Agriculture & Food

Supply and Demand Planning Copilot

Designed for Planner

Give the demand planner a consensus planning surface that reconciles statistical forecast, promotion lifts, open orders and weather-sensitive demand at the item-location level, explains every override, and routes exception-level changes for commercial sign-off.

Contract AGR-03 · primary aggregate PlanningUnitForecast · mode loading

Decision brief

Can we trust the 34% demand spike?

Here is how 34% landed: the forecast engine added 19 points for the approved feature-and-display event, 11 raw points for the overlapping coupon, and 4 points for warm weather. That 34-point stack moves 10,000 units to 13,400. The team must determine how much is real, duplicated, and uncertain.

Decision signals

The evidence and control points that shape this recommendation.

Why 34% landed

19 + 11 + 4 = 34

Feature and display, raw coupon lift, and warm-weather demand were stacked by the forecast engine.

Statistical forecast

10,000 → 13,400

(13,400 - 10,000) / 10,000 = 34%. Calculated upstream, not by GenAI.

Approved promotion

+19 points

Structured feature-and-display plan retained by deterministic rules.

Overlapping coupon

11 → 4 points

A deterministic overlap rule removes seven duplicate points.

Rules-adjusted forecast

+27%

13,400 minus 700 duplicated units produces 12,700 units.

Human-approved forecast

+26%

The planner applies a one-point risk adjustment and approves 12,600 units.

Business problem

A Midwest retailer scheduled a feature-and-display event and an overlapping digital coupon during an unusually warm Ohio week. The statistical engine stacked those signals and raised the forecast from 10,000 to 13,400 units.

Decision to make

Verify the calculation, remove duplicate promotion lift, review the GenAI evidence summary, and decide whether to approve 12,600 units.

Outcome to show

Lock one explainable forecast and hand it to supply planning and procurement.

  1. 1Confirm the baseline
  2. 2Remove double counting
  3. 3Explain the remaining uplift
  4. 4Get human approval
  5. 5Lock and hand off

Step 1 of 5

Ready: Start with the business question, then use the single button above.

The output appears here immediately. More technical evidence remains available below.

System logic, GenAI, and human accountability

The demonstration keeps deterministic controls, AI assistance, and approval separate.

Existing planning stack

The forecast engine calculates 34%; deterministic services net promotion overlap, validate workflow order, and lock records.

GenAI contribution

Hybrid retrieval finds prior minutes and the weather advisory. GenAI synthesizes them into a cited explanation and draft rationale.

Human accountability

The planner evaluates uncertainty, applies the final one-point risk adjustment, approves 26%, and remains accountable for the decision.

Workflow stages

Run the stages in order before using the Copilot. Every stage explains what happens, why it matters, the evidence it uses, and the complete API request and response.

Seeded record
Loading
Document type
Loading
Container
Loading
Partition key
Loading
Current stage
Loading

API demonstration guidance

Run a stage to see its explanation, evidence, full request, and full response.

Grounded chat

Ask a question and receive an answer grounded in retrieved PlanningUnitForecast evidence.

Answer and evidence

Select a starter question or enter your own.

Evidence

Domain implementation workspace

Loaded live from /api/config, /api/domain/model, and /api/domain/sample.

Domain objective
Give the demand planner a consensus planning surface that reconciles statistical forecast, promotion lifts, open orders and weather-sensitive demand at the item-location level, explains every override, and routes exception-level changes for commercial sign-off.
Primary aggregate
PlanningUnitForecast
Primary container
planningUnitForecasts
Hierarchical key
/tenantId + /planningUnitId
Schema version
1

Sample scenario

A 12-ounce ready-to-drink coffee SKU shipping from the Ohio plant has a 10,000-unit reference forecast. The statistical forecast engine produces 13,400 units for week 41, so (13,400 - 10,000) / 10,000 yields a 34 percent increase; GenAI does not calculate this baseline. The raw uplift contains 19 points from an approved retailer feature-and-display promotion, 11 points from an overlapping digital coupon, and 4 points from warm-weather demand. A deterministic overlap rule nets the coupon from 11 points to 4, removing 700 units and producing a 12,700-unit, 27 percent rules-adjusted forecast. GenAI retrieves the prior consensus minutes and weather advisory, then drafts a cited explanation and override rationale. The planner applies a one-point uncertainty adjustment, approves 12,600 units, a 26 percent increase, records accountability, and locks the consensus version for supply planning and procurement.

Physical Cosmos DB containers

Every container provisioned for this accelerator with its hierarchical partition key.
Container Role Hierarchical partition key Document types
planningUnitForecasts Primary /tenantId + /planningUnitId PlanningUnit, ForecastVersion, PromotionPlan, DemandSignal, PlanningNarrativeChunk, ForecastOverride

Document types

Contract document types, their owning container, and their identifier pattern.
Type Container Partition key Identifier pattern Embedding fields
PlanningUnit planningUnitForecasts /tenantId + /planningUnitId planningUnit::{planningUnitId} None
ForecastVersion planningUnitForecasts /tenantId + /planningUnitId planningUnit::{planningUnitId}::forecast::{versionLabel}::{bucketStartUtc} None
PromotionPlan planningUnitForecasts /tenantId + /planningUnitId planningUnit::{planningUnitId}::promotion::{promotionId} None
DemandSignal planningUnitForecasts /tenantId + /planningUnitId planningUnit::{planningUnitId}::signal::{signalTimestampUtc}::{signalType} None
PlanningNarrativeChunk planningUnitForecasts /tenantId + /planningUnitId planningUnit::{planningUnitId}::narrative::{narrativeDocId}::{chunkOrdinal} chunkVector
ForecastOverride planningUnitForecasts /tenantId + /planningUnitId planningUnit::{planningUnitId}::override::{overrideId} rationaleVector

Query patterns