Skip to content
Blueprint · FMCG

Explain every change in an order. From request to delivery.

A worked example: an order for 100 cases was confirmed at 80 and delivered at 60. Agents follow every quantity change, show the rule behind it and draft an answer the customer team can stand behind.

FMCG
The question

The customer asked for 100 cases. Why were 80 confirmed and only 60 delivered?

Inputs

What the agents read.

Order lines
Requested, confirmed and shipped quantities, with timestamps
Allocation rules
Customer priority, fair share and quotas
Stock and batches
On hand, reserved and remaining shelf life by location
Delivery records
Picking, loading and proof of delivery
Customer terms
Service level and minimum remaining shelf life
Run

What happens, step by step. Each step is logged with its sources.

  1. 01

    Find the case

    Pull together the order, customer, stock and delivery records.

  2. 02

    Trace the allocation

    Show which customer, stock or priority rule cut the requested quantity.

  3. 03

    Check the fulfilment

    Check batch eligibility, freshness, storage location and delivery constraints.

  4. 04

    Prepare the answer

    Write a plain explanation with each change linked to its rule and record.

  5. 05

    Review and learn

    An owner approves, corrects or escalates. Corrections join the test set.

Agent Hive · Run 1290
Why were 80 of 100 cases confirmed and only 60 delivered on order 4471?
  1. Readorder 4471, line 100.3s
  2. Tracedallocation run 03:00 · fair-share cut1.2s
  3. Checkedbatch 7740 shelf life 52%0.8s
  4. Matchedpick list and proof of delivery0.6s
  5. Wroteexplanation with 4 sources0.4s
Answer4 sources
100 → 60cases
Requested to delivered, order 4471

20 cases were cut by the fair-share rule during a constrained allocation. 20 more failed the customer's 60% shelf-life term at picking.

order_linesallocationswms.stockcustomer terms
Ask a follow-up…
Approvals

Where people decide.

  • Needs you

    Customer reply

    The service owner approves, corrects or escalates the explanation before anything goes to the customer.

  • Needs you

    Rule changes

    Allocation and fulfilment rules belong to supply planning. Agents can flag a rule; they never edit one.

Outputs

What you get back.

Explanation
Plain answer with a quantity waterfall and linked records
Pattern view
The same issue across customers, products and locations
Run trace
Sources, rules, agent steps and the reviewer decision
The answer

Every unit is accounted for. Each cut has a rule behind it.

The waterfall is the answer: requested, cut by allocation, confirmed, cut by fulfilment, delivered. Each step opens the record and the rule.

Fig. 1 · Order 4471, line 10Illustrative data
Allocation cutFair-share rule

Stock for SKU 220 was short at the 03:00 run, so every tier-2 customer got 80% of the request.

Fulfilment cutShelf-life term

Batch 7740 had 52% shelf life left. Customer B requires 60%, so FEFO picking skipped it.

DeliveredProof of delivery

Load 5520 delivered 60 cases; the proof of delivery matches the pick list.

Requested 100 cases, confirmed 80, delivered 60. Each cut opens the record and the rule behind it.
Data sources

Start with the question, then follow the evidence.

Agent Hive gathers only the context needed for this order. The first version stays read-only: no write-back until an owner says so.

Fig. 2 · Connected sources
Sources stay in place; Agent Hive reads them through read-only connectors and builds one shared model. Logos are trademarks of their respective owners.
  • Order and allocation

    Order lines, confirmations and allocation runs from the ERP.

  • Stock and warehouse

    Batches, reservations, shelf life and picking records.

  • Customer context

    Terms, service levels and the original request from CRM and email.

Agent workflow

Let the agents search. Keep the business owner in control.

Allocation and fulfilment are traced in parallel, then joined into one explanation that waits for the service owner.

Fig. 3 · Workflow runIllustrative data
The run as the platform draws it: tool steps, specialist agents, joins, branches and the approval it is waiting on. Drag to pan; Ctrl or ⌘ + wheel to zoom.
Knowledge model

The order trail, as one connected model.

