Demand Plan Reconciler
Says which forecast has been right before
- 8hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 9output fields
- A structured result, not a paragraph you have to re-read
- ≤6steps per run
- Bounded work: at most 4 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Sales forecasts to a target. Operations plans to what it can build. Finance budgets to what it promised the board. Three numbers for the same quarter, and the organisation resolves it by having each department act on its own.
The consequences land elsewhere. Operations builds to the conservative number and cannot supply when sales delivers, or builds to the optimistic one and the difference sits in a warehouse for a year. Either way the cost appears in a different department from the forecast that caused it, which is exactly why nobody fixes it.
The consensus meeting meant to resolve this resolves it by authority — the most senior or most confident person wins — rather than by evidence about who has been right before.
That evidence exists. Every forecast has a history against actuals, and it is usually biased in a consistent direction by a consistent amount, per product, per team. A forecast that has been 18% optimistic for eight quarters is not a forecast, it is a target with a known correction factor. Nobody applies the correction because nobody has measured it.
This agent measures it, shows where the plans diverge, and prices the consequence of each divergence in stockouts or excess. It sets no plan and commits no production — that is a decision with commercial context it cannot see.
Benefits
Consensus by evidence, not seniority
Which is the only way the meeting improves anything.
Bias corrected once it is measured
Usually the single biggest accuracy gain available.
The cost made visible
In the department that will actually pay it.
What it does
Measures each forecast's own bias
18% optimistic for eight quarters is not a forecast, it is a target with a known correction factor nobody applies.
Shows where the plans diverge
By product and period, rather than as one company-level gap that hides everything.
Prices the disagreement
What each divergence costs in stockouts or excess stock. The number that ends the meeting.
Names which forecast to trust where
Track record differs by product. One team is reliable on one line and not on another.
Finds forecasts that never move
A number repeated unchanged each cycle is a placeholder, not a plan, and it is treated as one.
Sets no plan
It has no route to a planning system and commits no production, purchase or capacity.
What this changes
What this changes about your week
- It takes over
- Reconciling sales, operations and finance forecasts against their own track records.
- Done today by
- A consensus meeting, resolved by whoever is most senior or most confident.
- How often
- Every planning cycle.
- Typical manual effort
- Bias has to be measured per source per product across many periods.
Why Operations in particular. A forecast that has been 18% optimistic for eight quarters is a target with a known correction factor, and nobody applies it because nobody has measured it. The cost lands in a different department from the forecast that caused it.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Demand Plan Reconciler has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolsextract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output, abridged.
Input
{
"horizon": "Q4 2026",
"currency": "USD",
"forecasts": [],
"minimumTrackRecord": 6
}Output
{
"escalate": true,
"disclaimer": "An analysis of your own forecasts against your own actuals. No plan has been set or published, no production, purchase or capacity committed, no service level recommended, and bias-corrected figures are existing forecasts with measured error applied rather than independent forecasts.",
"divergences": [
{
"period": "Q4 2026",
"spread": "Sales 41,000 · Finance 36,000 · Operations 31,000",
"product": "Product A",
"unitsApart": 10000,
"costIfWrongEitherWay": "Building to Operations and Sales being right costs 340,000 in stockouts. Building to Sales and Operations being right leaves 10,000 units held for roughly a year at 61,000. Both numbers land in departments that did not produce the forecast."
}
],
"biasBySource": [
{
"source": "Sales",
"product": "Product A",
"direction": "optimistic",
"consistency": "consistent-correctable",
"periodsMeasured": 8,
"magnitudePercent": 18.4
},
{
"source": "Operations",
"product": "Product A",
"direction": "conservative",
"consistency": "consistent-correctable",
"periodsMeasured": 8,
"magnitudePercent": 9.1
},
{
"source": "Sales",
"product": "Product C",
"direction": "unbiased",
"consistency": "noisy-not-correctable",
"periodsMeasured": 8,
"magnitudePercent": 2.1
}
],
"biasCorrected": [
{
"note": "Sales corrected for its measured 18.4% optimism gives 33,500. Operations corrected for its 9.1% conservatism gives 33,800. Both existing forecasts, with their own measured error applied — not a new forecast. They agree once the bias is removed, which is the finding.",
"product": "Product A"
}
],
"mostReliableBy": [
"Product B: Operations has been within 4% for eight periods. Product C: Sales, where Operations has been erratic. Reliability is not uniform across the range."
],
"escalationReason": "Product A forecasts are 32% apart with a priced consequence above threshold, and one source shows 18.4% consistent bias.",
"unchangedForecasts": [
"Product D's finance number has been identical for three cycles. That is a placeholder and should not carry equal weight in the meeting."
],
"insufficientTrackRecord": [
"Product E launched two quarters ago. No bias correction is applied to it."
]
}Requirements
Integrations
No integrations required.
Before you start
- Forecasts by source, product and period
- Actuals for prior periods against each forecast
- Cost of a stockout and of holding excess, where you have them
- Requires the All-Access plan or higher
Use cases
Before the consensus meeting
Divergences, track records and the cost of each.
After a stockout or a write-off
Which forecast caused it, and whether the bias was known.
New product planning
Which forecasts have been reliable on comparable launches.
Annual planning
Bias-corrected baselines rather than last year plus a percentage.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Does it produce its own forecast?
It produces a bias-corrected version of each existing forecast — the same forecast with its measured historical error applied — and labels it as exactly that. It does not invent a fourth number to argue with the other three.
Does it commit production or purchasing?
No. It has no route to a planning system. Committing capacity or stock is a decision with commercial context it cannot see.
Does it judge which team forecasts badly?
It reports measured bias per forecast and per product, as a property of the process rather than the people. A sales forecast is a target, and reporting it as biased is a description of what it is for.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent