Decision intelligence, applied where it matters
Twelve families of consequential supply-chain decisions. Governed end to end.
Aidecis is the decision intelligence layer for supply chain.
Each decision below is taken today by people, by your own algorithms or by a vendor's agent, and often by all three. Aidecis governs each one end to end, from the signal to the outcome, under one authority model, and links the decisions a change sets off next. Each case says how far it can move up the autonomy scale, and the one measure a pilot is judged on.
The decision, end to end
Context in. Governed action out. Evidence all the way through.
- Signal + context
- AI analysis + options
- Human / bot authority
- Approval + scoped action
- Outcome + learning
Decision-effectiveness scorecard
One scorecard for every case, so pilots can be compared.
Each use case below lists what gets measured under the same four groups. These are the measures a pilot is judged on, set against your own baseline.
- Decision latencyHow fast a decision is taken: time from signal to decision, time to decision.
- Decision qualityHow well it holds: policy exceptions, overrides, reversals.
- AdoptionWhether the flow is used: the share of decisions taken on the brief, the cases that run through the flow.
- Operational outcomeWhat it changes in operations: premium freight, stockout hours, on-time-in-full, the measure of each case.
Procurement · reference dossier
Procurement price, quantity or PO exception
The decisionA supplier asks for a higher price, a larger quantity or an expedited purchase outside the agreed tolerance, or confirms a PO line late.
Who takes partBuyers and budget owners; the purchasing rules; a vendor's contract or spot-buy agent that may accept the change itself.
What Aidecis doesCompares the contract, budget, stock, criticality and alternatives, routes the approval by total commitment and delegated authority, and grants an order amendment only within the approved scope, which the buyer or your integration executes (enforced control). A price change a vendor's agent accepts on its own is read back and checked against the price envelope (observed control).
How far it can goLevel 2 to 3 for small variances inside the envelope; above it, a person decides.
The measurePrice variance against contract. Also: maverick spend, approval time.
Opens nextA material shortfall and a production reschedule, and later an invoice variance to reconcile — each a decision of its own in the chain.
Scorecard
- Latency
- Approval time
- Quality
- Maverick spend
- Adoption
- Share of price and PO exceptions routed through the flow
- Outcome
- Price variance against contract
Price or PO changeEnvelope + contextAuthority + approvalScoped PO amendmentMargin + chain outcome
Decision map · reference case
A late PO line, from signal to customer promise.
One change followed across the chain, with the registered actor and the system connection at each step, and the decisions it opens next. Illustrative example: names, limits and amounts are set with each client.
Signal
A supplier confirms a purchase-order line late and above the agreed price, past the material cut-off for a production order.
- Actor
- The supplier; the buyer in SAP
- System
- SAP purchasing, read only
Context
Aidecis links the PO line to material availability, the production order it feeds and the customer promise downstream, and prices the options against the contract and budget.
- Actor
- Aidecis prepares the brief; it does not decide
- System
- SAP (purchasing, MARD, RESB), read only
Applicable policy
The delegation-of-authority policy sets who may accept the change, the price envelope per line and the total commitment above which it escalates.
- Actor
- Policy set by the client's supply-chain director
- System
- Client policy and limits held in Aidecis
Accountable owner
The brief goes to the buyer whose envelope covers it; above it, the category manager. The approval, the option chosen and the reason are recorded.
- Actor
- Named buyer or category manager
- System
- Aidecis decision record
Scoped action, and the chain it opens
Enforced: your integration checks the single-use grant, then executes the approved PO amendment in SAP; Aidecis reads it back and reconciles it. Observed: a vendor's agent that amends in SAP directly is read back and flagged against the envelope. The shortfall opens a production reschedule and a customer-promise check, each with its own owner.
- Actor
- The buyer or your integration, or the vendor's registered agent
- System
- SAP purchasing
Verified outcome
Aidecis follows the chain to the production order and the customer promise, and checks the later supplier invoice against the approved grant, measuring the outcome of the whole chain.
