How an AI agent can create a safe catalog cleanup queue
This workflow gives an AI agent a defined job, a bounded set of records, and a result a person can review. The agent reads the relevant Square context, applies the rules in the prompt, and keeps the source behind every recommendation. It returns a proposed handoff rather than taking consequential actions on its own.
Can an AI agent create a safe catalog cleanup queue?
Yes. Start with the scope, date range, decision rules, and fields that identify the right records. The agent can collect the evidence, compare states or sources, mark conflicts and missing data, and organize the result around the outcome above. A reviewer then checks the matches and judgment calls before approving messages, record updates, bookings, purchases, publishing, or other write actions. The guide below shows the records, boundaries, prompt, and handoff needed for this specific workflow.
Preserve current catalog values
A Square catalog audit starts with a named location, category, or item set and the business rules for required fields, naming, variations, prices, and status. The agent can identify blanks, inconsistent labels, duplicate candidates, unexpected prices, or variations that do not fit the supplied structure.
Each finding should include the item and variation identifiers, current values, rule, and proposed correction. The agent must not merge items based on name similarity alone or assume an inactive record is safe to delete.
Example starter prompt
Audit the Square catalog for [location, category, or item set] using these catalog rules: [rules].
Check item and variation names, categories, prices, SKUs, status, and required fields. For every finding, show the current value, source identifier, rule, and proposed correction. Put duplicate candidates and uncertain ownership in separate sections.
Do not edit, archive, merge, or publish items. Return a cleanup queue for the catalog owner.Review customer-facing impact
Before approval, check whether a proposed name or price change affects menus, checkout, reporting, or integrations. The final handoff should separate safe metadata corrections from changes that require operational coordination.
Questions this workflow answers
Which catalog items are duplicated, incomplete, or inconsistent before the next menu or product update?
The agent reviews a named location, category, or item population against written catalog rules. It records item and variation IDs, names, categories, SKUs, prices, status, modifier or option context where relevant, and required fields. Every finding shows the current value and the rule that made it a candidate. Similar spelling or price does not establish that two records are duplicates.
Duplicate review compares variation structure, location availability, reporting history, integrations, and customer-facing placement. One item may exist twice because locations use different tax or fulfillment setups. Inactive items may still support historical reporting. The agent marks the evidence and asks the catalog owner rather than proposing a merge or deletion from name similarity.
Inconsistencies are grouped by consequence. A missing SKU may affect inventory or integration. A naming mismatch may confuse customers or reports. An unexpected price needs comparison with the applicable location and variation, not a global assumption. The agent also flags rules that the current catalog repeatedly violates, which can point to a process or import problem rather than hundreds of manual edits.
The cleanup queue separates safe metadata corrections, merchandising decisions, price reviews, duplicate investigations, and records with unclear ownership. It includes downstream menu, checkout, reporting, and integration checks for each proposed change. The catalog owner approves and applies updates one by one or in a reviewed batch; the agent never archives, merges, prices, or publishes items during the audit.
The catalog owner chooses and applies each update in Square.
Duplicate review should compare item, variation, SKU, category, modifier set, location availability, price, tax behavior, and sales history. Two “Iced Latte” records may serve different locations or use different modifiers; merging them could break menus and reporting. The agent can identify likely duplicates, explain the matching fields, and show the downstream references that differ. Missing descriptions or inconsistent naming can form a safer correction queue, while price and tax discrepancies go to owners with authority to decide them.