Operations AI agent · Operational reporting
Operational Reporting AI Agent
Mark any figure that has moved since it was last published, carry the definition behind each number in the pack, and hold the operating review for the operations lead who releases it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Named ops lead
- Definition-led
- After launch
- Agent Care
What this agent does
Assembles the pack, never publishes it
In
- 01
A metric is defined, and that definition is written down beside every number it produces.
- 02
A figure is restated, and the record carries what moved, when it moved and what moved under it.
Reason
- 03
A number is pulled from a source of record, and the day it was pulled travels beside it.
- 04
A name is claimed by two teams, and the two measures behind it are reported apart, never added up.
- 05
A figure is set against the finance close, and the difference is stated as a quantity with a cause.
Decide
- 06
A definition changes mid-series, and the trend is broken at the change rather than drawn through it.
- 07
A late record lands in a closed period, and the figure it rewrote is marked as having moved.
Out
- 08
A number is lifted into a later paper, and the day it was published is carried along with it.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
Assembly, definition carriage and restatement marking belong to the agent. Publication belongs to a named operations lead, who also decides what a number means and whether a restatement is material.
Example workflow
One figure, source to release
- 1Source data receivedWarehouse tables, operational extracts, BI exports or the prior published pack
- 2Definition bound to the numberThe metric, the definition version it is reported at, the period and the day it was pulled
- 3Figure assembledThe number, its definition, what has moved since publication and confidence
- 4Controls appliedDefinition checks, source-currency checks, restatement checks and assembly confidence
No human action required
Stages 1 to 4 run unaided, and nothing is published at any of them — the agent is assembling, and the lead lane opens at the restatement gate.
Goes to the operations lead to release.
Adds a definition review first.
Definition review
The pack is held with each definition, each move since publication and the sources behind it.
Release · Amend figure · Send for definition review
Released — by the operations lead
- 6Reporting and pack records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedDefinition stability, restatement volume, lead corrections and what review found
Corrections
Each lead correction is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Publishing the operating review to the company.
- Deciding what a metric is meant to measure.
- Judging whether a restatement is material.
- Signing off the pack for the executive meeting.
Automation boundary
Agent acts unaided
- ✓Assemble the pack from the sources of record.
- ✓Carry each definition beside the number it produced.
- ✓Mark any figure that has moved since the last pack, and say why.
- ✓State the gap between the operational number and the finance close.
Nothing reaches a meeting except by a named operations lead, inside the agreed boundaries.
- Ruling which of two rival definitions is right.
- Telling a board that a number is final.
- Retiring a metric from the operating review.
- Changes to the pack, the metric or the definition.
Example output
One reported figure, annotated
Performance reporting answers to a client and board material preparation is governance papers; this one answers to the company that owns the number, where a restatement announces itself.
Reporting output · single figure
Illustrative example- Metric
- On-time completion, weekly
- Recorded as
- Moved since the last pack
- Definition
- Restated
- Evidence of record
- Warehouse pull, 3 August 2026
- Confidence
- Held unreleased
- Held for
- The operations lead, by name
As received
Built out of the warehouse of record and the stored definition — nothing beyond those two is claimed.
What the record holds
- Warehouse extract
- Stored definition
- Prior published pack
Why it moved
A source was backfilled and the figure moved; the note says which.
What the score decides
Below the configured threshold a figure gets a definition review before the lead sees it.
Action
ReleaseAmend figureSend for definition 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 figure
From the system that holds it
Where the definition is used
The agent does not vouch for a source, only for what that source held, when it was pulled, and how the figure was built out of it.
A number needs a definition
Move the system boundary, the filter or the date a record counts on, and the number moves with nobody having lied. When a released figure later turns out wrong, it is marked superseded where it was published and the lead who released it is told.
What was checked, and not found
No external regime governs the weekly and monthly operating review: no filing, no certification, no auditor and no statutory deadline attaches to an internal management pack. Statutory financial reporting and the certifications that ride with it belong to the finance function and are not carried here. What holds this page up is operational discipline set by your own reporting policy, and the only thing holding a figure back is a named person.
Build an evidence trail
The number, the definition it was built on and the analyst who published it stay together.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Warehouse of record
- Snowflake
- BigQuery
- dbt
- Modelled tables and the metric layer
Business intelligence
- Power BI
- Tableau
- Looker
- Published packs and the figures
Operational systems
- ERP
- WMS
- service desk
- The transactions a number is built out of
Agent
Operational reporting
- Reads the sources
- Assembles the pack
- Holds for the lead
Definitions and lineage
- dbt
- metric layers
- catalogues
- Definitions and their change history
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six filters between the model and the pack
Six filters in series, the last the tightest. What survives is set out in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a figure may stand
- L3
- Leaves the release to a named operations lead
- L4 – L5
- Keep the number and the definition behind it
- L6
- Withholds the restatement 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 the meeting reads
- E1Final-output evaluationDid the figure carry the definition it was actually built on?
- E2Step-level evaluationDid the agent read the right source, the right period and the live definition?
- E3Tool evaluationDid it read and write the correct metric and the correct period?