- Actor
- Buyer; production and commercial owners; finance for the invoice
- System
- SAP, read back
Inventory · allocation
Inventory reallocation
The decisionStock runs short at one site while another holds a surplus. Moving it costs transport and may leave the sending site short.
Who takes partSite planners and the supply planner; replenishment rules and the ERP's transfer proposals; a planning vendor's optimiser.
What Aidecis doesRanks the receiving sites by service risk, protects minimum stock and reserved orders at the sending site, and, after approval, grants the transfer, which your WMS or ERP integration checks before it executes (enforced control). A transfer the optimiser books on its own is read back, checked against its limit and flagged when it goes over (observed control).
How far it can goLevel 2 is realistic early: the optimiser proposes, a planner approves above a set quantity.
The measureStockout hours at the receiving site. Also: fill rate, transfer cost, policy exceptions.
Opens nextA customer-promise change if cover still falls short at the receiving site.
Scorecard
- Latency
- Time from shortage signal to approved transfer
- Quality
- Policy exceptions
- Adoption
- Share of transfers decided on the brief
- Outcome
- Stockout hours at the receiving site; fill rate; transfer cost
Planning · allocation
Demand surge allocation
The decisionA demand spike or a large order exceeds the supply available to promise, and someone decides which customers get what.
Who takes partDemand and supply planners, and the commercial owners; the allocation engine in the planning system.
What Aidecis doesCompares allocation options on customer commitments, margin, strategic rules and fairness, records who approved each change to a customer promise, and updates planning or the ERP only after approval.
How far it can goLevel 1 to 2. A customer promise is a commercial decision, and a person stays accountable for it.
The measureOn-time-in-full for the affected customers. Also: revenue protected, allocation overrides.
Opens nextA customer promise and backorder priority for the orders that have to wait.
Scorecard
- Latency
- Time from surge to approved allocation
- Quality
- Allocation overrides
- Adoption
- Share of allocation changes approved in the flow
- Outcome
- On-time-in-full for the affected customers; revenue protected
Procurement · resilience
Supplier disruption and alternate source
The decisionA supplier is late, cuts capacity, or is hit by a credit alert or a geopolitical event. Buying elsewhere costs a premium and needs a qualified source.
Who takes partBuyers, quality and the plant; the sourcing rules in the ERP; a procurement vendor's sourcing agent that may place the order itself.
What Aidecis doesMaps the affected materials and plants, compares approved alternates on lead time, price and qualification status, asks quality and procurement to approve, and grants the purchase-order change only within the approved scope (enforced control). An order a vendor's agent places directly is read back and checked against its cap and the approved-supplier list (observed control).
How far it can goLevel 1 to 2. Procurement is where vendors' agents arrive first, so this is often where observed control starts.
The measureDays of cover protected. Also: premium purchase cost, time to qualify an alternate.
Opens nextA quality review of the alternate source, and a production reschedule.
Scorecard
- Latency
- Time to qualify an alternate
- Quality
- Orders outside the cap or the approved-supplier list
- Adoption
- Share of supplier disruptions handled in the flow
- Outcome
- Days of cover protected; premium purchase cost
Quality · disposition
Quality hold, release or substitute
The decisionA QMS hold, a lab result or a nonconformance blocks a lot. Releasing, reworking or substituting it touches safety, compliance and customer orders.
Who takes partQuality assurance and the operations owners; the rules in the QMS and the laboratory system.
What Aidecis doesLinks the affected lots and orders to the customers downstream, blocks any release the rules do not allow, routes the disposition to the authorised quality owners and records the evidence behind every release or substitution.
How far it can goLevel 1. A release stays with a person; Aidecis makes it faster and fully evidenced, not automatic.
The measureTime to disposition. Also: affected volume, compliance deviations.
Opens nextA customer-promise change for the orders the held lot was meant for.
