Main site

Automotive AI agent · Appraisal & listing

Used-EV Appraisal & Listing AI Agent

Assemble the condition, battery-health, warranty and recall evidence on a used EV and draft the listing — with the value, the range claim and the decision to publish left to a named person.

Typical delivery
4–6 weeks
Deployment
Your DMS
Valuation
Your manager
After launch
Agent Care

What this agent does

Assembles the evidence, not the number

In

  1. 01

    Take the unit in from a trade-in, a lease return, an auction buy or a car already standing in inventory.

  2. 02

    Pull the service records, prior damage, title brands and the battery and drivetrain warranty as recorded.

Reason

  1. 03

    Record every battery state-of-health reading on file with where it came from, when, and on what tool.

  2. 04

    Run the VIN through the recall and campaign lookups your store uses and note what each returned, and when.

  3. 05

    Set out charging capability, connector and conditioning as the build record states them for this VIN.

Decide

  1. 06

    Hold back any value, offer, range figure or claim that the battery or the coverage is sound.

  2. 07

    Flag thin, stale or disagreeing evidence and say what the appraiser should verify before it is listed.

Out

  1. 08

    Hand over an appraisal file and a draft listing carrying the disclosures the store requires.

  2. 09

    Retain every reading, its source and date, what the manager changed, what was published and by whom.

Product statement

The agent gathers, dates and drafts. The value, the condition claim and the decision to publish stay with a named person at the dealership.

Example workflow

One unit, end to end

  1. 1Unit taken inTrade-in at the desk, lease return, auction buy or a car already standing in inventory
  2. 2VIN decoded and matchedResolved to one VIN and one build record before any history or reading is read
  3. 3Evidence gatheredService records, prior damage, title brands, warranty terms and the recall and campaign lookups
  4. 4Battery evidence datedEach state-of-health reading with its source, its tool and the conditions it was taken under

No human action required

Stages 1 to 4 run without a person in the loop — the decode, the retrieval and the battery evidence are finished before a manager opens the file, and nothing has been listed yet.

5DecisionSplits on how the evidence stands, and on what a listing would have to assert
Evidence agrees and is dated

Reaches the used-car manager ready to read.

Readings disagree or a claim is needed

Held for the used-car manager to resolve.

Used-car manager

Reads the file, the readings behind it and what the agent refused to state, sets the number themselves, then publishes under the store's name.

Approve file · Correct · Order inspection

Approved — handed back

  1. 6File saved against the stock numberHeld unpublished on the stock record; nothing reaches a marketplace at this stage
  2. 7Outcome evaluatedWhat the manager corrected, what was published, and disclosure and evidence misses by cohort

Corrections

Every claim the manager adds back, or strikes out, is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Stating a value, an offer or a trade figure.
    • Pricing a unit or setting a listing price.
    • Calling a battery healthy, degraded or sound.
    • Stating a range the vehicle will achieve.

  • Automation boundary

    Agent acts unaided

    • ✓Decode the VIN and match one build record.
    • ✓Gather service records, damage, title brands and warranty terms.
    • ✓Record each battery reading with its source, tool and date.
    • ✓Assemble the appraisal file and draft the listing copy, unpublished.

    The appraisal file is written to the stock record, never to a marketplace. A named person sets the value and publishes.

    • Confirming coverage, or that a warranty transfers.
    • Saying a unit has no open recall.
    • Deciding the as-is or warranty disclosure.
  • Publishing a listing to any marketplace.

Example output

One appraisal file, annotated

Everything the agent puts up is attached to the reading and the record it was taken from.

Appraisal output · one used EV

Illustrative example
Unit
Trade-in EV, one prior owner
Battery reading
Dash estimate, cold morning
Matched to
One VIN, one build
File status
Unpublished — manager to value
Confidence
88%
Battery condition
Not stated by the agent

As received

The unit as it arrived and the reading as it was taken, with its conditions beside it — nothing here is adjusted.

Evidence used

  • VIN and build match
  • Battery reading, dated
  • Recall lookup, timestamped

Why no health claim

A dash estimate off a cold pack is not a diagnostic result. The claim is the manager's.

What the score decides

Confidence decides how much the manager verifies, not what the listing is allowed to claim.

Action

Approve fileCorrectOrder inspection

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every unit taken in

Trade-in, lease return, auction or lot stock

03Evidence & draft

Date every reading to its source

Use the build record, the service and damage history, the warranty terms and the readings as they were actually taken.

01Approved path

The evidence arrives assembled

Battery readings, warranty terms, recall lookups and damage history land in one dated file instead of across four systems.