- E4Confidence calibrationDo low-confidence figures actually attract more lead corrections?
- E5Slice evaluationHow does performance change across specific metrics?
- E6Business outcomeHow many figures needed a correction before the pack was released?
Floor — the number a decision rests on
Failure modes
Where each failure originates in the agent
Seven failure modes, each set at the stage where it first becomes visible.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- NJ-03
Source changed underneath
The system of record moved and the pull did not.
Stage gathers
The sources, the metrics, the periods and the pulls
02 · Reasoning
2 modes
- NJ-04
Figure asserted, not assembled
A number appears with no source behind it.
- NJ-06
Superseded definition read as live
A retired definition is worked as the current one.
Stage proposes
The figure, its definition and what has moved
03 · Tool / write
2 modes
- NJ-02
Thin figure passed forward
A figure moves on without the definition read.
- NJ-05
Figure bound to wrong metric
The number is filed against another metric.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- NJ-01
Released, restatement unmarked
The record shows the new figure but not the move.
Stage returns
The pack a meeting reads and a lead signs
05 · Change / Version
1 mode
- NJ-07
Series crosses a redefinition
A trend runs through a definition change unmarked.
Stage tracks
Model, prompt, metric rules and source dates
- Sev-1 · a figure released with no source
- Sev-2 · a restated figure reaches the pack
- Sev-3 · source degrades, pack held back
Affected slices
Newly defined metrics absorb the restatements
A metric-level definition-stability figure can read clean while newly defined metrics carry most of the restatements. Nestack reports the restatement rate by metric, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Newly defined metrics | 7.6% | 3.7× | Review | |
| Metrics shared across teams | 5.4% | 2.6× | Review | |
| Series crossing a redefinition | 3.3% | 1.6× | Watch | |
| Stable long-running metrics | 1.8% | 0.9× | Normal |
Bar: restatement-rate lift vs. stable-metric baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What a silent restatement costs
A loop ends when the silently restated number becomes a case the next release must pass. That suite is what the next pack is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Restatement rate rises on newly defined metrics.
- 02
Diagnose
The number that moved between the pack and the meeting with no note saying so is taken apart until a single cause is left.
- 03
Improve
The change ships numbered, and the restatements that drove it ride underneath.
- 04
Verify
One number case still failing holds the entire release back.
- 05
Learn
It is retained permanently, and the definition rules move with 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
- 01Metric-definition and automation-boundary mapping.Week 1
- 02Warehouse, BI and operational sources.Week 1, Week 2
- 03Source-to-figure and definition-ownership mapping.Week 1, Week 2
- 04Source-of-record ingestion.Week 2
- 05Figure, definition and period binding.Week 2, Week 3
- 06Restatement scoring and review routing.Week 3
- 07Lead release workflow.Week 3, Week 4
- 08Reporting-system integration.Week 5
- 09Definition and restatement cases.Week 4, Week 5
- 10Guardrails and publication controls.Week 4
- 11Number-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 pack, one cycle | ProductionProduction reporting workflow | AdvancedMultiple entities / metric sets |
|---|---|---|---|
| Introduced at Pilot | |||
| Pack assembly to your definitions | ✓ | ✓ | ✓ |
| Named operations lead release | ✓ | ✓ | ✓ |
| Metric-definition baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by metric | — | ✓ | ✓ |
| Definition review workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Warehouse-of-record integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-source reconciliation | — | — | ✓ |
| Cross-entity reporting packs | — | — | ✓ |
| Large metric catalogues | — | — | ✓ |
| Multi-entity definition controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, metric 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 metrics and the definition each is reported at
→Definition capture and restatement marking
Week 102Representative warehouse, BI and operational sources
→Source binding, assembly logic and the pack baseline
Week 203Your reporting calendar and the leads it names
→Definition mapping, source binding and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Warehouse, BI and operational source assessment, then integration setup
Week 205Numbers you would not want re-derived
→Restatement cases and the evaluation round
Week 406What no reported number may settle
→Restatement scoring, review routing, guardrails and release controls
Week 307A named operations lead who releases the pack
→Release to the named operations lead, 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 grid is not decoration. Where two phases truly run together, the fifth week shows them both.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Definition discovery, metric ownership and the automation boundary
W2Source integration and the metric-definition baseline
W3Pack assembly, restatement logic and release controls
W4Evaluation suite, definition cases and failure-mode testing
W5Reporting-system integration, pilot packs and targeted corrections
W6One reporting quarter run under the operations lead, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its own work is named for, and week five carries two.
- At the end of W6
- Once the definition 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 · Operations AI agent
Build a reporting agent around the number that moved without saying so.
Show us one weekly pack and the number in it people argue about. Not a figure that never moves. A figure that moves and says so, with the cause beside it and the operations lead who released it named on the same page.
- Typical delivery
- 4–6 weeks
- Deployment
- Your live metrics
- Named ops lead
- Definition-led
- After launch
- Agent Care available
Nestack Agents · Operational reporting
AGT-OP-04 · Agent Care available after launch