Member-Retention & Churn-Prediction AI Agent

What this agent does

Scores the risk, never releases the offer

Attendance, billing and contract signals ingested from supported club-management, billing or CRM sources.

Field labels and formats normalised, with each signal carried forward with the system it came from.

Reason

A retention-risk score per member, built from the input set configured and disclosed with the operator.

Cohorts and thresholds applied as the operator configured them, per brand and per state.

Health-adjacent inputs marked, and the features behind each score recorded for a person to read.

Decide

A cancellation request anywhere in the estate, which takes that member out of the offer queues.

Offer copy drafted against the approved set and held for the named retention manager.

Out

The score, the features, the offer and the release retained against the member record.

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

→ Product statement

The agent scores and drafts; the retention manager decides whether a member is contacted, and a member who has asked to cancel is out of scope.

Example workflow

One member, signal to release

AgentHuman

1Member record receivedClub-management record, billing history, contract state or freeze log

2Signals assembledVisits, billing events, tenure and contract state, each with the system it came from

3Risk scoredA risk band, the features that moved it, the cohort and confidence

4Controls appliedCancellation suppression, configured state offer rules, protected-attribute checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing reaches a member at any of them — the agent is scoring, and the retention manager's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold

High confidence

Goes to the retention manager to release.

Low confidence

Adds a compliance read first.

Retention-manager approval

The offer is held with its score, the features behind it and the rules that applied.

Approve · Amend · Send to compliance review

Approved — released to send▼

6Member systems updatedOnly where write access and approval policy allow it

7Outcome evaluatedOffer take-up, suppression failures, cancellations after contact and complaints

Amendments

Every retention-manager amendment is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Releasing any save offer to a member.

Contacting a member who has asked to cancel.

Deciding whether an offer may be sent in a state.

Changing a member's price, term or billing.

Automation boundaryAgent acts unaided

✓Score retention risk from configured attendance and billing signals.

✓Carry each score forward with the features behind it.

✓Sort scored members into the configured retention cohorts.

✓Draft offer copy, flag the exposure and hold it for a person for the named owner.

Any write happens inside the boundaries agreed at implementation, never ahead of release.

Wording or placing an algorithmic-pricing notice.

Adjudicating a statutory cancellation ground.

Adding a new signal to the scoring inputs.

Changes to thresholds, offer rules or suppression.

Example output

One score, annotated

Everything the agent scores is attached to the signals it was built from.

Retention output · single memberIllustrative example

Member

Scored driver

Risk band

Source of record

Confidence

Offer basis

Twelve-month member

Visit frequency down against this member's own twelve-month pattern

Elevated

Attendance and billing

88%

Standard configured offer

As receivedTaken from the club's own attendance and billing records — the score sits on the other side.

Features used Visit-frequency change Billing event history Contract state and tenure

Why this scoreIt ranks a hypothesis about a member who has not decided and a person still decides.

ActionApproveAmendSend to compliance review

What the score decidesBelow the configured threshold the offer picks up a compliance read before 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

Every scored memberFrom the club systems

03Scoring

Score from your own data

Draw on the attendance and billing signals the operator configured, with the input set documented.

01Approved path

A score is not a decision

Routine cohorts arrive already ranked, with the features attached.

02Human review

Send review to the exposed offers

Suppressed members and low-confidence scores are marked, so the retention manager's read starts where exposure concentrates.

04Build an evidence trail

The score, the features behind it and the offer released stay on the member record.

Integrations

Typical integrations

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

Club managementMindbody · ABC Ignite
Glofox · Zen Planner

Billing and duesDirect debit · card on file
Dues processing · payment APIs

Member CRMClub OS · Keepme
Salesforce · HubSpot

Agent

Member retention & churn scoring

Reads the signals
Scores the risk
Holds for release

Messaging and offersBraze · Klaviyo
Twilio · consent records

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 score and the member

The controls are nested. What survives all of them appears in the map underneath.

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

L6Rollback / safe modePull retention back to scoring-only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, input-set and offer-rule changes.Track

L4TraceabilityRecord the signals, the score, the features, the suppression and the release.Record

L3Manager releaseHold offers for the named manager; it governs release, not whether an approved offer is right.Gate

L2Policy guardrailsTest each offer against the suppression list and the configured state rules; a failure returns it.Restrict

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

Model coreScore produced — the risk band, the features behind it, the cohort and confidence

L1 – L2Test whether an offer may stand

L3Puts the release in a manager's hands

L4 – L5Show what the score was built from

L6Holds offers for staff when signals degrade

How Nestack evaluates it

Evaluate the retention workflow — not only the score.

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

Surface — the offer the member is sent

Depth of coverage ▼

E1Final-output evaluationDid each released offer match the rules configured for that member?

E2Step-level evaluationDid the agent use the right signals, cohorts and suppression state?

E3Tool evaluationDid it read and write the correct member and the correct field?

