Lead-Qualification & Follow-Up Voice AI Agent

What this agent does

Dials the list, never builds it

Check that the campaign, list and script carry the supervising broker's release before a number enters the run.

Verify each number against the internal do-not-call record and the suppression services the brokerage subscribes to.

Reason

Resolve the recipient's local calling window and attempt cap to the stricter of the federal and state rule.

Check that a consent artefact is attached to the record and is the one the broker marked in scope for this campaign.

Speak the responsible entity, a callback number and the AI disclosure the brokerage configured for that jurisdiction.

Decide

Qualify against the questions on the approved script, and nothing outside it.

Resolve a clear stop, remove-me or equivalent request to the suppression queue, and end the call.

Out

Retain the consent reference, the recording state, the script version and the disposition against the lead record.

Execute write actions only inside the approval boundaries agreed during implementation.

→ Product statement

An AI voice is an artificial voice under the FCC's February 2024 ruling, so the agent sits inside 47 USC §227 by construction; the brokerage is the caller and the supervising broker releases a campaign.

Example workflow

One attempt, list to disposition

AgentHuman

1Released campaign receivedBroker-approved list, script, cadence and calling-hour rules

2Consent state assembledConsent artefact, suppression status, recipient state and prior attempt history

3Attempt preparedDialling window, script path, recording state and confidence

4Controls appliedSuppression checks, calling-hour and frequency rules, recording-state gate and confidence threshold

No human action required

Stages 1 to 4 run unaided and dial nothing — a placed call cannot be unplaced, so the broker's lane opens at the confidence gate, ahead of the dial.

5DecisionSplits at the confidence threshold

High confidence

Waits for the supervising broker.

Low confidence

Adds a compliance-officer read first.

Broker approval

The attempt is held with its consent artefact, its state rules and the confidence.

Approve · Amend · Send to compliance

Approved — the call is placed▼

6Dialer and CRM updatedOnly where write access and approval policy allow it

7Outcome evaluatedConsent evidence, disclosure delivery, opt-out handling and complaints after the call

Amendments

Each broker amendment is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Adding a number to a dialling list.

Deciding that a consent record is valid or in scope.

Launching or resuming a campaign.

Turning on call recording or transcription.

Automation boundaryAgent acts unaided

✓Speak the responsible entity's identity and callback number.

✓Speak the AI disclosure configured for the recipient's jurisdiction.

✓Capture a clear stop request and write it to the suppression queue.

✓Transfer the call to the named licensed agent.

Writes happen inside the boundaries agreed at implementation; only a stop request goes without waiting.

Deciding that a message is not telemarketing.

Interpreting an ambiguous revocation.

Setting or overriding calling-hour and frequency rules.

Approving script or voice-persona changes.

Example output

One attempt, annotated

Everything the agent prepares is attached to the record it was drawn from.

Voice-agent output · single leadIllustrative example

Lead

Planned opener

Local dial time

Consent on file

Confidence

Recording

Warm inbound enquiry

Automated assistant for the brokerage, callback number spoken first

2:40 pm local

Signed written consent

91%

Off — not captured on this call

As receivedTaken from the lead record and the consent artefact on file — nothing on this side is decided by the agent.

Consent evidence used Signed consent record Broker-released list Suppression check

Why this call nowThe window, the cap and the suppression state cleared before the broker saw it.

ActionApproveAmendSend to compliance

What the score decidesBelow the threshold the attempt picks up a compliance read before the broker sees it.

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

All released leadsFrom the broker's list

03Qualification

Call from the approved script

Work from the consent artefact on file and the calling rules configured for the recipient's state.

01Approved path

Reach the leads that consent covers

Routine first-contact and follow-up attempts run on the approved script.

02Human review

Point review at the doubtful attempts

Attempts with thin consent provenance or an unresolved state rule are held, so the broker's read starts where exposure concentrates.

04Build an evidence trail

Hold the consent reference, the recording state, the script version and the disposition together.

Integrations

Typical integrations

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

Telephony and CPaaSTwilio · Bandwidth
Voice APIs · SIP trunking

Contact centre and dialerFive9 · Genesys Cloud CX
Campaign and cadence tooling

Brokerage CRM and leadsBoldTrail · Lofty
Zillow Premier Agent · Realtor.com

Agent

Lead-qualification voice

Reads the consent state
Places the released call
Hands to the licensed agent

Consent and DNC servicesPossibleNOW · Gryphon.ai
Internal DNC · registry access

Observability & evaluationOpenTelemetry · 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 dial

Each control wraps the one inside it. What a layer misses is named in the failure-mode map that follows it.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modePull the agent back to preparing attempts only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, voice, prompt, script and calling-rule configuration changes.Track

L4TraceabilityRecord the consent reference, the disclosure spoken, the disposition and the suppression writes.Record

L3Broker approvalHold attempts for the supervising broker; it governs whether a call is placed, not whether the consent behind it is good.Gate

L2Policy guardrailsTest attempts against configured suppression, calling-hour and recording-state rules; a failure returns the attempt.Restrict

L1Confidence thresholdsRoute low-confidence attempts to a compliance read before the broker sees them.Require review

Model coreAttempt prepared — dialling window, script path, recording state and confidence

L1 – L2Test whether an attempt may stand

L3Puts the dial in the broker's hands

L4 – L5Keep the record and the version honest

L6Pulls calling back when signals degrade

How Nestack evaluates it

Evaluate the calling workflow — not only the words on the call.