02Human review

Hold the listing where a claim starts

Copy that would have to state a value, a range or a sound battery waits for the manager who answers for the advert.

04

Build an evidence trail

Retain every reading, its source and date, the recall lookup, the draft, the manager's correction and what was published — 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 & inventory

    • CDK Global
    • Tekion
    • vAuto
    • Stock and appraisal records

  • Listing & syndication

    • Dealer website
    • marketplace feeds
    • HomeNet
    • OEM certified programme

  • History, title & recalls

    • Carfax
    • AutoCheck
    • NHTSA VIN lookup
    • OEM campaign portal

Agent

Used-EV appraisal & listing

  • Decodes the build
  • Dates the evidence
  • States no value

  • Battery & warranty

    • OEM diagnostic reports
    • scan tools
    • Telematics readings
    • warranty terms

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers around what the file may claim

Each control wraps the one inside it. The six wrap the file; the number the manager writes and the moment it is published are outside them.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeReturn appraisal and listing work to staff if evaluations or production signals degrade.Roll back
L5TraceabilityRecord the sources read, the readings, the draft, the correction and what was published.Record
L4Human publishA listing is published by the used-car manager who set the value on it.Gate
L3Disclosure checkThe draft is held against the disclosures your store requires for this unit.Check
L2Evidence provenanceNo reading is carried without its source, its tool, its date and its conditions.Attribute
L1Claim withholdingThe value, the range figure and the battery claim are the manager's to write.Withhold
Model coreAppraisal file assembled — the dated evidence, the warranty read and the draft listing copy
L1 – L2
Decide what the file is allowed to assert
L3
Decides what the listing has to carry
L4 – L5
Keep the publish with a person and the trail intact
L6
Pulls automation back when signals degrade

How Nestack evaluates it

Evaluate the whole file — not only the listing copy at the end.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the appraisal file the manager opens

  1. E1Final-output evaluationWas the drafted listing true to the evidence behind it?
  2. E2Step-level evaluationDid it decode the build and read the right history and warranty terms?
  3. E3Tool evaluationDid it query the correct VIN in the recall, history and diagnostic sources?
  4. E4Claim-withholding recallDid a value, a range or a battery-condition claim survive into a draft?
  5. E5Slice evaluationHow does file quality change across vehicle and evidence cohorts?
  6. E6Business outcomeHow much did the manager correct, and what was published wrong anyway?

Floor — the listing a buyer finally reads

Failure modes

Where each failure originates in the agent

Seven failure modes plotted against the five stages of the agent lifecycle.

Agent lifecycle

Direction of processing →

  1. 01 · Decode / retrieval

    2 modes

    • UE-01

      Recall lookup gone stale

      A campaign was posted after the VIN was last checked.

    • UE-02

      Original-owner term carried over

      The second owner's shorter coverage is the one that applies.

    Stage resolves

    The VIN, the build record, the history and the warranty terms

  2. 02 · Battery evidence

    2 modes

    • UE-03

      Dash estimate read as a test

      A cold-pack number is carried as a measured capacity.

    • UE-04

      Reading taken on a warm pack

      Captured straight off a fast charge, unrested and unrepeatable.

    Stage records

    Each reading, its source, its tool and its conditions

  3. 03 · Drafting

    1 mode

    • UE-05

      Range implied by the copy

      A draft turns a capacity figure into what the car will do.

    Stage drafts

    The listing copy, the disclosures and the blanks left

  4. 04 · Handoff / publish

    1 mode

    • UE-06

      Disclosure dropped in the feed

      Syndication truncates the line the store requires.

    Stage presents

    The file a manager values and the listing they publish

  5. 05 · Change / Version

    1 mode

    • UE-07

      Template revision not re-checked

      A wording change quietly weakens what a listing must carry.

    Stage tracks

    Model, prompt, disclosure and marketplace rules

  • Sev-1 · a claim reaches a buyer
  • Sev-2 · the file misstates the vehicle
  • Sev-3 · evidence thins, more is verified
See the complete Automotive failure-mode catalogue

Affected slices

A reading is not the same as a test

A number read off the dash, a pack replaced under a campaign and an auction unit with no service file are where the manager sends the evidence back before anything is listed. File quality is reported by cohort, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Dash-only battery readings5.6%2.8×Review
Recall-replaced battery packs4.4%2.2×Review
Auction units, thin history3.6%1.8×Watch
Recent same-model trade-ins1.4%0.7×Normal

Bar: manager-correction rate lift vs. same-model trade-in baseline · scale 0–4.0× · tick at 2.0×

2 of 4 slices over threshold

Evidence-linked improvement

A reading without a source is not evidence

A manager sending the file back is not one bad appraisal. It changes the provenance rule, the disclosure list or the source the agent is allowed to read.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Manager corrections, disclosure misses or disputed readings rise in a cohort.

