-flagged repair ordersClosed in your DMS
03Match & check
Read from the published guide
Use the labour operations as the manufacturer publishes them, the allowance already set against each, the terms that apply and the campaigns open on that VIN.
01Approved path
Stop rebuilding the claim by hand
The operation, the parts, the punches and the coverage arrive already assembled, so the administrator starts from a file instead of a repair order.
02Human review
Find the gap before the auditor does
A missing cause line, an unsigned authorisation or a part nobody kept is raised while the vehicle is still on site, not at the audit two years on.
04Build an evidence trail
Retain the repair order as written, the operation matched, the terms read, the gaps raised, the administrator's edits and what the manufacturer paid — on both paths.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
DMS & repair ordersCDK Global · Reynolds · Tekion Dealertrack · Time-clock and dispatch data
Manufacturer systemsClaim portals · labour-time guides Campaign and VIN lookup
Parts & service historyParts inventory · retention records Sublet and outside-purchase invoices
Agent
Warranty-claim triage
Reads the repair order Matches the operation Holds weak files
Documents & signaturesScanned repair orders · e-signature Technician time tickets
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the claim you submit
Each control wraps the one inside it. A file clears every layer before an administrator sees it, and none of them adds an hour the manufacturer has not published.
L6 · Outermost — last line of defenceInward → L1 · closest to the model
L6Rollback / safe modeReturn claim prep to your existing process if evaluation or debit signals degrade.Roll back
L5TraceabilityRecord the repair order as written, the operation matched, the gaps raised and the edits made.Record
L4Administrator gateSubmission, the coverage question and any change to an operation stay with the people who own them.Gate
L3Documentation checkConcern, cause, correction, authorisation and parts retention are checked before a file is offered.Check
L2Published operationThe operation and its allowance come from the manufacturer's guide, never from the agent.Pin
L1Coverage readTerms are read from the in-service date and the write-up odometer on the repair order.Verify
Model coreClaim file drafted — operation matched, terms read, documentation checked and confidence
L1 – L2Keep the vehicle and the operation right
L3Decides whether the file could be audited
L4 – L5Leave the signature with a person, trail kept
L6Pulls automation back when signals degrade
How Nestack evaluates it
Evaluate the whole claim file — not only the operation it matched.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the file the administrator opens
Depth of coverage ▼
E1Final-output evaluationDid the claim the administrator sent match the file the agent built?
E2Step-level evaluationDid it read the right terms, operation and allowance for that VIN?
E3Tool evaluationDid it write to the correct repair order, line and claim draft?
E4Documentation-gap recallWere the gaps an auditor later found raised before the claim went?
E5Slice evaluationHow does performance change across franchise, model year and repair type?
E6Business outcomeWhat was rejected, resubmitted, or debited back months afterwards?
Floor — what the manufacturer paid and kept
Failure modes
Where each failure originates in the agent
Seven failure modes plotted against the five stages of the agent lifecycle. None of them submits a claim — the administrator's check and the manufacturer's own adjudication are the controls that stop them.
Agent lifecycleDirection of processing →
01 · Retrieval2 modes
WT-01
Coverage read at the wrong date
Terms judged on today's odometer, not the one at write-up.
WT-02
Signed authorisation not on file
The customer's approval never left the paper copy.
Stage gathersThe repair order, the parts, the punches and the terms
02 · Matching2 modes
WT-03
Allowance superseded last week
The operation was re-timed after the guide was read.
WT-04
Correction with no cause behind it
The story names what was replaced, not what was found.
Stage matchesThe story to a published operation and its allowance
03 · Draft / write1 mode
WT-05
Overlapping operations drafted
Two lines claim time the manufacturer pays once.
Stage draftsThe claim draft written back to the repair order
04 · Output1 mode
WT-06
Repeat visit drafted as new
A comeback loses its link to the first repair.
Stage returnsThe file the administrator checks and submits
05 · Change / Version1 mode
WT-07
Bulletin lands after the drafts
A programme change moves the rule and nothing re-checks.
Stage tracksModel, rule, labour-guide and bulletin changes
Sev-1 · a debit lands months laterSev-2 · the file is built on wrong termsSev-3 · the file goes back to the shop
A new platform is what the guide has least to say about
A first-year platform changes its operations and allowances faster than anything the agent has read, and a high-voltage repair may have almost no comparable history behind it. Nestack reports performance by slice, not only in total.
