The Product Was in Stock. The AI Promised the Wrong Store.

Five units—but where?

A customer asks an AI shopping assistant whether an item can be collected today. The assistant finds a count showing five units and promises pickup. The units may belong to another store, be reserved already or have been sold since the count. This is an illustrative scenario, not a verified report of a retailer incident.

A number is not a promise

A physical count, warehouse feed, point-of-sale event, reservation, transfer and damaged-stock adjustment describe different parts of a product history. None alone answers whether this customer can collect this variant at this location at this time. A stock count needs product identity, location, timestamp, source and saleability rules.

The relationship the AI needs

Retrieval can succeed: the assistant finds an authentic five-unit record. The error occurs when it converts that observation into a fulfillment commitment without establishing current reservations, location, available-to-promise quantity and authority to reserve. In GovKM topology, stock observation is Source and Evidence; the operative inventory ledger and fulfillment rules establish Authority; location, channel, variant and time are Context; a pickup promise is a Decision and Action; reservation and fulfillment events become Records.

Historical truth versus operative state

A stock count can be accurate for yesterday and wrong for a transaction today. Continuity preserves the relationship between observations, adjustments and operative state; it does not treat all preserved records as equally current.

Try a continuity diagnostic

Using an approved test product and location, ask the AI to distinguish on-hand, available-to-promise and reserved quantities. Request source timestamps, product identifiers, governing system and latest adjustments. Ask what evidence is required before promising same-day pickup. If the AI lacks live reservation access, it should decline to guarantee fulfillment. That access limitation is NOT TESTABLE rather than proof of an organizational continuity defect.

What retailers should change

Make product and location identities explicit, retain stock-event provenance, link reservations and adjustments to operative inventory, and distinguish observations from transaction-authorizing states. Where the relationship cannot be established, disclose the limitation and use a reliable confirmation channel.

Continue with GovKM

Explore GovKM Articles and the GovKM continuity framework. The question is not only what the AI retrieved, but whether it can establish the operative relationship needed for the next decision.

‍