Business Context for Ecommerce AI: Shared Company Context

A product record says what a thing is: material, capacity, price, category. Business context says why it matters: which market it sells in, whether its collection is a priority, who buys it and what they ask before buying, how you position it, and which claims stay fixed. With both, Paz can tell an important correction from a suggestion you would ignore.

Two bottle collections are missing capacity in their descriptions. One is the range you are promoting this quarter; the other is being phased out. The product information points to fixing both. The context says fix the growth range first. That ordering is the difference between a list of issues and a plan.

Retrieval-augmented generation, or RAG, pulls relevant information into a model response. Business context tells the model which sources to trust, which priorities apply now, and which rules never bend.

In practice that means a populated capacity field answers the product question, and an instruction like "prioritize the travel range and keep approved fabric claims unchanged" shapes the action. Both are needed for a response that fits your business.

Your buyer personas are part of the context. A running shoe sold to first-time marathoners needs cushioning, drop and break-in notes on the page. The same shoe sold to experienced racers needs weight, stack height and plate material. The product is identical. The product information that earns the recommendation differs.

Paz uses the personas you describe, plus the questions they ask in reviews, support tickets and AI assistants, to decide which attributes to complete first, which questions a page should answer and what tone to write in. The copy comes out in your brand voice, for the shoppers you sell to, and goes to your queue for approval like any other fix. A Paz expert gathers this with you at the start and keeps it current as your range and audience change.

Each fact Paz uses carries its source and how fresh it is. A specification from your catalog is treated differently from an inference drawn from a review. A standing rule such as "no medical claims" is treated differently from a quarterly priority. When two sources disagree, the conflict is recorded and resolved against the one you have named as authoritative.

The task or Routine decides which context is relevant to the check at hand. Publishing permissions decide which destination and fields a resulting fix may change. Connected signals carry their reporting period, so you can see how current the evidence behind a finding is.

Context keeps building. Every approved fix, every rejected suggestion, every answer your team gives to a question from Paz becomes part of what the next check knows. A confirmed campaign pause steers a traffic investigation toward the affected channel and period. A rejected description change tells the next run to leave that claim alone.

Change a priority or enter a new market and the next round of work changes with it. The result is a picture of your business that gets sharper with every round.