Scorecard
- Latency
- Time to disposition
- Quality
- Compliance deviations
- Adoption
- Share of dispositions recorded with their evidence
- Outcome
- Affected volume
Manufacturing · recovery
Production rescheduling
The decisionA machine stops, a team is short or material is missing, and the schedule has to change without breaking the commitments downstream.
Who takes partProduction planners and the production owner; the scheduler in the APS or MES.
What Aidecis doesCompares feasible schedules and their downstream effects within the labour, maintenance, quality and customer constraints, asks the production owner to approve, and then grants the approved change, which your MES or APS integration checks before it executes (enforced control).
How far it can goLevel 2 for routine changes inside a set envelope; larger changes stay with the production owner.
The measureTime to a recovered plan. Also: schedule adherence, changeover loss, overtime.
Opens nextA customer-promise change, and an expedite to protect the date.
Scorecard
- Latency
- Time to a recovered plan
- Quality
- Changes outside the approved envelope
- Adoption
- Share of schedule changes approved in the flow
- Outcome
- Schedule adherence; changeover loss; overtime
Logistics · routing
Logistics reroute or carrier choice
The decisionPort congestion, weather or a carrier failure disrupts a lane, and the freight has to move another way.
Who takes partLogistics planners; the routing rules in the TMS; a forwarder's or visibility platform's agent that proposes or books reroutes.
What Aidecis doesCompares route and carrier options on service, cost, emissions and contract terms, checks the approved-carrier list and the site caps, and grants exactly the approved booking change, which your TMS integration checks before it executes (enforced control). A reroute a forwarder's agent books on its own is read back and flagged when it breaks the contract or the cap (observed control).
How far it can goLevel 2 within contracted carriers and a cost cap.
The measureResponse time from disruption to booked reroute. Also: transit time, premium cost.
Opens nextAn invoice variance check on the premium that was booked.
Scorecard
- Latency
- Response time from disruption to booked reroute
- Quality
- Reroutes flagged against the contract or the cap
- Adoption
- Share of reroutes decided on the brief
- Outcome
- Transit time; premium cost
Customer service · prioritisation
Customer promise and backorder priority
The decisionSupply falls short, or a priority order risks missing its date, and someone decides whose delivery moves.
Who takes partCustomer service and the commercial owners; the order-promising rules in the ERP.
What Aidecis doesShows the affected customers and order dependencies, applies the client's priority policy, proposes new dates, splits or substitutions, and asks an authorised commercial owner to approve.
How far it can goLevel 1 to 2. Who waits is a commercial choice, and a person stays accountable for it.
The measurePromise accuracy for the affected orders. Also: on-time-in-full, override rate.
Opens nextAn inventory reallocation or an expedite to hold the date.
Scorecard
- Latency
- Time from shortfall to a new promise
- Quality
- Override rate
- Adoption
- Share of priority changes approved in the flow
- Outcome
- Promise accuracy for the affected orders; on-time-in-full
Logistics · dossier
Expedite freight after disruption
The decisionA late arrival puts a production line or a customer order at risk. A carrier quotes an expedite, and someone has to decide whether the premium is worth it.
Who takes partThe planner and the logistics manager; the carrier-booking rules in the TMS; more and more often, a carrier's or visibility platform's agent that proposes the expedite.
What Aidecis doesPrepares one brief with the options against the total cost of the disruption and the past outcomes at the site, and routes it to the person whose limit covers it. The planner books it in the TMS, and Aidecis reads the booking back and reconciles it against the approval; with enforced control, your TMS integration checks the single-use grant for exactly the approved booking before it books. The invoice is later checked against that approval.
How far it can goStarts at autonomy level 1: agents prepare, a person decides. Small, reversible expedites under a tight cap can move to level 2 or 3 once the record shows they hold.
The measurePremium freight spend per month against the baseline. Also: line-stoppage hours avoided, time to decision, invoice variance.
