Automotive AI agent · Service scheduling
Service-Scheduling & Capacity AI Agent
Map the request to the operations your shop books, read the roster, the bays, the parts and the loaners, then hold a slot — the promised time stays with your service advisor.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Who promises
- Your advisor
- After launch
- Agent Care
What this agent does
Holds the slot, promises nothing
In
- 01
Take the request from the phone line, the web scheduler, the OEM app or a reply to a recall letter.
- 02
Read the VIN, its open campaigns, its declined work and what the vehicle was last in for.
Reason
- 03
Match the request to the operations your shop books it under, and the skill level each of those needs.
- 04
Read the grid as it stands — rostered technicians and certifications, bays, hoists and calibration space.
- 05
Check the parts the job would need against the shelf, the loaner list and what can arrive before the vehicle.
Decide
- 06
Offer only slots the shop can staff, equip and part, and say what is missing when none of them work.
- 07
Hold the request for a person when it needs a certification, a part or a loaner that is not there.
Out
- 08
Put a held slot in front of the advisor with the constraint behind it, and the promise left unmade.
- 09
Retain what was offered, what was held, who confirmed it, and what the shop did with the day.
Product statement
The agent maps, checks capacity and holds. The promised time, the price, the fault and the repair stay with your service advisor and your technicians.
Example workflow
One service request, end to end
- 1Request receivedPhone line, web scheduler, OEM app, a recall-letter reply or a declined-work follow-up
- 2Vehicle and history matchedOne VIN, its open campaigns, its declined work and what it was last in for
- 3Work mapped to operationsThe operations your shop books this request under, and the skill level each one needs
- 4Capacity readRostered technicians and certifications, bay and equipment, parts on hand, loaners and transport
No human action required
Stages 1 to 4 run without a person in the loop — matching, mapping and the capacity read all finish before an advisor is asked to confirm anything, and nothing has been promised yet.
Reaches the advisor as a slot to confirm.
Goes to the advisor with the gap named.
Service advisor or shop foreman
Reads the hold, the constraint behind it and what the shop would have to move, then sets the time the customer is given under their own name.
Confirm · Move · Call the customer
Confirmed — written back
- 6Slot held against the gridWritten to the schedule as a hold; the promised time is not set at this stage
- 7Outcome evaluatedHolds kept, jobs moved, days the shop could not work what was booked, by cohort
Moves
Every hold an advisor moves or drops is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Quoting a price, a repair time or a fee.
- Naming the fault behind the customer's description.
- Promising the vehicle will be ready by a time.
- Booking work no rostered technician is certified for.
Automation boundary
Agent acts unaided
- ✓Map the request to the operations your shop books.
- ✓Read the roster, certifications, bays and equipment.
- ✓Check parts on hand and the loaner list before a slot is offered.
- ✓Hold a slot, and give it back when the constraint does not clear.
A slot is held, not promised. The advisor sets the time the customer is given, and any figure comes from a person.
- Telling a customer the vehicle is safe to drive.
- Overbooking a day past the capacity a person set.
- Committing the last loaner, rental or mobile visit.
- Changing the capacity model or the skills matrix.
Example output
One held slot, annotated
Everything the agent holds is attached to the request and the capacity read behind it.
Scheduling output · one held slot
Illustrative example- Customer said
- Grinding at the front when I brake
- Channel
- Web scheduler, evening
- Matched to
- One VIN, one owner
- Slot
- Held — advisor to confirm
- Confidence
- 88%
- Ready time
- Not set by the agent
As received
The customer's own words and the channel they arrived on — the agent does not turn them into a fault.
Evidence used
- Operations this shop books
- Certified technician rostered
- Pads and rotors on hand
Why nothing is promised
A grinding noise is not yet a brake job, and the time it takes is the technician's to find.
What the score decides
Confidence decides how hard the advisor re-checks the mapping, not the time given.
Action
ConfirmMoveCall the customerValue
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 service request
Phone, web scheduler, OEM app or recall reply
Book the shop you actually have
Use the roster, the certifications, the bays and equipment, the parts on the shelf and the loaner list as they stand today.
