Automotive AI agent · ADAS calibration
ADAS Calibration & Verification AI Agent
Determine from the repair whether a calibration is required and which type, pin the OEM procedure in force on the repair date, and draft the record a technician signs before the vehicle goes back.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Technician sign-off
- Pre-delivery
- After launch
- Agent Care
What this agent does
Reads the repair, drafts the record
In
- 01
A repair order arrives — VIN, damage lines, parts list and estimate read from the shop or estimating system.
- 02
A windscreen is replaced: glass part, bracket, pre-repair scan and sublet are normalised onto one job.
Reason
- 03
A bumper comes off, so the radar behind it counts as touched and the calibration requirement is determined.
- 04
A battery is disconnected or a ride height changes — triggers the OEM procedure names, not only crash damage.
- 05
A procedure version is pinned to the repair date and retained, not whichever revision happens to be live today.
Decide
- 06
A recycled or aftermarket ADAS part is flagged against the OEM position statement before a bay is booked.
- 07
A calibration report comes back, and its elements, VIN, tolerances and drive cycle are checked for gaps.
Out
- 08
A job in a notice state is matched to that state's own duty — Utah 2020 through New York 2025, and they differ.
- 09
A record is drafted for the technician, and any write-back stays inside the boundaries agreed at implementation.
Product statement
The agent determines what the repair requires and drafts the record; a technician signs that it was performed and verified, and the shop makes the statement.
Example workflow
One repair, order to signature
- 1Repair order receivedEstimate lines, VIN, damage report, parts list or scan file
- 2Requirement determinedThe OEM procedure in force on the repair date, its triggers, prerequisites and parts rules
- 3Evidence reconciledPre- and post-repair scans, calibration report, drive cycle and confidence
- 4Controls appliedElement checks, VIN and model-year match, sequence checks, parts category and confidence threshold
No human action required
Stages 1 to 4 run unaided, and nothing is certified at any of them — the agent is assembling, and the technician's lane opens at the confidence gate.
Goes to the technician to sign.
Adds a foreman read first.
Technician sign-off
The record is held with its procedure version, its evidence and the named gaps.
Sign · Return to bay · Escalate to foreman
Signed — released to the file
- 6Shop systems updatedOnly where write access and approval policy allow it
- 7Outcome evaluatedRequirement accuracy, returned jobs, supplement timing and post-delivery comebacks
Returns
Jobs sent back to the bay are counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Performing the calibration on the vehicle.
- Certifying that a calibration passed.
- The written statement made to the customer.
- Releasing the vehicle back to the owner.
Automation boundary
Agent acts unaided
- ✓Determine from the repair what calibration is required.
- ✓Pin the OEM procedure in force on the repair date and keep that version.
- ✓Reconcile the scans, the report and the drive cycle.
- ✓Name the missing elements and hold the record for the technician.
Any write-back runs inside the boundaries agreed at implementation, never ahead of a signature.
- Fitting a recycled or aftermarket ADAS part.
- Overriding an OEM procedure or position statement.
- Whether a glass-facility notice duty binds this shop.
- Changes to procedure sources or evidence rules.
Example output
One calibration record, annotated
Everything the agent assembles is attached to the job it was drawn from.
Calibration file · single job
Illustrative example- Vehicle
- Camera-equipped saloon
- Repair line
- Windscreen replaced, forward-camera bracket refitted
- Procedure
- Static, then dynamic
- Procedure of record
- Ford position, 1 May 2026
- Confidence
- 93%
- Statement
- Drafted for the shop to make
As received
Taken from the repair order and the procedure pinned to the repair date — nothing on this side is written by the agent.
Evidence assembled
- Pre-repair scan
- Calibration report
- Drive-cycle record
Why calibration applies
The bracket was touched during the repair and a person still decides.
What the score decides
Below the configured threshold the record picks up a foreman read before it.
Action
SignReturn to bayEscalate to foremanValue
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 ADAS-equipped repair
From the repair order
Read the repair, not the damage
Use the OEM procedure in force on the repair date, its trigger list and its parts rules.
A clean scan proves nothing
A sensor can sit out of specification without setting a code, so the proof is the report and the drive cycle.
Put the read where the gaps are
Jobs missing an element, a VIN match or drive-cycle evidence are marked, so the technician's read starts there.
Build an evidence trail
The calibration, the procedure it was run to and the technician who verified stay on the job.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Estimating systems
- CCC ONE
- Mitchell
- Audatex
- Web-Est
Scan and calibration tools
- asTech
- Bosch
- Autel
- Hunter
- Texa
- OEM tools
OEM repair procedures
- Ford
- GM
- Toyota portals
- ALLDATA
- Identifix
Agent
ADAS calibration & verification
- Reads the repair
- Determines the requirement
- Drafts the record
Shop management
- Shop-Ware
- Tekmetric
- R.O. Writer
- sublet vendors
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the signed record
The six sit one within another, narrowing. What survives the lot is listed in the map beneath.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether the file may stand
- L3
- Puts the signature in a technician's hands
- L4 – L5
- Keep the calibration and the procedure behind it
- L6
- Holds the job at unverified when signals degrade
How Nestack evaluates it
Evaluate the whole determination — not only the calibration report.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the record the customer is handed
- E1Final-output evaluationWas the calibration requirement determined correctly?
- E2Step-level evaluationDid the agent use the procedure version in force on the repair date?
- E3Tool evaluationDid it read the right repair order and the right scan file?