Opens nextAn invoice variance check against the approved grant when the carrier bills.
Scorecard
- Latency
- Time to decision
- Quality
- Invoice variance against the approved grant
- Adoption
- Share of expedites decided on the brief
- Outcome
- Premium freight spend per month; line-stoppage hours avoided
Finance · reconciliation · later phase
Invoice variance, claim and close-out
The decisionAn invoice differs from the quote, the order, the receipt or the approved expedite.
Who takes partAccounts payable and the finance controllers; the matching rules in the ERP; an invoice-processing vendor's agent that may release payments.
What Aidecis doesMatches the invoice against the approved decision and what actually executed, sends differences outside the tolerance to finance, prepares the evidence for a claim or chargeback, and closes the case only after reconciliation. A payment an agent releases on its own is checked against the tolerance (observed control).
How far it can goLevel 3 is realistic early for small differences inside the tolerance: the check rests on facts the decision record already holds.
The measureLeakage recovered per month. Also: claim success, days to close.
Opens nextA supplier claim or chargeback, and the close-out.
Scorecard
- Latency
- Days to close
- Quality
- Claim success
- Adoption
- Share of invoices matched against the decision record
- Outcome
- Leakage recovered per month
Maintenance · asset failure
Unplanned failure: run reduced or stop to repair
The decisionA priority-1 notification on a critical asset, such as a seal leak on a refinery feed pump, with the spare part days away.
Who takes partThe site owner, maintenance and operations; the maintenance system and its work-order rules.
What Aidecis doesCompares running at a reduced rate, stopping to repair and switching to standby on lost throughput, margin, safety, permits and spare-part lead time, routes the choice to the site owner within the stop-duration envelope, and schedules the work order and the rate plan only after approval.
How far it can goLevel 1. With safety at stake a person decides; it starts as a shadow brief.
The measureTime to decision. Also: throughput protected, safety deviations.
Opens nextA production reschedule, and a customer-promise change where the date is at risk.
Scorecard
- Latency
- Time to decision
- Quality
- Safety deviations
- Adoption
- Share of priority-1 notifications decided on the brief
- Outcome
- Throughput protected
Trading · decision chain
Buy a cargo, and the decisions it sets in motion
The decisionA trader sees a spread to buy an extra cargo against a monthly position limit. The purchase will need tank space, transport and an accrual downstream.
Who takes partThe trader and the head of trading; the terminal and logistics schedulers; the trading system and its limit checks.
What Aidecis doesChecks the commitment against the position limit and routes it to the head of trading. Once it is approved, opens the downstream decisions for tank allocation and barge nomination, each with its own owner, in one chain.
How far it can goLevel 1 for the trade itself; each downstream step can move up once it holds on the record.
The measurePosition within limits. Also: downstream decisions on time, demurrage avoided.
Opens nextTank allocation and a barge nomination, each with its own owner, in one chain.
Scorecard
- Latency
- Downstream decisions on time
- Quality
- Position within limits
- Adoption
- Share of downstream steps opened in the chain
- Outcome
- Demurrage avoided
The decision library
72 consequential decisions in twelve families.
Every consequential supply-chain decision Aidecis governs belongs to one of twelve families. In a guided decision assessment we map with you which decisions are yours, how they score and where to start.
- Trading and commodity positions
- Procurement and supply
- Demand and customer commitments
- Planning and inventory
- Production and operations
- Maintenance and assets
- Energy and utilities
- Quality and product compliance
- Health, safety and environment
- Logistics and transport
- Warehousing and terminals
- Finance and close-out
Book a guided decision assessment
The library holds the structure of each decision, not the answer: your own limits, systems and history are added with you, and examples stay labelled illustrative until real cases prove them.
Built around your operating rules
Start with the decision that costs you the most when it goes wrong.
Use client-defined authority, policies and approved data. Keep people accountable as automation grows, and measure outcomes before increasing a bot's decision scope.
Build your business case