Fewer bookings the shop cannot work
A slot is offered only where a certified technician, the bay the job needs and the parts are all there on that day.
The shortfall shows up early
A missing certification, a part on back-order or the last loaner already out reaches the advisor while the booking can still change.
Build an evidence trail
Retain the request, the operations mapped, the capacity read, the hold, the advisor's confirmation and what the shop did with the day.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Scheduling & DMS
- Xtime
- myKaarma
- Tekion
- CDK Service
- Reynolds appointment grid
Shop capacity
- Technician roster
- certifications
- Bay, hoist and equipment
- skills matrix
Parts & inventory
- Parts inventory
- special-order status
- Supplier ETA
- recall parts availability
Agent
Service scheduling & capacity
- Maps the operations
- Reads real capacity
- Holds, never promises
Customer contact & transport
- Texting platform
- telephony
- Consent lists
- loaner and rental
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 your shop's day
Each control wraps the one inside it. A slot clears every layer before it is held, and the promised time sits outside all six.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Decide whether the shop can work the job
- L3
- Decides which channel a confirmation may use
- L4 – L5
- Leave the promise to a person, log every move
- L6
- Pulls automation back when signals degrade
How Nestack evaluates it
Evaluate the appointment against the day it became.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the slot the customer is offered
- E1Final-output evaluationWas the appointment one the shop could actually work?
- E2Step-level evaluationDid it map the request to the operations your shop books it under?
- E3Tool evaluationDid it read the live roster, grid, stock and loaner list?
- E4Certification recallDid any job reach a technician not certified to do it?
- E5Slice evaluationHow does booking quality change across job and vehicle cohorts?
- E6Business outcomeHow many holds were kept, and how many promises had to move?
Floor — the day the shop actually worked
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 →
01 · Request / match
2 modes
- SV-01
Recall booked before the remedy
The campaign is open but the remedy parts have not shipped.
- SV-02
Declined work booked again
Work already done elsewhere returns as still due.
Stage reads
The request, the VIN, its campaigns and its declined work
02 · Operation mapping
1 mode
- SV-03
A noise mapped to a repair
A grinding complaint books a brake job, not a diagnosis.
Stage maps
The operations your shop books, and the skill needed
03 · Capacity & parts
2 modes
- SV-04
Booked to an uncertified technician
The only one certified for that operation is off that day.
- SV-05
Skills matrix read as current
A lapsed high-voltage certification is still on the roster.
Stage checks
Roster, certifications, bays, equipment and stock
04 · Hold / handoff
1 mode
- SV-06
The last loaner promised twice
Two holds land on one vehicle before either is confirmed.
Stage holds
The slot an advisor confirms and the customer is given
05 · Change / Version
1 mode
- SV-07
Capacity rule loosened quietly
A grid change lets in more than the shop can staff.
Stage tracks
Model, prompt, capacity rules and programme changes
- Sev-1 · work is booked nobody may do
- Sev-2 · the shop cannot work what was booked
- Sev-3 · capacity leaks, advisors re-book
Affected slices
Some jobs are booked against capacity the shop barely has
A noise nobody can name yet, a high-voltage job only two technicians may touch, and a calibration that needs the one bay with the space — that is where the grid and the day part company. Nestack reports quality by slice, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Vague symptom descriptions | 6.9% | 3.7× | Review | |
| EV high-voltage bookings | 4.6% | 2.5× | Review | |
| ADAS calibration after repair | 3.5% | 1.9× | Watch | |
| Routine oil-service bookings | 1.5% | 0.8× | Normal |
Bar: re-booked-or-overrun rate lift vs. routine oil-service baseline · scale 0–4.0× · tick at 2.0×
2 of 4 slices over threshold
Evidence-linked improvement
A broken promise is a capacity number, not an apology
A job that ran past its promised time is not closed by calling the customer back. It moves the operation mapping, the skill match or the hours the grid sells.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Moved holds, overruns or same-day re-books rise in a cohort.
- 02
Diagnose
Separated into the request read, the operations mapped, the skill match, the parts hold and the grid itself.