- E4Confidence calibrationDo low-confidence requirement calls actually get overturned more?
- E5Slice evaluationHow does performance change across specific sensor systems?
- E6Business outcomeHow many jobs came back, or needed a supplement after delivery?
Floor — the outcome the shop answers for
Failure modes
Where each failure originates in the agent
Seven failure modes, placed at the stage each one originates.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- XT-03
Superseded procedure
Today's OEM procedure pulled, not the repair-date version.
Stage gathers
Repair order, scans, procedures and parts data
02 · Reasoning
2 modes
- XT-04
Trigger not read
A battery disconnect or alignment is not treated as a trigger.
- XT-06
Wrong state's duty
A notice is drafted from another state's template.
Stage proposes
Requirement, procedure type, gaps and confidence
03 · Tool / write
2 modes
- XT-02
File closed unverified
Job marked done on a scan rather than a report.
- XT-05
Report filed twice
One calibration report is attached to two jobs.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- XT-01
Element gap missed
A required report element is present but does not match.
Stage returns
The record a technician signs and the file keeps
05 · Change / Version
1 mode
- XT-07
Silent procedure regression
A source or rule change narrows what the agent flags.
Stage tracks
Model, prompt, procedure sources and evidence rules
- Sev-1 · a record closes without evidence
- Sev-2 · a wrong statement reaches the customer
- Sev-3 · source degrades, job routes to review
Affected slices
One system can carry most of the misses
A shop-level calibration rate can look healthy while a single sensor system absorbs most of the missed requirements. Nestack reports the missed-requirement rate by slice, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Front camera behind glass | 7.7% | 3.7× | Review | |
| Blind-spot and rear radar | 5.1% | 2.4× | Review | |
| Sublet mobile calibration | 3.5% | 1.7× | Watch | |
| Single-system in-house jobs | 1.9% | 0.9× | Normal |
Bar: missed-requirement lift vs. single-system baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
The loop closes on a regression case
The loop shuts when the skipped calibration is a regression case, not when it has been explained. That suite is what the next job closed is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Missed-requirement rate rises in one sensor system.
- 02
Diagnose
The car that drove out with a camera three degrees left is traced back to the repair line that missed the trigger.
- 03
Improve
Any change goes out numbered, and the jobs that caused it are attached.
- 04
Verify
The release waits on the touched calibration cases clearing a second run.
- 05
Learn
One case added to the suite, one line added to the procedure record.
Learn → Detect
The return edge. The next miss is looked for against a suite one case longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, requirement logic, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Calibration workflow discovery and boundary setting.Week 1
- 02Scan-tool, DMS and estimate assessment.Week 1, Week 2
- 03OEM procedure, trigger and evidence-rule mapping.Week 1, Week 2
- 04Repair-order ingestion and normalisation.Week 2
- 05Requirement logic and procedure pinning.Week 2, Week 3
- 06Confidence scoring and gap routing.Week 3
- 07Technician sign-off workflow.Week 3, Week 4
- 08Scan-tool and estimate-system integration.Week 5
- 09Requirement and evidence cases.Week 4, Week 5
- 10Guardrails and verification controls.Week 4
- 11Job-trail instrumentation.Week 5, Week 6
- 12Deployment, documentation 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 tier | PilotOne make, one shop | ProductionProduction shop systems | AdvancedMultiple shops / makes |
|---|---|---|---|
| Introduced at Pilot | |||
| Requirement calls on your procedures | ✓ | ✓ | ✓ |
| Technician sign-off | ✓ | ✓ | ✓ |
| Requirement-detection baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by system | — | ✓ | ✓ |
| Sign-off workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Scan-tool and estimate integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-state notice rules | — | — | ✓ |
| Multi-stage shop approvals | — | — | ✓ |
| High job volume | — | — | ✓ |
| Multi-make procedure controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls 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.
You bring
We build with it
01Your repair-order fields and job structure
→Repair-order ingestion and trigger mapping
Week 102Representative closed calibration jobs
→Requirement baseline, procedure pinning and evidence binding
Week 203Your OEM procedure sources and position statements
→OEM procedure, trigger and evidence-rule mapping
Week 104Access to relevant APIs, feeds or exports
→Scan-tool, DMS and estimate assessment, then integration setup
Week 205Jobs you would not want delivered
→Missed-calibration cases and failure-mode testing
Week 406What no delivery may assume
→Confidence scoring, gap routing, guardrails and sign-off controls
Week 307Named technicians and a foreman to review records
→Technician sign-off workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
The chart follows the shop rather than the slide, so evaluation and pilot both land in week 5.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Calibration workflow discovery, procedure mapping and the boundary
W2Source integration and the requirement baseline
W3Requirement logic, confidence scoring and sign-off controls
W4Evaluation suite, evidence-element checks and failure-mode testing
W5Scan-tool integration, pilot jobs and targeted corrections
W6One repair cycle run under the shop foreman, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its work is named in. The week 5 overlap is real, not padding.
- At the end of W6
- With the cycle validated, the running agent passes to Agent Care.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Automotive AI agent
Build an ADAS calibration agent around your shop's sign-off.
Show us your repair orders, your procedure sources and who signs the record. Who makes the statement to the customer? The shop does — so the agent stops at a drafted file.
- Typical delivery
- 4–6 weeks
- Deployment
- Your shop systems
- Technician sign-off
- Pre-delivery
- After launch
- Agent Care available
Nestack Agents · ADAS calibration & verification
AGT-AUT-11 · Agent Care available after launch