Strategy AI agent · Market sizing
Market Analysis AI Agent
Version the boundary before the number: which geographies, which buyer types, which adjacent categories count, then derive top-down and bottom-up apart and hand the reconciled estimate to the analyst who accepts it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Named analyst
- Boundary first
- After launch
- Agent Care
What this agent does
Derives the estimate, never accepts it
In
- 01
A market is sized, and the boundary that produced the number is written down beside it.
- 02
A definition shifts, and the older figure and the newer one stop being comparable.
Reason
- 03
A top-down build is raised, and a bottom-up one rises apart from it, never off the same base.
- 04
A residual opens between the two builds, and it is reported at its width rather than averaged away.
- 05
A forecast is drawn, and every assumption sitting under its base year travels with it.
Decide
- 06
A source ages, and each estimate leaning on that source is flagged for re-derivation.
- 07
A figure is reused in a planning paper, and the day it was derived is carried alongside it.
Out
- 08
A market resists definition, and that resistance is recorded as the finding, not as a gap.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
Derivation, reconciliation and versioning belong to the agent. Acceptance belongs to a named analyst, who takes the estimate into the plan and owns it there.
Example workflow
One estimate, sources to acceptance
- 1Source material receivedTrade series, filings, panel data, tender registers or internal sales history
- 2Definition fixed and versionedThe geographies, the buyer types, the adjacent categories counted in and the day that boundary was set
- 3Two builds raised apartThe top-down build, the bottom-up build, the residual and its width
- 4Controls appliedBoundary checks, source-age checks, arithmetic checks and derivation confidence
No human action required
Stages 1 to 4 run unaided, and nothing enters a plan at any of them — the agent is deriving, and the analyst lane opens at the reconciliation gate.
Goes to the named analyst to accept.
Adds a senior analyst read first.
Analyst review
The estimate is held with its definition, its residual and the sources it was drawn from.
Accept · Append source · Send to analyst review
Accepted — by the named analyst
- 6Planning and model records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedDefinition stability, source currency, analyst corrections and what review found
Corrections
Each analyst correction is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Accepting an estimate into the planning cycle.
- Setting the boundary a market is reported at.
- Deciding which adjacent category counts as in.
- Signing off a forecast for a board paper.
Automation boundary
Agent acts unaided
- ✓Write the market definition down and version it.
- ✓Raise the top-down and the bottom-up builds on separate bases.
- ✓Report the residual between the two builds at its measured width.
- ✓Flag any figure whose source has aged past its term.
Nothing enters a planning paper except by a named analyst, inside the agreed boundaries.
- Judging whether a market can be defined at all.
- Telling a board that an estimate is defensible.
- Choosing the growth rate a forecast is built on.
- Changes to the plan, the model or the definition.
Example output
One market estimate, annotated
This serves a strategy team who may have to defend a number in a planning paper years after it was drawn; below is one estimate exactly as the agent leaves it.
Market estimate · single segment
Illustrative example- Estimate
- Serviceable market, mid-market buyers
- Recorded as
- Reconciled across both builds
- Segment
- Boundary, versioned
- Evidence of record
- Trade series, 3 August 2026
- Confidence
- Held unaccepted
- Held for
- The named analyst, by name
As received
Built from a published trade series and internal sales history, and it claims nothing beyond them.
What the record holds
- Trade series
- Sales history
- Tender register
Why no acceptance here
Taking a number into the plan is a judgement a named analyst makes.
What the score decides
Below the configured threshold an estimate gets a senior read before the analyst sees it.
Action
AcceptAppend sourceSend to analyst 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 estimate
From the source that carries it
Where the evidence is used
The agent does not vouch for a source, only for what that source says, when it said it, and how the estimate was built from it.
A market is a definition
Move the geography, the buyer type or the question of whether services count, and the number moves too, with nobody having lied.
What was checked, and not found
No published series consulted gives one agreed boundary for this category, and they do not treat services alike, so the spread between them is reported rather than a consensus figure that none of them supports.
Build an evidence trail
The estimate, the definition it rests on and the analyst who accepted it stay together.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Published statistics
- Trade bodies
- government series
- Category and geography series
Internal commercial data
- CRM
- billing
- sales history
- Won and lost demand records
Panel and survey data
- Research panels
- buyer surveys
- Buyer populations and purchase rates
Agent
Market sizing and segmentation
- Reads the sources
- Raises both builds
- Holds for the analyst
Tenders and public filings
- Tender registers
- annual reports
- Competitor and tender disclosures
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six weirs between the model and the analyst
Six weirs along one channel, the last the highest. What still flows through is drawn in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether an estimate may stand
- L3
- Leaves the acceptance to a named analyst
- L4 – L5
- Keep the estimate and the definition behind it
- L6
- Reverts to source listing when signals degrade
How Nestack evaluates it
Evaluate the whole derivation — not only the estimate that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the estimate a planning paper carries
- E1Final-output evaluationDid the estimate record the definition it was actually built on?
- E2Step-level evaluationDid the agent read the right series, the right period and the live definition?
- E3Tool evaluationDid it read and write the correct segment and the correct estimate?
