Business Context for Ecommerce AI: Shared Company Context

A product record describes material, capacity, price, or category. Business context adds the intended market, collection priority, positioning, and constraints on claims. These inputs help an AI task distinguish an important correction from an irrelevant suggestion.

For example, two bottle collections lack capacity details in their descriptions. One is the current focus range; the other is approaching phase-out. The facts support correcting both, but the priority directs attention to the focus range first. Context makes that reasoning explicit rather than leaving the AI to invent a business objective.

Retrieval-augmented generation, or RAG, brings retrieved information into a model response. It can supply part of the context, but retrieval alone does not establish which source is authoritative or whether a commercial priority is still current.

Shared context can combine structured product records, supplied instructions, and retrieved notes. It is not just a vector database, and it does not require retraining the model. Catalog completeness is also different: a populated capacity field answers a product question, while “prioritize the focus range and retain approved claims” guides the action.

Identify the source and freshness requirement for each important fact. Distinguish a product specification from an inference and a standing constraint from a temporary priority. If sources disagree, preserve the conflict rather than treating the most recently retrieved text as truth.

Unknown margin, stock, or performance specifications remain unknown until supported by an authorized source. The task or Paz Routine determines which context is relevant. Execution permissions determine what may change; adding a fact to context does not grant permission to update a storefront.

In Paz, connected signals carry reporting periods, source quality, and freshness. If a refresh fails, the last successful results can remain visible with their original timestamp and a warning. This preserves usable evidence without presenting it as a new observation or proof of recovery.

Paz Routines define recurring checks; shared follow-ups develop investigations of specific findings. A follow-up can ask for a missing fact through Inbox, such as whether an acquisition campaign was paused. Paz can reuse the answer for another relevant investigation or link to an already open question.

Relevance depends on the account, catalog, and reporting period. An answer about one campaign is not a universal explanation for lower traffic. Answers remain attributed context, distinct from connected-system measurements. They can help choose the next check without proving a cause or authorizing a product update.