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

Surface — the call the recipient answers

Depth of coverage ▼

E1Final-output evaluationDid the opener carry the identity, the callback number and the AI disclosure?

E2Step-level evaluationDid the agent use the right consent artefact, state rules and script version?

E3Tool evaluationDid the opt-out reach the brokerage-wide do-not-call record, not just the campaign?

E4Confidence calibrationDo low-confidence attempts actually attract more broker amendments?

E5Slice evaluationHow does performance change across specific recipient cohorts?

E6Business outcomeHow many attempts could not produce a consent artefact, or drew a complaint afterwards?

Floor — the outcome the brokerage answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each shown at the stage where it starts.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

VO-01

Co-registration artefact

A checkbox on a shared form is read as consent for this caller.

Stage gathersConsent artefact, suppression status and campaign rules

02 · Reasoning2 modes

VO-02

Area-code time inference

Calling time is taken from the area code, not the recipient.

VO-03

Revocation read narrowly

A stop request is treated as an objection to one property.

Stage proposesDialling window, script path and confidence

03 · Tool / write2 modes

VO-04

Partial suppression write

The opt-out reaches the campaign table, not the internal list.

VO-05

Recording on by default

Transcription starts on a call resolving to an all-party state.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

VO-06

Identification omitted

The opener drops the responsible entity and the callback number.

Stage returnsThe opener the recipient hears and what it collects

05 · Change / Version1 mode

VO-07

Silent disclosure regression

A voice-model swap drops the AI disclosure as redundant.

Stage tracksModel, voice, script and calling-rule config

Sev-1 · a call placed outside the boundary Sev-2 · the opener or the recording state is wrong Sev-3 · consent evidence degrades, attempt held

See the complete Real Estate failure-mode catalogue →

Affected slices

A safe-looking total, one exposed cohort

The unconsented-contact rate counts attempts placed where a valid, in-scope, unrevoked consent artefact cannot be produced at the moment of dialling. Nestack reports it by slice and by recipient state.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Aged or purchased leads, strict states	9.1%	3.8×	Review
Ported and reassigned numbers	7.2%	3.0×	Review
Recipients in all-party states	3.6%	1.5×	Watch
Warm inbound leads	1.2%	0.5×	Normal

Bar: unconsented-contact-rate lift vs. the all-attempts baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

What did this failure not test for?

That is the question a cycle answers, and it answers it by leaving a regression case the next release must pass. Detection runs against that suite.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Unconsented-contact rate rises in a recipient cohort.

02Diagnose

The broker replays the attempts and the consent records behind them until the cause narrows to one.

03Improve

Version-stamp the change, and attach the attempts that exposed it.

04Verify

Nothing ships until the affected cases pass a second time.

05Learn

The suite grows by one case; so does the calling policy.

Learn → DetectThe return edge. The next cycle is measured 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, calling workflow, evaluation, integration, then production validation and handover.

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

01Calling workflow discovery and boundary definition.

02Telephony, CRM and consent-source review.

03Consent, suppression and state calling-rule mapping.

04Lead-record ingestion and consent linking.

05Script logic and disclosure sequencing.

06Confidence scoring and compliance routing.

07Broker approval workflow.

08Telephony, dialer and CRM integration.

09Evaluation suite and regression cases.

10Suppression guardrails and release controls.

11Call-record retention and tracing.

12Deployment, documentation and Agent Care handover.

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 campaign, one office ProductionProduction telephony AdvancedMulti-office campaigns

Introduced at Pilot

Calling to your consent rules✓✓✓

Broker approval✓✓✓

Call-level tracing and baseline✓✓✓

Introduced at Production

Per-state contact reporting—✓✓

Approval workflow in your systems—✓✓

Write-back into your dialer and CRM—✓✓

Trace export to your stack—✓✓

Introduced at Advanced

Multi-brand and multi-office rollout——✓

Multi-stage brokerage approvals——✓

High call volume——✓

Enterprise call controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls 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 lead sources and consent artefacts → Lead-record ingestion and consent linkingWeek 1

02Representative approved scripts and cadences → Calling baseline, script mapping and disclosure sequencingWeek 2

03Your written do-not-call policy and calling rules → Consent, suppression and state calling-rule mappingWeek 1

04Access to relevant APIs, feeds or exports → Telephony, CRM and consent-source assessment, then integration setupWeek 2

05Attempts that drew complaints → Evaluation suite, regression cases and failure-mode testingWeek 4

06The rules that decide when a number may be dialled → Confidence scoring, compliance routing, guardrails and approval controlsWeek 3

07A named supervising broker to release campaigns → Broker approval workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Phases occupy the weeks they actually run. Week 5 carries two bars because evaluation and pilot overlap.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Consent provenance, calling rules and where the boundary sits W2Consent, suppression and telephony data in place W3Calling workflow, confidence logic and approval controls W4Consent-artefact tests, recording-state checks and failure modes W5Telephony connected, one campaign run under supervision, nothing autonomous W6One campaign completed under broker release, then monitoring passes to Agent Care

Reading the bandNothing is padded to fill a week. Evaluation and pilot both land in week 5, which is why it doubles.

At the end of W6The last checks clear and monitoring moves to Agent Care.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Real Estate AI agent

Build a voice follow-up agent around your brokerage's consent record.

Show us your lead sources, your consent records and your calling rules. Whoever owns consent at your brokerage keeps the approval, and we map the calling workflow around them.

Discuss this agent →

Nestack Agents · Lead-qualification voiceAGT-RE-06 · Agent Care available after launch