  2. 02

    Diagnose

    Narrowed to the build decode, the source read, the age of the reading, the warranty term or the draft.

  3. 03

    Improve

    The provenance rule or the disclosure list changes under your used-car manager's sign-off, version-linked.

  4. 04

    Verify

    Re-run on stored units from that cohort, including the ones that were listed and pulled.

  5. 05

    Learn

    The pulled listing is kept as a test case, and the claim it carried enters your store's advertising notes.

  6. Learn → Detect

    The return edge. The next unit is drafted against one more claim this store has already had to take back.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, decode and retrieval, battery evidence, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Workflow discovery and automation-boundary definition.Week 1
  2. 02DMS, inventory and listing-feed access assessment.Week 1, Week 2
  3. 03Disclosure rules and unstated claims.Week 1, Week 2
  4. 04VIN decode and build-record matching.Week 2
  5. 05History, title-brand and recall retrieval.Week 2, Week 3
  6. 06Battery-evidence and provenance rules.Week 3
  7. 07Warranty term, transfer and coverage read.Week 3, Week 4
  8. 08Appraisal file assembly and listing drafting.Week 3, Week 4
  9. 09Claim-withholding and disclosure testing.Week 4, Week 5
  10. 10Cohort slice testing across evidence sources.Week 5
  11. 11Stock-record write-back and manager handoff.Week 5, Week 6
  12. 12Observability, deployment and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final 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 sourceProductionLive inventory and feedsAdvancedWhole group inventory
Introduced at Pilot
Appraisal file with dated, sourced readings✓✓✓
Withheld value, range and battery claims✓✓✓
A named person values and publishes✓✓✓
Store disclosure and wording templates✓✓✓
History, title-brand and recall retrieval✓✓✓
Baseline evaluation✓✓✓
Introduced at Production
Stock-record write-back and manager handoff—✓✓
Drafts shaped for your syndication feeds—✓✓
Observability and evaluation—✓✓
Introduced at Advanced
Additional battery-diagnostic sources——✓
Group inventory and rooftop controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on rooftops, DMS and inventory integrations, battery-evidence sources, listing feeds, disclosure approvals and deployment requirements.
Separate from build
Build 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.

  1. 01Your approved disclosures and what a listing may never claim

    Disclosure rules and unstated claims

    Week 1

  2. 02Read access to the DMS, the inventory tool and your listing feeds

    DMS, inventory and listing-feed access assessment

    Week 2

  3. 03The battery-evidence sources you trust, and who actually takes them

    Battery-evidence and provenance rules

    Week 3

  4. 04How warranty terms and transfers are recorded for your franchises

    Warranty term, transfer and coverage read

    Week 3

  5. 05Listings that had to be corrected or pulled, and what they claimed

    Evaluation suite, regression cases and failure-mode testing

    Week 4

  6. 06The model mix you actually take in, thin-history units included

    Cohort slice testing across evidence sources

    Week 5

  7. 07A named used-car manager to value the unit and publish it

    Manager handoff, then supervised valuation and publishing

    Weeks 5–6

Nothing else is required

Deployment, 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 is where the cohort testing ends and the first units are listed.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Withheld claims, store disclosures and the automation boundary

W2DMS and inventory access, VIN decode and build-record matching

W3History, recall and warranty retrieval, and battery-evidence provenance

W4Disclosure checks, claim-withholding tests and the evaluation suite

W5Cohort testing, stock-record write-back, then the first units listed

W6Managers value and publish from live files, then Agent Care starts

Reading the band
Provenance rules are fixed in week 3, so nothing reaches evaluation carrying an undated reading.
At the end of W6
Managers have valued and published from live files, with every claim traceable to the reading and the date behind it.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Automotive AI agent

Build a listing agent around the evidence you can stand behind.

Show us a month of used EVs you took in — the battery readings you had, the warranty terms you quoted and the listings that had to be corrected. We'll rebuild those files from the same evidence and show you which readings could never have carried the claim the listing made.

Typical delivery
4–6 weeks
Deployment
Your DMS, inventory and listing feeds
Valuation
Your used-car manager
After launch
Agent Care available

Nestack Agents · Used-EV appraisal & listing

AGT-AUT-08 · Agent Care available after launch

Build a used-EV listing agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your setup

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.