Customer Success AI agent · QBR preparation
QBR Preparation AI Agent
Source each figure where it was produced, put its as-at date on the slide with it, mark whatever moved since last quarter, and hold the pack for the account manager who presents it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Account manager
- Traceable pack
- After launch
- Agent Care
What this agent does
Assembles the pack, never presents it
In
- 01
A pack is assembled, and each figure on it arrives with the system it came from and the day it was true.
- 02
A number is placed on a slide, and the source behind it is recorded there rather than remembered.
Reason
- 03
A figure has moved since it was last shown, and both readings appear with the reason the earlier one changed.
- 04
A chart is carried over from a previous pack, and it stays unpopulated until its source is read again.
- 05
A comparison period is chosen, and the rule that chose it is stated rather than left to the assembly.
Decide
- 06
A slide rests on a system the customer cannot open, and that limit is marked on the slide itself.
- 07
An export ages past its term, and each slide leaning on it is flagged for re-sourcing before assembly.
Out
- 08
A deck is quoted back a quarter later, and the pack still holds the source and the date behind each figure.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
Assembly, sourcing and restatement marking belong to the agent. What the customer is told belongs to a named account manager, who chooses the narrative and sends the pack.
Example workflow
One QBR pack, sources to sign-off
- 1Source material receivedProduct telemetry, support records, billing systems, CRM and the last pack shown
- 2Figures sourced and datedEach figure, the system of record it came from, the day it was true and the pack it last appeared in
- 3Slides assembledThe slides, the figures on them, what moved since last time and confidence
- 4Controls appliedSource checks, as-at date checks, restatement checks and sourcing confidence
No human action required
Stages 1 to 4 run unaided, and nothing reaches a customer at any of them — the agent is assembling, and the account manager lane opens at the sourcing gate.
Goes to the named account manager.
Adds a reporting review first.
Manager review
The pack is held with its sources, the dates behind them and what moved since the last review.
Approve · Append source · Send to reporting review
Approved — by the account manager
- 6Reporting and CRM records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedSource coverage, restatement rate, manager corrections and what review found
Corrections
Each 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
- Deciding what the customer should be told.
- Choosing the narrative the pack argues.
- Sending the deck to the customer.
- Signing off a figure quoted in a contract.
Automation boundary
Agent acts unaided
- ✓Assemble the pack from the sources of record.
- ✓Mark any figure that has moved since it was last shown.
- ✓Carry the source and the as-at date onto each figure on a slide.
- ✓Flag what could not be sourced rather than leaving it off the pack.
Nothing reaches a customer except by a named account manager, inside the agreed boundaries.
- Judging whether an account is doing well.
- Explaining a restated number to a customer.
- Choosing which metrics the review covers.
- Changes to the pack, the sources or the rules.
Example output
One QBR slide, annotated
This serves an account team who will have the numbers in this pack quoted back at them a quarter later; below is one slide exactly as the agent leaves it.
QBR pack output · single slide
Illustrative example- Slide
- Quarterly review, one account
- Figure shown
- Adoption across licensed teams
- Account
- Restated, both shown
- Source of record
- Product telemetry, 3 August 2026
- Confidence
- Held, not sent
- Held for
- The named account manager, by name
As received
Built from one telemetry export and one billing record, and it claims nothing about the account beyond them.
What the slide holds
- Product telemetry
- Billing record
- Support history
Why nothing is sent
Deciding what a customer is told is a call the account manager makes.
What the score decides
Below the configured threshold a slide gets a reporting read before the manager sees it.
Action
ApproveAppend sourceSend to reporting 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 QBR pack
From the sources of record
Where the pack is used
The operational-reporting agent in Operations reports a number to the company that owns it; this pack is shown to the customer, who checks it against figures of their own.
The deck is not the account
A pack is a selection — which metrics appear, which period is shown, which comparison is drawn — and each of those choices is set by you, not baked in.
What was checked, and not found
No figure here is presented as one the customer has agreed: their own reporting may compute the same measure on a different basis, and where the two do not meet, the gap is reported rather than a single number neither side supports.
Build an evidence trail
The slide, the number on it and the source that produced 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
- Usage events
- entitlement data
- Adoption by team and by feature
Support and ticket systems
- Zendesk
- Salesforce Service
- Ticket volumes and outcomes
Billing and contracts
- Billing platform
- contract records
- Entitlements, terms and invoices
Agent
QBR assembly and sourcing
- Reads the sources
- Assembles the pack
- Holds for the manager
Reporting and presentation
- Data warehouse
- BI dashboards
- Templates and the packs already shown
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six presses between the data and the deck
Six presses in a row, the last the heaviest. What holds shape is set out in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a slide may stand
- L3
- Leaves the sending to a named manager
- L4 – L5
- Keep the slide and the source behind it
- L6
- Leaves the slide unpopulated when signals degrade
How Nestack evaluates it
Evaluate the whole assembly — not only the pack that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the pack a customer sits through
- E1Final-output evaluationDid each figure carry the source it was actually drawn from?
- E2Step-level evaluationDid the agent read the right system, the right period and the live sourcing rules?