- 03
Improve
The mapping, the skills matrix or the hours the grid sells is re-approved by your service manager and version-linked.
- 04
Verify
Replayed against real weeks in your own grid, including the days the shop lost.
- 05
Learn
The lost day is replayed on every later release, and the constraint it exposed is written into the booking rules.
Learn → Detect
The return edge. The next request is booked against a grid that already knows which jobs this shop has been unable to hold.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, mapping and capacity, holds and consent, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Workflow discovery and automation-boundary definition.Week 1
- 02Scheduler, DMS and roster access assessment.Week 1, Week 2
- 03Operation mapping from customer language.Week 1, Week 2
- 04Skills matrix and certification rules.Week 2
- 05Bay, hoist and equipment constraints.Week 2, Week 3
- 06Parts-on-hand and special-order checks.Week 3
- 07Loaner, rental and transport availability.Week 3, Week 4
- 08Capacity model and hold-and-release rules.Week 3, Week 4
- 09Contact channel, consent and quiet-hours checks.Week 4
- 10Evaluation suite and certification-recall testing.Week 4, Week 5
- 11Advisor confirmation and schedule write-back.Week 5, Week 6
- 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 tier | PilotOne rooftop, one lane | ProductionLive grid, live holds | AdvancedMulti-rooftop / group |
|---|---|---|---|
| Introduced at Pilot | |||
| Operation mapping and slot proposals | ✓ | ✓ | ✓ |
| Skill and certification gate | ✓ | ✓ | ✓ |
| An advisor sets every promised time | ✓ | ✓ | ✓ |
| Consent and contact-channel checks | ✓ | ✓ | ✓ |
| Baseline evaluation | ✓ | ✓ | ✓ |
| Bay, equipment, parts and loaner constraints | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Grid utilisation and overrun reporting | — | ✓ | ✓ |
| Schedule write-back and advisor handoff | — | ✓ | ✓ |
| Observability and evaluation | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Mobile service and multi-lane capacity | — | — | ✓ |
| Multi-rooftop and group controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on rooftops, scheduler and DMS integrations, appointment volume, roster and capacity complexity, loaner systems 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
01The words your customers use, and the operations you book them under
→Operation mapping from customer language
Week 102Read access to your scheduler, DMS and technician roster
→Scheduler, DMS and roster access assessment
Week 203Who may work what, and which certifications you hold today
→Skills matrix and certification rules
Week 204Your bays, hoists and calibration space, and the loaner fleet
→Equipment constraints, then loaner and transport availability
Week 305The channels a customer will take a booking confirmation on
→Contact channel, consent and quiet-hours checks
Week 406Days the shop could not work what was booked, and what happened
→Evaluation suite, regression cases and failure-mode testing
Week 407Named advisors and a shop foreman to confirm what the agent holds
→Advisor confirmation workflow, then supervised booking and validation
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. Nothing is held against your live grid until week 5, after the certification tests.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Automation boundary, and the operations your shop books each request under
W2Scheduler, DMS and roster access, and the skills matrix behind it
W3Bay, equipment, parts and loaner constraints, then the capacity model
W4Evaluation suite, certification-recall testing and consent checks
W5Schedule write-back, slice testing and the first confirmed holds
W6Advisors book from live holds, then Agent Care starts
- Reading the band
- Weeks 2 and 3 fix the roster, the bays and the parts before week 4 tests them. A slot the shop cannot staff is the thing being tested for.
- At the end of W6
- Your advisors have confirmed live slots through a full week of the shop's own mix, and every promised time was theirs to set.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Automotive AI agent
Build a scheduling agent around the shop you actually have.
Show us a month of service appointments — the grid they were booked into, the technicians rostered, the parts on the shelf and the days that fell over. We'll re-book that month against your real capacity and mark every appointment your shop was never going to be able to work.
- Typical delivery
- 4–6 weeks
- Deployment
- Your scheduler, DMS and roster
- Who promises
- Your service advisor, always
- After launch
- Agent Care available
Nestack Agents · Service scheduling & capacity
AGT-AUT-02 · Agent Care available after launch