Automotive AI agent · Outbound BDC

Outbound BDC & Follow-Up AI Agent

Build the follow-up list from the consent record, hold every number a check cannot clear and draft from approved wording — your BDC manager owns the campaign and any figure comes from a person.

Typical delivery: 4–6 weeks
Deployment: Your stack
Any figure: From a person
After launch: Agent Care

What this agent does

Works the list the consent record allows

01
Take the follow-up list from the CRM, the DMS, the service file and the campaign your BDC manager approved.
02
Read the consent on each number — what was given, for what, on which channel, and every stop that has landed since.

Reason

03
Check the number against the suppression lists it is connected to, and name the ones it could not reach.
04
Read the local hour at the number rather than the area code, against the calling hours that state allows.
05
Draft from the wording approved for the reason this customer is on the list, with every figure left out.

Decide

06
Hold a number when a check does not clear, and say which check held it instead of working it anyway.
07
Route a reply that reads as a stop, a wrong number or a complaint to a person before the cadence runs again.

Out

08
Put the draft in front of the named BDC person with the consent evidence and the checks that ran beside it.
09
Retain what went out, on what consent, at what local hour, from whom, and every stop that landed after it.

Product statement

The agent works only numbers the consent record clears and only wording your BDC manager approved. Any figure in a message, and any reply that is not a simple answer, comes from a person.

Example workflow

One follow-up, end to end

Agent

Human

1List builtUnsold showroom traffic, a lead gone quiet, declined service, an open campaign or a lease maturing
2Consent record readWhat consent was given, for what, on which channel, and every stop that has landed since
3Contactability checkedInternal suppression, the registries and reassigned-number data, then the local hour at the number
4Message draftedApproved wording for the reason this customer is on the list, with every figure left out

No human action required

Stages 1 to 4 run without a person in the loop — the list, the consent record and the contactability checks all finish before anyone is asked to read a draft, and nothing has gone out.

5DecisionSplits on whether the consent record and the checks clear the number

Cleared, and no figure needed

Reaches the BDC person ready to send.

A check held it, or a figure is wanted

Goes to the BDC with the check named.

BDC manager or the named salesperson

Reads the draft, the consent evidence behind it and the check that held the number, then sends under their own name or takes the number off the list.

Send · Edit · Suppress

Sent — logged against the consent

6Sent by a named person, or heldGoes out under the named BDC person who sent it, inside the boundary agreed at implementation
7Outcome evaluatedStops honoured and how fast, wrong numbers, complaints, held numbers and replies, by cohort

Stops

Every stop, wrong number and complaint is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

Sending a message that states a price or payment.
- Quoting an APR, a term or a rebate.
- Putting a trade or payoff figure in writing.
- Releasing a number the consent check held.
Automation boundary
Agent acts unaided
- ✓Build the list from your BDC manager's approved campaign.
- ✓Read the consent, the suppression lists and the local hour.
- ✓Draft from the wording approved for that reason, with every figure left out.
- ✓Hold a number when a check does not clear, and name the check.
A message goes out under the BDC person named against it, and only inside the boundary agreed at implementation.
- Adding a customer to a campaign they never joined.
- Answering a complaint, a dispute or a legal demand.
- Making a commitment the dealership must keep.
Changing the cadence, the wording or the suppression rules.

Example output

One follow-up message, annotated

Everything the agent drafts is attached to the consent record and the checks that ran behind it.

Outbound draft · one customer, one channel

Illustrative example

On the list because: Service declined at the last visit
Channel: Text, written consent on file
Local hour: Inside hours
Message: Drafted — cleared to send
Confidence: 86%
Figures in it: None — left to the advisor

As received

The reason this customer is on the list and the channel their consent covers — the agent adds neither.

Evidence used

Written consent, this store
No stop on any channel
Not on a suppression list

Why no figure

A price or a payment in a text is an advertisement, and the store can be held to it.

What the score decides

Confidence decides how closely the BDC reads the draft, not who may be contacted.

Action

SendEditSuppress

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 number on the list

CRM, DMS, service file or campaign

03Consent & cadence

Check the record before the message

Use the consent on file, the stops already landed, the suppression lists you connect and the hours that state allows.

01Approved path

The list arrives already checked

Consent, suppression, reassigned-number and calling-hours checks have all run before a BDC person opens the first draft.

02Human review

A stop reaches the next campaign too

A reply that reads as a stop goes to a person, and the number is suppressed across the lists we connect before the cadence runs again.

