Customer Success AI agent · Expansion signals
Expansion Signal AI Agent
Name what an account did — usage measured against the entitlement it actually holds — and let the motive stay unnamed, then hand the signal to the growth manager who qualifies it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Growth manager
- Usage first
- After launch
- Agent Care
What this agent does
Raises the signal, never qualifies it
In
- 01
A signal is raised, and the usage under it travels with it rather than being summarised away.
- 02
Usage crosses a line, and the line it crossed is named beside the entitlement record it came from.
Reason
- 03
A signal says what was observed, and it stops there instead of guessing at why anyone did it.
- 04
Usage climbs on retries, and the signal says so rather than reading a failing job as appetite.
- 05
An entitlement record is old, and the signal carries the day it was read instead of implying it is live.
Decide
- 06
A signal is raised on a trial, and the terms the customer was actually given are carried beside it.
- 07
Usage lands under a subsidiary, and the signal names the legal entity it was measured on.
Out
- 08
A signal cannot be attributed to anyone, and that is written up as a gap rather than assigned to a team.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
Watching entitlement against use, raising the signal and showing the unknowns belong to the agent. Deciding an account is ready to buy belongs to a named growth manager, who qualifies it and owns that call.
Example workflow
One signal, usage to qualification
- 1Usage and entitlement receivedProduct telemetry, seat and API records, entitlement terms, order forms and billing history
- 2Entitlement fixed and datedThe limits the contract sets, the limits the system enforces and the day each record was read
- 3Signal raisedThe observation, the usage under it, the entitlement it was measured against and confidence
- 4Controls appliedEntitlement checks, retry checks, attribution checks and signal confidence
No human action required
Stages 1 to 4 run unaided, and no customer is approached at any of them — the agent is observing, and the growth manager lane opens at the entitlement gate.
Goes to the named growth manager to qualify.
Adds a senior growth read first.
Growth manager review
The signal is held with its usage, the entitlement it was measured against and what it cannot say.
Accept · Annotate · Send to senior review
Accepted — by the named growth manager
- 6Account and expansion records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedEntitlement accuracy, attribution quality, manager corrections and what review found
Corrections
Each growth manager correction is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Qualifying a signal as an opportunity.
- Pricing anything an account might buy.
- Contacting a customer about their usage.
- Deciding an account is ready to buy more.
Automation boundary
Agent acts unaided
- ✓Raise the signal and carry the usage under it.
- ✓Separate usage that came from retries from usage that came from people.
- ✓Mark the day each entitlement record was read.
- ✓State plainly what the signal cannot say about why usage moved.
No customer is approached except by a named growth manager, inside the agreed boundaries.
- Judging why usage actually moved.
- Telling a seller an account will expand.
- Choosing which limit an account is sold up to.
- Changes to the thresholds, terms or entitlements.
Example output
One expansion signal, annotated
This serves a growth team who may have to explain to an account team why a signal was raised at all; below is one signal exactly as the agent leaves it.
Expansion signal · single account
Illustrative example- Account
- Mid-market, seat-based plan
- Recorded as
- Raised, entitlement dated
- Usage
- Usage carried
- Evidence of record
- Entitlement read 3 August 2026
- Confidence
- Held unqualified
- Held for
- The named growth manager
As received
Taken from the usage recorded and the entitlement it was measured on, and it concludes nothing beyond that.
What the record holds
- Seat records
- API call volume
- Entitlement terms
Why no approach here
Calling a signal an opportunity is a judgement the manager makes.
What the score decides
Below the configured threshold a signal gets a senior read before the manager sees it.
Action
AcceptAnnotateSend to senior reviewValue
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 signal
From the usage under it
Where the entitlement comes from
An ancillary-offer agent in travel and hospitality presents an offer to a guest at the point of sale. This raises a signal for an internal team and presents nothing to anyone.
Usage is not intent
An account near a limit may be buying more, leaving, hearing from its own procurement team or refactoring what it built, and the usage reads alike in all four.
What was checked, and not found
Checked and not found: no statute or published standard governs when a vendor may read its own telemetry for commercial ends, so there is no external rule to point at and this page claims none.
Build an evidence trail
The signal, the usage under it and the manager who qualified it stay together.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Product telemetry
- Pendo
- Amplitude
- Snowplow
- Seat activity and API call volume
Entitlement and billing
- Zuora
- Chargebee
- Stripe
- Plan limits and overage terms
CRM and account records
- Salesforce
- HubSpot
- Gainsight
- Contract terms and account tiering
Agent
Expansion signal detection
- Reads the usage
- Raises the signal
- Holds for the manager
Support and ticketing
- Zendesk
- Statuspage
- Jira
- Throttling errors and limit complaints
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six assays between the model and the manager
Six assays in sequence, the last the sharpest. What assays clean is drawn in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a signal may stand
- L3
- Leaves the qualification to a manager
- L4 – L5
- Keep the signal and the usage behind it
- L6
- Raises no signal at all when signals degrade
How Nestack evaluates it
Evaluate the whole detection — not only the signal that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the signal a manager works
- E1Final-output evaluationDid the signal record the usage and the entitlement it was actually built on?
- E2Step-level evaluationDid the agent read the live entitlement, the right period and the right account?
- E3Tool evaluationDid it read and write the correct account and the correct signal?
- E4Confidence calibrationDo low-confidence signals actually attract more manager corrections?
