Main site

Marketing AI agent · campaign sends

Email & SMS Campaign AI Agent

Refuse the send the record cannot support: stage the audience, test it against suppression and the frequency cap at the hour of release, and leave that release to a named marketer.

Typical delivery
4–6 weeks
Deployment
Your stack
Named marketer
Send-time test
After launch
Agent Care

What this agent does

Stages the send, and holds the release

In

  1. 01

    A list is uploaded, and a contact who unsubscribed in March can re-enter the audience in June.

  2. 02

    A send is staged, and the audience meets suppression at that moment, not when the list was built.

Reason

  1. 03

    A segment is assembled from inferred attributes, and it says on its face that it was inferred.

  2. 04

    A send leaves, and there is no recall — which is why the release gate is the whole product.

  3. 05

    An email and a text draw on one audience, and the cap you set is applied across both channels.

Decide

  1. 06

    A footer is drafted, and 16 CFR Part 316 wants a working opt-out and a physical postal address.

  2. 07

    A bounce signal moves, and it describes the state of a list rather than granting any permission.

Out

  1. 08

    A sync runs overnight, and each synced list is a fresh re-entry risk until suppression clears it.

  2. 09

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

Product statement

The agent assembles the audience, tests it against suppression and the frequency cap at send time, and drafts the send. A named marketer releases it — nothing in CAN-SPAM is signed or certified.

Example workflow

One send, audience to release

  1. 1Send request receivedCampaign briefs, segment definitions, uploaded lists and platform syncs
  2. 2Audience context assembledThe segment, the list it came from, the channels it reaches and the hour of release
  3. 3Send package draftedThe audience, its suppression state, the caps it consumes and completeness
  4. 4Controls appliedSuppression checks, frequency checks, footer checks and completeness confidence

No human action required

Stages 1 to 4 run unaided, and nothing is released at any of them — the agent is staging, and the compliance lane opens at the completeness gate.

5DecisionSplits at the completeness gate
Evidence sufficient

Goes to the named marketer to release.

Anything thin

Adds a compliance analyst read first.

Compliance review

The send is held with its audience, its suppression state and the caps behind it.

Release · Append evidence · Send to compliance

Released — by a named marketer

  1. 6Suppression and send records updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedSuppression integrity, cap adherence, analyst corrections and what the read found

Corrections

Every analyst correction is carried into the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Releasing a send to a live audience.
    • Deciding a revocation was validly made.
    • Certifying anything about a consent record.
    • Setting the frequency cap a brand sends under.

  • Automation boundary

    Agent acts unaided

    • ✓Assemble the audience from the segments your brief actually names.
    • ✓Re-test the staged audience against suppression.
    • ✓Draw email and text against one shared frequency budget per contact.
    • ✓Surface the bounce, complaint and authentication.

    Nothing is released or sent except by a named person, inside the agreed boundaries.

    • Judging whether a suppression file is complete.
    • Answering a regulator about a past campaign.
    • Deciding which audience a campaign may reach.
  • Changes to suppression, audience or send records.

Example output

One send, annotated

Our sibling consent and revocation evidence agent owns the consent record, the revocation intake, the honouring clocks and the litigation evidence file; this page assembles no consent evidence and adjudicates no revocation — it refuses to send what the record cannot support.

Send record · single release

Illustrative example
Send
Reactivation send, uploaded list
Recorded as
Tested against suppression at staging
Channel
Email and SMS
Evidence of record
List import, 3 August 2026
Confidence
Held unreleased
Held for
The releasing marketer, by name

As received

Taken from the import log and the suppression file — it reaches as far as those sources do.

What the record holds

  • Suppression check log
  • Import provenance
  • Frequency budget trail

Why no send here

Whether an audience may be sent at all is a marketer call, not a model output.

What the score decides

Below the configured threshold a send picks up a compliance read before the marketer sees it.

Action

ReleaseAppend evidenceSend to compliance

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 send

From the list it drew on

03Evidence

Where the evidence is used

Our advertising pages are agency-side and bill a client for the media; this one answers to the brand legal, privacy and finance functions that live with the list long after a campaign ends.