- E3Tool evaluationDid it read and write the correct account and the correct period?
- E4Confidence calibrationDo low-confidence slides actually attract more manager corrections?
- E5Slice evaluationHow does performance change across specific accounts?
- E6Business outcomeHow many slides needed a correction before the manager approved?
Floor — the source a figure rests on
Failure modes
Where each failure originates in the agent
Seven failure modes, each placed at the stage that first reveals it.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- NN-03
Stale export read
The export read is older than the live source.
Stage gathers
The systems, the exports, the periods and the dates
02 · Reasoning
2 modes
- NN-04
Figure asserted, not sourced
A number appears with no source behind it.
- NN-06
Comparison chosen to flatter
The period shown is the one that reads best.
Stage proposes
The slide, the figures on it and their sources
03 · Tool / write
2 modes
- NN-02
Thin sourcing passed forward
A slide moves on without the second read.
- NN-05
Figure bound to wrong account
The number is filed against another account.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- NN-01
Shown, restatement unrecorded
The deck shows the new figure but not the change.
Stage returns
The pack a customer reads and later quotes back
05 · Change / Version
1 mode
- NN-07
Silent chart carry-over
A chart is reused while its source has moved underneath it.
Stage tracks
Model, prompt, sourcing rules and pack dates
- Sev-1 · a figure shown with no source
- Sev-2 · a restated number reaches the deck
- Sev-3 · source unclear, slide held back
Affected slices
Multi-product accounts absorb the corrections
An account-level source-traceability figure can read clean while multi-product accounts carry most of the unsourced slides. Nestack reports the correction rate by account, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Multi-product accounts | 7.7% | 3.7× | Review | |
| Accounts on migrated data | 5.5% | 2.6× | Review | |
| Accounts on legacy reporting | 3.4% | 1.6× | Watch | |
| Single-product accounts | 1.9% | 0.9× | Normal |
Bar: correction-rate lift vs. single-product baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What an unsourced slide costs
The loop shuts when the unsourced slide is a regression case. That suite is what the next deck assembled is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Correction rate rises on multi-product accounts.
- 02
Diagnose
The number on slide nine that three people in the room could each source differently is taken apart until one source is left standing.
- 03
Improve
Every change goes out numbered, with the decks that caused it attached.
- 04
Verify
Nothing releases while one touched deck case is still red.
- 05
Learn
The case stays on, and the sourcing rules are rewritten alongside it.
Learn → Detect
The return edge. The next pack 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, pack assembly, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01QBR pack workflow and automation-boundary work.Week 1
- 02Telemetry, support, billing and BI sources.Week 1, Week 2
- 03Source-traceability and pack-coverage mapping.Week 1, Week 2
- 04Reporting source ingestion.Week 2
- 05Figure, source and as-at date binding.Week 2, Week 3
- 06Sourcing confidence and review routing.Week 3
- 07Account-manager review workflow.Week 3, Week 4
- 08Reporting and presentation integration.Week 5
- 09Slide and source cases.Week 4, Week 5
- 10Guardrails and sourcing controls.Week 4
- 11Deck-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 account, one quarter | ProductionProduction review workflow | AdvancedMultiple accounts / regions |
|---|---|---|---|
| Introduced at Pilot | |||
| Pack assembly to your sources | ✓ | ✓ | ✓ |
| Named manager approval | ✓ | ✓ | ✓ |
| Reporting-pack baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by account | — | ✓ | ✓ |
| Reporting review workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Presentation-tooling integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-entity account packs | — | — | ✓ |
| Multi-stage pack approvals | — | — | ✓ |
| Large pack volumes | — | — | ✓ |
| Multi-account pack controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, pack 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 quarterly review each one gets
→Source capture and as-at date versioning
Week 102Representative telemetry, billing and support sources
→Source binding, assembly logic and the pack baseline
Week 203Your review calendar and the managers it names
→Source mapping, sourcing rules and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Telemetry, billing and support source assessment, then integration setup
Week 205Decks you would not want re-derived
→Source cases and the evaluation round
Week 406What no slide may settle
→Sourcing confidence, review routing, guardrails and release controls
Week 307A named account manager who presents it
→Release to the named account 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
The fifth band holds two phases, and that is a measurement rather than a flourish in the drawing.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Review-pack discovery, sourcing rules and the automation boundary
W2Source integration and the reporting-pack baseline
W3Assembly logic, restatement marking and release controls
W4Evaluation suite, source cases and failure-mode testing
W5Presentation-tool integration, pilot packs and targeted corrections
W6One review cycle run under the account manager, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its own work is named for, and week five is shared by design.
- At the end of W6
- Once the source 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 a QBR preparation agent around the number your last pack could not source.
Show us one account you review each quarter and its last pack. Not the figure that felt right on the day. The figure with the system it came from and the date it was true on, because somebody will hold you to it a quarter later. An unsourced slide comes back as a finding.
- Typical delivery
- 4–6 weeks
- Deployment
- Your live accounts
- Account manager
- Traceable pack
- After launch
- Agent Care available
Nestack Agents · QBR preparation
AGT-CX-02 · Agent Care available after launch