Orders, lines, allocation runs, batches, deliveries and the rules that govern them, each linked to its source table or document.

Fig. 4 · Knowledge modelIllustrative data
Order line 10 in context: the allocation run and its rules, the stock and batches it could draw on, the delivery, and the customer terms.
Question types

Use the same pattern for the next question.

  • Allocation

    Why was 100 confirmed as 80?

    Show which customer, stock or priority rule changed the requested quantity.

  • Fulfilment

    Why did only 60 ship?

    Check batch eligibility, freshness, storage location and delivery constraints.

  • Pattern

    Where does it keep happening?

    Compare similar orders across customers, products, locations and weeks.

Evaluation

Agree on a good answer before the pilot begins.

Representative orders and expert review check that each explanation is correct, complete, useful and supported by the right evidence.

Fig. 5 · Question set by versionIllustrative data
Evaluation results by question and version
Questionv1v2v3
Why was order 4471 confirmed at 80 of 100?Allocation✓✓✓
Which priority rule applied to Customer B?Allocation✗✓✓
Why did 20 cases fail the shelf-life check?Fulfilment~✗✓
Was the short delivery a picking or a stock issue?Fulfilment~✓✓
Which DC shorted the most lines last month?Pattern✓~✓
Do other customers on the same SKU see the same cut?Pattern✗✗✓
Draft the reply to the customer teamAnswer✓~✓
v3: 7 of 7 pass · release needs all 7
Each row is a question agreed with the team; each column a version of the workflow. The candidate ships only if it holds the release bar.
  • 01

    Owner checks the answer

    Service and planning review real explanations.

  • 02

    The case joins the set

    Corrections become repeatable test cases.

  • 03

    New versions are checked

    Every change reruns the whole set.

  • 04

    Release criteria apply

    Only versions that hold the bar go live.

Outcomes

Fewer reopened tickets. Answers the customer team can send.

To a reviewed answer
6 min
from a day of email threads
Units accounted for
100%
every cut linked to a rule
Reopened cases
−42%
over the first 8 weeks
Eval pass, v3
7 of 7
release bar
Illustrative data
Deployment

Keep the order context private, whichever model setup you choose.

The first workflow can use approved read-only access. Any later action or write-back stays behind permissions, review and audit.

Run inside the enterprise boundary.

Private environment

Agent Hive, the operating context and the model runtime stay in a private cloud, VPC or on-premises environment.

  • Approved read-only connectors
  • Enterprise SSO and roles
  • Source and decision trace

Use approved models through a policy gateway.

Controlled hybrid

The order context stays private while selected requests pass through redaction, routing and audit controls.

  • Per-agent model routing
  • One controlled egress path
  • Approval before write-back
Fig. 6 · Deployment options
Deployment options compared: Managed, Dedicated cloud, On-prem, Air-gapped
Runs inQpiAI managed environmentA dedicated network in your cloudYour data centre (Docker / Compose)An isolated network
Operated byQpiAIAgreed per engagementYour teamYour team
TenancyMulti-tenant, isolatedSingle-tenantSingle-tenantSingle-tenant
Data at restManaged environmentYour cloud and regionYour networkInside the perimeter
ModelsManaged modelsApproved endpointsLocal, or via a policy gatewayLocal models only
OutboundOperated by QpiAIPrivate connectionsOne controlled egress path (hybrid)None
IdentityYour identity rulesEnterprise identitySSO and RBACLocal SSO and RBAC
UpdatesRolled out by QpiAIAgreed per engagementNew image version, pulledSigned offline packages
AuditAudit historyAudit historyLocal audit historyLocal audit history

Air-gapped Graph, agents and models run inside an isolated network; updates arrive as signed offline packages.

What runs where in each option. Select a column to move the boundary.

Run this blueprint on your data. We set up the first one with you.