Build an evidence trail

Retain the consent read, the checks that ran, the wording drafted, who sent it, and every stop, wrong number and complaint that followed.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

CRM & DMS
- VinSolutions
- Elead
- DealerSocket
- Tekion
- CDK and Reynolds records
Consent & suppression
- Consent records
- internal do-not-call
- Registry scrubs
- reassigned-number data
Messaging & telephony
- Texting platform
- dialler and IVR
- Email delivery
- call recordings

Agent

Outbound BDC & follow-up

Reads the consent
Checks suppression
Drafts the wording

Campaign sources
- Equity and lease lists
- service declined
- Open campaigns
- third-party leads
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 send button

Each control wraps the one inside it. A message clears every layer before it leaves, and any figure in it sits outside all six.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeReturn outreach to your BDC if evaluations or production signals degrade.Roll back

L5TraceabilityRecord the consent read, the checks run, the wording, the sender and every stop.Record

L4Named senderA message leaves under the BDC person named against it, figure or not.Gate

L3Hours and cadenceThe local hour at the number and how often that campaign may run.Limit

L2Suppression checksInternal list, registries and reassigned-number data are read before a draft exists.Screen

L1Consent gateA number is worked only where the record shows consent for that channel and purpose.Verify

Model coreMessage drafted — the reason for contact, the channel, the wording and the consent read behind it

L1 – L2: Decide whether a number may be worked at all
L3: Decides when, and how often, it may be worked
L4 – L5: Name the sender, and log every message sent
L6: Pulls automation back when signals degrade

How Nestack evaluates it

Evaluate who was contacted, not only what was said.

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

Surface — the message the customer receives

E1Final-output evaluationWas this a message the customer had consented to receive?
E2Step-level evaluationDid it read the consent record, the stops and the suppression lists?
E3Tool evaluationDid the stop reach every system that holds that number?
E4Revocation latencyHow long after a customer said stop did the last message go out?
E5Slice evaluationHow does contact quality change across list and channel cohorts?
E6Business outcomeHow many complaints, wrong numbers and opt-outs did the campaign cost?

Floor — the stop, the complaint or the appointment that follows

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 · List & consent
2 modes
- OB-01
  A stop that never said STOP
  Plain-English revocation is not read as revocation.
- OB-02
  Sold customer still in the cadence
  The deal closed last week; the win-back list has not.
Stage reads
The list, the consent on file and every stop since
02 · Contactability
2 modes
- OB-03
  The stop landed on one platform
  The texting tool has it, the dialler list does not.
- OB-04
  Area code read as location
  A ported number is worked outside its own state's hours.
Stage checks
Suppression lists, registries and the local hour
03 · Message draft
1 mode
- OB-05
  A payment reaches the wording
  A merge field pulls an incentive into a text.
Stage drafts
Approved wording for the reason for the contact
04 · Send / handoff
1 mode
- OB-06
  A line that commits the store
  Wording a customer can hold the dealership to.
Stage hands
The message that goes out, and the person behind it
05 · Change / Version
1 mode
- OB-07
  Cadence loosened quietly
  A campaign change raises frequency past what was approved.
Stage tracks
Model, prompt, cadence and suppression-rule changes

Sev-1 · contact goes out where it may not
Sev-2 · the message says what it may not
Sev-3 · the same customer is worked again

See the complete Automotive failure-mode catalogue

Affected slices

Not every stop arrives in a word the system knows

A customer who asks in their own words to be left alone has revoked consent as surely as one who writes STOP, and a reassigned number carries consent its new owner never gave. Nestack reports contact quality by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example

Slice	Failure rate	Lift	Status
Freeform stop wording	6.6%	4.1×	Review
Reassigned or ported numbers	4.2%	2.6×	Review
Aged consent records	2.9%	1.8×	Watch
Recent service customers	1.1%	0.7×	Normal

Bar: unwanted-contact rate lift vs. recent-service baseline · scale 0–4.0× · tick at 2.0×

2 of 4 slices over threshold

Evidence-linked improvement

A complaint is a list problem before it is a person problem

Taking one customer off one list does not close it. The same gap sits in the consent read, the suppression sync or the cadence behind it.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

01
Detect
Complaints, stops honoured late or wrong-number replies rise in a cohort.
02
Diagnose
Read back through the consent read, the suppression sync, the hours check, the wording and the cadence itself.
03
Improve
The suppression rule, the cadence or the approved wording changes under your BDC manager's sign-off, version-linked.
04
Verify
Re-run on stored campaigns from that cohort, including the numbers that should never have been worked.
05
Learn
The complaint becomes a standing check, and the number it came from stays suppressed.
Learn → Detect
The return edge. A stop written in a customer's own words becomes a case the next campaign is tested against before it is sent.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, consent and suppression, cadence and wording, evaluation, integration, then production validation and handover.