Slice performance — reported separately, not only in aggregateIllustrative example
Slice
Failure rate
Lift
Lift vs. threshold
Status
First model year, new platform
6.5%
3.4×
Review
EV high-voltage repairs
4.9%
2.6×
Review
Thin technician narratives
3.4%
1.8×
Watch
Single-line routine repairs
1.3%
0.7×
Normal
Bar: rework-rate lift — corrected before submission or debited after · scale 0–4.0× · tick at 2.0×2 of 4 slices over threshold
Evidence-linked improvement
A debit in March scores a claim from November
Every file an administrator had to correct, and every line the manufacturer later took back, is scored again against what the agent put in front of them at the time.
Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect
Administrator edits, manufacturer rejections, or a debit on an audited sample.
02Diagnose
Separated into terms read, operation matched, documentation gap, approval routing and repeat-visit linking.
03Improve
Match rules and gap checks change under your fixed-ops change control, with a named approver.
04Verify
Re-run on stored repair orders from that franchise, model year and repair type, including the debited ones.
05Learn
The debited line is added to the pre-submission checks, and the reason goes into your warranty office's own guide.
Learn → DetectThe return edge. A debit closes when the check that would have caught it is running on the next repair order — not when the appeal is filed.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, records and terms, matching and checks, evaluation, integration, then supervised submission and handover.
WorkstreamWeek 1Week 2Week 3Week 4Week 5Week 6
01Warranty-claim discovery and boundaries.
02Coverage terms and campaign-lookup mapping.
03DMS repair-order and time-punch access.
04Labour-operation guide and allowance sync.
05Repair-order and narrative normalisation.
06Story-to-operation matching logic.
07Documentation sufficiency checks.
08Approval routing to service management.
09Evaluation suite, slices and past-debit replay.
10Claim-draft write-back to the DMS.
11Audit-pack and retention-record assembly.
12Observability, deployment and Agent Care handover.
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallelFinal scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
Capability✓ in scope · — not at this tierPilotOne rooftop, one franchiseProductionProduction DMS and portal integrationAdvancedMulti-rooftop / multi-franchise
Introduced at Pilot
Repair orders read and claim files drafted✓✓✓
Operations matched to the published guide✓✓✓
Coverage and campaign lookup by VIN✓✓✓
Documentation sufficiency checks✓✓✓
Submission by a named administrator✓✓✓
Baseline evaluation✓✓✓
Introduced at Production
Approval routing to service management—✓✓
Claim-draft write-back to the DMS—✓✓
Observability and evaluation—✓✓
Audit-pack and retention-record assembly—✓✓
Introduced at Advanced
Multi-rooftop rules and enterprise controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build priceConfirmed after discovery based on rooftops and franchises in scope, DMS and manufacturer-portal integrations, claim volume, approval routing and deployment requirements.
Separate from buildBuild pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bringWe build with it
01Your manufacturers' warranty policy and labour-time guides→Labour-operation guide and allowance syncWeek 1
02Coverage terms, in-service data and campaign lookup→Coverage terms and campaign-lookup mappingWeek 1
03Access to DMS repair orders, parts and technician time punches→DMS repair-order and time-punch accessWeek 2
04A year of closed warranty repair orders as they were written→Repair-order normalisation and the matching baselineWeek 2
05Who approves added operations, straight time and repeat repairs→Approval routing to service managementWeek 3
06The debits and rejections you have taken, with the claims behind them→Evaluation suite, past-debit replay and failure-mode testingWeek 4
07Named warranty administrators and a service manager→Claim-draft write-back, then supervised submissionWeeks 5–6
Nothing else is requiredDeployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Phases are drawn over the weeks they actually occupy. Week 5 carries both the replayed debits and the first files your administrators submit from.
PhaseW1W2W3W4W5W6
DiscoveryW1
BuildW2 – W3
EvaluateW4 – W5
Pilot & LaunchW5 – W6
Week focusW1Warranty-claim discovery, coverage terms and campaign lookupW2Repair-order, parts and time-punch access, then normalisationW3Story-to-operation matching and documentation sufficiency checksW4Evaluation suite, past-debit replay and approval routingW5DMS write-back, first supervised files and targeted correctionsW6Your administrators submit from the file, then Agent Care starts
Reading the bandDocumentation checks are built in week 3, before any file is drafted for submission in week 5. A file your own auditor would debit should not reach an administrator.
At the end of W6Files have run alongside your existing claim prep and been submitted by the administrators who sign them, with the operation and the allowance still the manufacturer's, then Agent Care takes over monitoring.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Automotive AI agent
Build a claim-triage agent around your warranty office.
Show us one franchise, the labour-time guide behind it and a year of closed warranty repair orders as they were written. We'll rebuild those claims against the operations in force at the time and mark which of the debits you actually took we would have raised before submission.