- E5Slice evaluationHow does performance change across specific product lines?
- E6Business outcomeHow many signals needed a correction before the manager accepted?
Floor — the usage a signal rests on
Failure modes
Where each failure originates in the agent
Seven failure modes, each set at the stage where it first surfaces.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- NT-03
Stale entitlement read
The record read is not the one now in force.
Stage gathers
The usage, the terms, the seats and the dates
02 · Reasoning
2 modes
- NT-04
Signal raised on a retry storm
Usage climbs because a job keeps failing.
- NT-06
Motive asserted, not observed
A signal states why, having seen only what.
Stage proposes
The usage, the entitlement and what is unknown
03 · Tool / write
2 modes
- NT-02
Thin signal passed forward
A signal moves on without the senior read.
- NT-05
Signal bound to wrong entity
The usage is filed against another subsidiary.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- NT-01
Ceiling hit, nobody was buying
One script against the API filled the seats.
Stage returns
The signal a manager works and a seller acts on
05 · Change / Version
1 mode
- NT-07
Silent threshold drift
A limit moves while the stored signal keeps the old one.
Stage tracks
Model, prompt, threshold rules and term changes
- Sev-1 · the agent approaches a customer
- Sev-2 · a stale signal reaches a seller
- Sev-3 · usage degrades, no signal raised
Affected slices
API-driven accounts absorb the corrections
A product-level qualification figure can read clean while accounts driven by API use carry most of the corrections. Nestack reports the correction rate by product line, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Accounts with API-driven use | 8.6% | 3.7× | Review | |
| Multi-entitlement estates | 6.1% | 2.6× | Review | |
| Accounts inside a trial | 3.8% | 1.6× | Watch | |
| Established single-plan accounts | 2.1% | 0.9× | Normal |
Bar: correction-rate lift vs. single-plan-account baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What a scripted seat ceiling costs
The loop shuts when the signal raised on nothing is a standing case. That suite is what the next signal qualified is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Correction rate rises on accounts with API-driven use.
- 02
Diagnose
The team that hit the seat ceiling because one person wrote a script against the API is worked backwards until one cause is left standing.
- 03
Improve
A change goes out numbered, and the signals that forced it travel with it.
- 04
Verify
Nothing ships while one signal case remains red.
- 05
Learn
The case is kept, and the qualification rules are amended in that same commit.
Learn → Detect
The return edge. The next signal raised 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, signal detection, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Entitlement-inventory and automation-boundary work.Week 1
- 02Telemetry, entitlement and billing sources.Week 1, Week 2
- 03Usage-to-entitlement and threshold-source mapping.Week 1, Week 2
- 04Usage ingestion and entitlement checks.Week 2
- 05Entitlement, threshold and account binding.Week 2, Week 3
- 06Attribution scoring and review routing.Week 3
- 07Growth manager review workflow.Week 3, Week 4
- 08Entitlement-system integration.Week 5
- 09Signal and qualification cases.Week 4, Week 5
- 10Guardrails and qualification controls.Week 4
- 11Signal-trail instrumentation.Week 5, Week 6
- 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 tier | PilotOne product line, one quarter | ProductionProduction growth workflow | AdvancedMultiple products / regions |
|---|---|---|---|
| Introduced at Pilot | |||
| Signal detection to your entitlements | ✓ | ✓ | ✓ |
| Named growth manager release | ✓ | ✓ | ✓ |
| Entitlement-and-usage baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by product | — | ✓ | ✓ |
| Manager review workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Entitlement-system integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-entitlement rule sets | — | — | ✓ |
| Cross-product signal packs | — | — | ✓ |
| Large account estates | — | — | ✓ |
| Multi-entitlement signal controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, entitlement 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.
You bring
We build with it
01Your live accounts and the entitlement each holds
→Usage capture and entitlement checks
Week 102Representative accounts, including ones that churned
→Entitlement binding, threshold logic and the signal baseline
Week 203Your expansion cadence and the managers it names
→Entitlement mapping, threshold review and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Telemetry, entitlement and billing source assessment, then integration setup
Week 205Signals you would not want qualified
→Qualification cases and the evaluation run
Week 406What no signal may promise
→Attribution scoring, review routing, guardrails and release controls
Week 307A named growth manager who qualifies it
→Release to the named growth manager, 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
Six weeks measured out rather than six columns ruled; the fifth doubles up because the work does.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Entitlement discovery, threshold review and the automation boundary
W2Source integration and the entitlement-and-usage baseline
W3Signal detection, attribution logic and release controls
W4Evaluation suite, signal cases and the qualification run
W5Growth-platform integration, pilot signals and targeted corrections
W6One expansion quarter run under the growth lead, then Agent Care handover
- Reading the band
- A bar covers only the weeks its own work is named for, and week five carries two by design.
- At the end of W6
- When the signal record validates, Agent Care assumes the agent.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Customer Success AI agent
Build an expansion signal agent around the ceiling your last account hit without telling anyone.
Show us one product line and the entitlement each account on it holds. Which of them sits near a limit nobody has told them about — and if nobody can answer that today, the limit is being watched by you and not by them. A signal that names a motive comes back as a finding.
- Typical delivery
- 4–6 weeks
- Deployment
- Your product line
- Growth manager
- Usage first
- After launch
- Agent Care available
Nestack Agents · Expansion signals
AGT-CX-05 · Agent Care available after launch