01Approved path

The send is the last gate

A suppression list is only as good as the last import, so an assembled audience is tested at the moment of release rather than when the list was built.

02Human review

What was checked, and not found

The 1 August 2026 amendment shown against the CAN-SPAM rule was not read back to primary text; nothing under it is signed, filed or certified, so no filing record exists to inspect; no authority publishes a bounce, complaint or authentication threshold that would make a list fit to send; and no sending platform documents how its own list sync behaves against an existing suppression file.

04

Build an evidence trail

The send, the audience it drew on and the marketer who released it stay on file.

Integrations

Typical integrations

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

  • Sending platforms

    • Email service providers
    • SMS gateways
    • Send and suppression logs

  • Audience sources

    • CRM
    • CDP
    • uploaded and synced lists
    • Segment and list records

  • Deliverability signals

    • Bounce, complaint and authentication feeds
    • List health records

Agent

Email and SMS campaign

  • Assembles the audience
  • Tests it at send time
  • Holds for the marketer

  • Records and case systems

    • Preference centre
    • ticketing
    • Cap and release records

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

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

Agent controls

Six shutters between the model and the marketer

Six shutters closing in order, the last the tightest. What still gets out is set out in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to audience assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and suppression rules; a rule amended after the last import is the one that lets a contact back in.Track
L4TraceabilityRecord each send, the audience behind it, the suppression state it met and every read of that record.Record
L3Marketer releaseHold the send for a named marketer; the hold governs release, not whether the underlying permission was ever valid.Gate
L2Send guardrailsTest each staged send against your suppression file, your frequency cap and the footer mechanics 16 CFR Part 316 sets: a working opt-out and a physical postal address.Restrict
L1Confidence thresholdsRoute a thin audience to a compliance read first; a segment built on inferred attributes has to declare the inference.Require review
Model coreEvidence assembled — the audience, its suppression state, the caps and completeness
L1 – L2
Test whether a send may go
L3
Puts the release in a person's hands
L4 – L5
Keep the send and the audience behind it
L6
Holds the send unreleased when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the send package that comes out.

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

Surface — the message a recipient receives

  1. E1Final-output evaluationDid the send reach only the audience it was cleared for?
  2. E2Step-level evaluationDid the agent read the right list, the right cap and the live suppression state?
  3. E3Tool evaluationDid it read and write the correct audience and the correct suppression file?
  4. E4Confidence calibrationDo low-confidence audiences actually attract more compliance corrections?
  5. E5Slice evaluationHow does performance change across specific list sources?
  6. E6Business outcomeHow many sends needed a correction before the marketer released?

Floor — the audience the brand answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed at the stage where it first shows.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • KZ-03

      Stale audience read

      The audience read is not the one now staged.

    Stage gathers

    The lists, the contacts, the caps and the sends

  2. 02 · Reasoning

    2 modes

    • KZ-04

      Suppression assumed, not tested

      An audience moves on without a send-time check.

    • KZ-06

      Cap read from one channel

      Email and text are budgeted apart, not together.

    Stage proposes

    The audiences, their caps and suppression state

  3. 03 · Tool / write

    2 modes

    • KZ-02

      Thin audience passed forward

      A staged send skips the compliance read.

    • KZ-05

      Bound to the wrong list

      A suppression entry lands on the wrong contact.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • KZ-01

      Released, audience unrecorded

      A send sits on file with no audience behind it.

    Stage returns

    The send a marketer releases and a recipient opens

  5. 05 · Change / Version

    1 mode

    • KZ-07

      Silent suppression regression

      A platform change moves the audience, not the record.