E4Confidence calibrationDo low-confidence scores actually attract more manager amendments?

E5Slice evaluationHow does performance change across specific member cohorts?

E6Business outcomeHow many offers were amended, suppressed late or complained about after sending?

Floor — whether the member stays or goes

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed where it starts.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

RV-03

Stale attendance feed

Visits recorded late are read as absence.

Stage gathersAttendance, billing, contract state and suppression flags

02 · Reasoning2 modes

RV-04

Health-derived feature

An input describing physical health enters the score unmarked.

RV-06

Proxy for a protected trait

A signal stands in for a characteristic an offer may not turn on.

Stage proposesA risk band, the features behind it and confidence

03 · Tool / write2 modes

RV-02

Late suppression

An offer is released before the cancellation request reaches the agent.

RV-05

Duplicate save contact

One member is reached twice from two retention cohorts.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

RV-01

Offer outside the state rules

An unsolicited offer is sent where consent was required.

Stage returnsThe offer the member receives and answers

05 · Change / Version1 mode

RV-07

Silent input-set drift

A model or rule change widens what the score is built from.

Stage tracksModel, prompt, input set and offer rules

Sev-1 · an offer sent outside the boundary Sev-2 · a suppressed member is contacted Sev-3 · a signal degrades, the score routes to review

See the complete Sports and Fitness failure-mode catalogue →

Affected slices

Overall suppression can hide one cohort

In aggregate, the rate at which an offer reached a member it should not have looks settled. In one cohort it is not. Nestack reports that rate by slice, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Members inside a cancellation request	6.4%	3.3×	Review
Consent-restricted states	5.3%	2.7×	Review
Paused or frozen contracts	2.7%	1.4×	Watch
Standard monthly members	1.2%	0.6×	Normal

Bar: suppression-failure lift vs. standard-monthly baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

Every cycle leaves one more case

Explaining a bad save does not close the cycle. A standing regression case does, and that suite is what the next scored member is measured against.

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

01Detect

Suppression failures rise in a member cohort.

02Diagnose

Not the score. The features behind it — read until the cause narrows to one signal or rule.

03Improve

The fix is versioned against the scores that produced it.

04Verify

The release stops until every affected case is passing.

05Learn

The case stays in the suite, and the offer policy is amended.

Learn → DetectThe return edge. The next detection runs against a longer suite than this one.

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, scoring workflow, evaluation, integration, then production validation and handover.

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

01Retention workflow discovery and boundary definition.

02Club-management and billing assessment.

03Input-set, cohort and state offer-rule set mapping.

04Signal ingestion and normalisation.

05Scoring logic and feature attribution.

06Confidence scoring and suppression routing.

07Retention-manager approval workflow.

08Billing and messaging-system integration.

09Offer-eligibility cases.

10Guardrails and offer controls.

11Score-trail instrumentation.

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 club, one billing system ProductionProduction member systems AdvancedMultiple brands / markets

Introduced at Pilot

Scoring to your inputs and cohorts✓✓✓

Retention-manager sign-off✓✓✓

Score-quality baseline✓✓✓

Introduced at Production

Reporting by member cohort—✓✓

Approval workflow in your systems—✓✓

Approved write-back—✓✓

Billing-system integration—✓✓

Introduced at Advanced

Multi-system member estates——✓

Multi-stage retention approvals——✓

High member volume——✓

Multi-brand retention 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 retention playbook and approved offer set → Input-set, cohort and offer-rule mappingWeek 1

02Representative member and billing history → Scoring baseline, feature attribution and cohortingWeek 2

03The signals you are willing to score on → Offer-rule mapping and the automation-boundary definitionWeek 1

04Access to relevant APIs, feeds or exports → Club-management, billing and messaging assessment, then integration setupWeek 2

05Offers that should never have gone out → Eligibility cases and the evaluation suiteWeek 4

06Where a score has to stop and wait for a person → Confidence scoring, suppression routing, guardrails and release controlsWeek 3

07A named retention manager to release offers → Retention-manager 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 the work really needs, so week 5 runs evaluation and launch side by side.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Retention workflow discovery, offer-rule mapping and the automation boundary W2Source integration and the scoring baseline W3Scoring workflow, confidence logic and release controls W4Evaluation suite, suppression cases and failure-mode testing W5Messaging integration, pilot cohorts and targeted corrections W6One billing cycle scored under the retention team, then handover

Reading the bandEach bar spans only the weeks its work is named in. The week 5 overlap is real, not padding.

At the end of W6The cycle closes validation and Agent Care picks up the monitoring.

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

Next step · Sports & Fitness AI agent

Build a retention agent that stops at the cancellation request.

Show us the signals you score on, the offers you make and the states you sell in. The named retention manager signs the offers that reach members. A score built from attendance is health-adjacent data, so the input set is agreed first.

Discuss this agent →

Nestack Agents · Member retention and churn predictionAGT-SF-06 · Agent Care available after launch