Workstream

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

01Workflow discovery and automation-boundary definition.Week 1
02CRM, DMS and telephony access assessment.Week 1, Week 2
03Consent-record mapping and evidence rules.Week 1, Week 2
04Internal suppression list and registry scrubs.Week 2
05Reassigned and wrong-number handling.Week 2, Week 3
06Calling-hours and quiet-hours rules by state.Week 3
07Campaign lists, cadence and frequency caps.Week 3, Week 4
08Approved wording and withheld-figure rules.Week 3, Week 4
09Reply handling and stop routing to a person.Week 4
10Evaluation suite and revocation-latency testing.Week 4, Week 5
11Named-sender workflow and CRM 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 list, one channel	ProductionLive campaigns, live sending	AdvancedMulti-rooftop / group
Introduced at Pilot
Message drafting from approved wording	✓	✓	✓
Consent and calling-hours checks on every number	✓	✓	✓
Suppression, registry and reassigned-number checks	✓	✓	✓
Stop and complaint routing, with stop write-back	✓	✓	✓
Every message leaves under a named sender	✓	✓	✓
Baseline evaluation	✓	✓	✓
Introduced at Production
Campaign lists, cadence and frequency caps	—	✓	✓
Lead and appointment write-back to the CRM	—	✓	✓
Observability and evaluation	—	✓	✓
Introduced at Advanced
Multi-channel outreach and language coverage	—	—	✓
Multi-rooftop and group controls	—	—	✓
Build price	From $5,000	From $8,000	Custom quote

Final build price: Confirmed after discovery based on rooftops, CRM, DMS and telephony integrations, contact volume, the number of campaigns and channels, consent-record complexity 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 consent records, and where the evidence for each one actually lives
→
Consent-record mapping and evidence rules
Week 1
02Read access to your CRM, DMS and texting or dialler platform
→
CRM, DMS and telephony access assessment
Week 2
03Your internal do-not-call list, and everywhere a stop has to land
→
Internal suppression list and registry scrubs
Week 2
04The campaigns you run, and why a customer lands on each list
→
Campaign lists, cadence and frequency caps
Week 3
05The wording you already approve, and the figures you keep out of writing
→
Approved wording and withheld-figure rules
Week 3
06Complaints, wrong numbers and the messages you wish had not gone out
→
Evaluation suite, regression cases and failure-mode testing
Week 4
07Named BDC people to send what the agent drafts and to own each campaign
→
Named-sender workflow, then supervised campaigns 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. Week 5 carries both the last of the evaluation and the first messages a named person sends.

Phase

DiscoveryW1
BuildW2 – W3
EvaluateW4 – W5
Pilot & LaunchW5 – W6

Week focus

W1Automation boundary, and the consent record behind every number you work

W2CRM, DMS and telephony access, the suppression list and the registry scrubs

W3Calling hours, cadence caps and the wording a message is allowed to carry

W4Evaluation suite, revocation-latency testing, reply and stop routing

W5CRM write-back, slice testing and the first messages a named person sends

W6Your BDC works live lists from drafts, then Agent Care starts

Reading the band: Revocation-latency testing runs in week 4, before a message reaches a customer in week 5. A stop honoured late is worse than a campaign that never ran.
At the end of W6: Drafts have run beside your existing follow-up, with every figure still coming from a person and each stop written back across the systems we connect.
Duration: Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Automotive AI agent

Build an outbound agent around the consent you actually hold.

Show us a quarter of outbound follow-up — the lists, the messages that went out and the complaints, wrong numbers and opt-outs behind them. We'll re-run those campaigns against your own consent record and mark every number that should never have been worked.

Typical delivery: 4–6 weeks
Deployment: Your CRM, DMS and consent record
Any figure: From a named person, always
After launch: Agent Care available

Nestack Agents · Outbound BDC & follow-up

AGT-AUT-03 · Agent Care available after launch

Outbound BDC & Follow-Up AI Agent

Check the record before the message

The list arrives already checked

A stop reaches the next campaign too

CRM & DMS

Consent & suppression

Messaging & telephony

Outbound BDC & follow-up

Campaign sources

Observability & evaluation

A stop that never said STOP

Sold customer still in the cadence

The stop landed on one platform

Area code read as location

A payment reaches the wording

A line that commits the store

Cadence loosened quietly

Detect

Diagnose

Improve

Verify

Learn

Build an outbound agent around the consent you actually hold.