- E4Confidence calibrationDo low-confidence estimates actually attract more analyst corrections?
- E5Slice evaluationHow does performance change across specific segments?
- E6Business outcomeHow many estimates needed a correction before the analyst accepted?
Floor — the base an estimate rests on
Failure modes
Where each failure originates in the agent
Seven failure modes, each placed at the stage it first becomes visible.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- LO-03
Stale series read
The series read is not the one now published.
Stage gathers
The series, the panels, the periods and the buyers
02 · Reasoning
2 modes
- LO-04
Estimate asserted, not derived
A number appears with no build behind it.
- LO-06
Superseded boundary read as live
A retired definition is worked as the current one.
Stage proposes
The definition, both builds and the residual
03 · Tool / write
2 modes
- LO-02
Thin estimate passed forward
An estimate moves on without the senior read.
- LO-05
Figure bound to wrong segment
The number is filed against another segment.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- LO-01
Accepted, derivation unrecorded
The record shows acceptance but not the build under it.
Stage returns
The estimate a plan carries and a board reads
05 · Change / Version
1 mode
- LO-07
Silent definition drift
A boundary widens while the stored estimate keeps the old one.
Stage tracks
Model, prompt, boundary rules and source dates
- Sev-1 · an estimate accepted on no build
- Sev-2 · a stale figure reaches the plan
- Sev-3 · source degrades, estimate held back
Affected slices
Emerging categories absorb the corrections
A segment-level definition-stability figure can read clean while emerging adjacent categories carry most of the re-derivation. Nestack reports the correction rate by segment, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Emerging adjacent categories | 11.6% | 3.7× | Review | |
| Multi-country roll-ups | 8.3% | 2.6× | Review | |
| Services-inclusive segments | 5.2% | 1.7× | Watch | |
| Established core categories | 2.4% | 0.8× | Normal |
Bar: correction-rate lift vs. established-category baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What a widened boundary costs
A cycle shuts when the undefined market is a regression case. That suite is what the next estimate issued is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Correction rate rises on emerging adjacent categories.
- 02
Diagnose
The market that grew by a third because somebody quietly widened what counted is worked backwards until one cause is left standing.
- 03
Improve
The change ships numbered, and the estimates that forced it ride with it.
- 04
Verify
Nothing releases while one touched estimate case is still red.
- 05
Learn
It is retained for good, and the definition rules are amended in that same commit.
Learn → Detect
The return edge. The next estimate 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, estimate derivation, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Market-definition and automation-boundary work.Week 1
- 02Statistical, panel and internal sources.Week 1, Week 2
- 03Source-to-estimate and definition-coverage mapping.Week 1, Week 2
- 04Market source ingestion.Week 2
- 05Definition, source and segment binding.Week 2, Week 3
- 06Residual scoring and review routing.Week 3
- 07Analyst acceptance workflow.Week 3, Week 4
- 08Planning-model integration.Week 5
- 09Definition and boundary cases.Week 4, Week 5
- 10Guardrails and acceptance controls.Week 4
- 11Estimate-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 segment, one cycle | ProductionProduction planning workflow | AdvancedMultiple segments / geographies |
|---|---|---|---|
| Introduced at Pilot | |||
| Estimate derivation to your definitions | ✓ | ✓ | ✓ |
| Named analyst acceptance | ✓ | ✓ | ✓ |
| Segment-definition baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by segment | — | ✓ | ✓ |
| Analyst review workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Source-library integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-source reconciliation | — | — | ✓ |
| Cross-segment estimate packs | — | — | ✓ |
| Large source libraries | — | — | ✓ |
| Multi-geography definition controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, source 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 segments and the definition each is reported at
→Definition capture and boundary versioning
Week 102Representative statistical, panel and internal sources
→Source binding, build logic and the estimate baseline
Week 203Your planning calendar and the analysts it names
→Definition mapping, source binding and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Statistical, panel and internal source assessment, then integration setup
Week 205Estimates you would not want defended
→Boundary cases and failure-mode testing
Week 406What no market estimate may establish
→Residual scoring, review routing, guardrails and release controls
Week 307A named analyst who accepts the estimate
→Release to the named analyst, 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
Nothing below is padded to fill a row: where two phases genuinely coincide, they share week five.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Definition discovery, boundary versioning and the automation boundary
W2Source integration and the segment-definition baseline
W3Estimate derivation, reconciliation logic and release controls
W4Evaluation suite, boundary cases and failure-mode testing
W5Planning-model integration, pilot estimates and targeted corrections
W6One planning cycle run under the strategy lead, 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
- Validation closes on live estimates, and Agent Care picks up the watch.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Strategy AI agent
Build a market analysis agent around the definition your last estimate never wrote down.
Show us one market you size every year and the definition behind the last number. If that figure has ridden through more than one planning cycle without being re-derived, then it is being trusted rather than defended. A market that cannot be defined comes back as a finding.
- Typical delivery
- 4–6 weeks
- Deployment
- Your live segments
- Named analyst
- Boundary first
- After launch
- Agent Care available
Nestack Agents · Market sizing
AGT-CS-01 · Agent Care available after launch