    Stage tracks

    Model, prompt, suppression rules and send fields

  • Sev-1 · a send to a suppressed contact
  • Sev-2 · wrong audience reaches the send
  • Sev-3 · source degrades, send held back
See the complete Marketing failure-mode catalogue

Affected slices

Uploaded lists absorb the corrections

A send-level suppression-integrity figure can read clean while uploaded and synced lists carry most of the rework. Nestack reports the correction rate by list source, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Uploaded and synced lists10.9%3.6×Review
Partner and co-registration feeds7.8%2.6×Review
Behaviour-triggered sends4.9%1.6×Watch
Native sign-up capture1.8%0.6×Normal

Bar: correction-rate lift vs. native sign-up baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What an unsuppressed send costs

A loop ends when the resurrected contact has become a case the next release must pass. That suite is what the next send cleared is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on uploaded and synced lists.

  2. 02

    Diagnose

    The contact who unsubscribed in March and reappeared in an uploaded list in June is traced back until a single cause is left.

  3. 03

    Improve

    Number the change; the sends that drove it are filed underneath it.

  4. 04

    Verify

    Each touched send case is run once more, and one red holds it back.

  5. 05

    Learn

    It stays on as a standing test, and the suppression rules travel with it.

  6. Learn → Detect

    The return edge. The next send 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, audience assembly, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Suppression-integrity discovery and boundary work.Week 1
  2. 02List, platform and delivery sources.Week 1, Week 2
  3. 03Send-to-suppression and frequency-budget mapping.Week 1, Week 2
  4. 04Audience and suppression ingestion.Week 2
  5. 05Contact, list and send binding.Week 2, Week 3
  6. 06Completeness scoring and review routing.Week 3
  7. 07Marketer release workflow.Week 3, Week 4
  8. 08Send-platform and CRM integration.Week 5
  9. 09Suppression and footer cases.Week 4, Week 5
  10. 10Guardrails and send controls.Week 4
  11. 11Send-trail instrumentation.Week 5, Week 6
  12. 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 tierPilotOne list source, one send yearProductionProduction send workflowAdvancedMultiple brands / regions
Introduced at Pilot
Audience assembly to your lists✓✓✓
Named marketer release✓✓✓
Addressable-audience baseline✓✓✓
Introduced at Production
Reporting by send—✓✓
Release workflow in your systems—✓✓
Approved write-back—✓✓
Platform-and-list integration—✓✓
Introduced at Advanced
Multi-region estates——✓
Cross-channel release packs——✓
Large sending volumes——✓
Multi-channel frequency controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, send 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.

  1. 01Your sending platforms and the lists behind them

    List inventory mapping and suppression capture

    Week 1

  2. 02Representative sends, imports and suppression files

    Record binding, cap logic and the suppression baseline

    Week 2

  3. 03Your frequency caps and the channels they cover

    List mapping, cap binding and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Platform, list and deliverability source assessment, then integration setup

    Week 2

  5. 05Sends you would not want recalled

    Suppression cases and the evaluation run

    Week 4

  6. 06What no send record may settle

    Completeness scoring, review routing, guardrails and release controls

    Week 3

  7. 07A named marketer to release the send

    Release 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 widths are weeks of actual work rather than layout, which is why one band has to carry two.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Campaign workflow discovery, list mapping and the automation boundary

W2Source integration and the suppression-integrity baseline

W3Audience assembly, cap logic and release controls

W4Evaluation suite, suppression cases and failure-mode testing

W5Platform integration, pilot sends and targeted corrections

W6One campaign year run under the lifecycle lead, then Agent Care handover

Reading the band
Each bar covers only the weeks its own work is named for. The fifth takes a pair because the work does.
At the end of W6
When the send record validates, Agent Care takes the agent on.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Marketing AI agent

Build a campaign send agent around the contact who unsubscribed in March and came back on a list in June.

Show us one staged send and the list it drew on. A named marketer releases it, never the agent. Suppression is tested at the hour of release rather than at list build, and nothing in CAN-SPAM is signed or certified. Consent capture and revocation intake are a different agent.

Typical delivery
4–6 weeks
Deployment
Your audiences
Named marketer
Send-time test
After launch
Agent Care available

Nestack Agents · campaign sends

AGT-MK-04 · Agent Care available after launch

Build a campaign send agent

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

About you
Your audiences

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