What happens when two users or AI agents change the same product at the same time?

Short answer

Use conflict detection, coordination, and reconciliation rather than allowing competing writes to silently win.

Core explanation

As ecommerce systems become more automated, multiple actors may try to modify the same product simultaneously.

Those actors might include:

• employees
• agencies
• AI agents
• apps
• scheduled automations
• PIM or ERP systems
• bulk import processes

The problem is that all of them can be individually authorized and still create conflicting changes.

For example:

1. An AI agent reads a product title and proposes an update.
2. A merchandiser edits the same title manually.
3. Another automation generates a different update.
4. Each action is valid based on the state it originally observed.
5. Whichever write executes last may silently overwrite the others.

This is not necessarily a permissions problem.

It is a concurrency and state-consistency problem.

A safer workflow may need to track the state each proposal was created against.

Before execution, the system can ask:

Has this product or field changed since this proposal was created or approved?

If the answer is yes, possible responses include:

• block the stale write
• mark the mutation as conflicting
• require re-review
• regenerate the proposal against the latest state
• compare the competing changes
• escalate the decision to an appropriate operator

Controls can also become more granular.

Two actors modifying different fields may not actually conflict.

A description update and an SEO title update could potentially proceed independently, while two changes targeting the same title should require conflict resolution.

This suggests that concurrency control may need to operate at the mutation or field level, not simply lock the entire product.

The issue becomes even more important with multiple AI agents.

Agents can generate and execute changes much faster than humans, increasing the likelihood that several valid-looking mutations are based on different versions of the same production state.

So the important question may not be:

“Which actor should win?”

It may be:

“Which mutation is still valid against the current production state?”

CommerceGov position

CommerceGov’s position is that individual business changes should be governed according to context, authority, policy, and outcome, rather than access alone.

Key concepts

• proposal authority
• approval authority
• execution authority
• risk-based policy
• verified production outcome

Related resources

• QuestionCan multiple AI agents create conflicting actionsYes. Multiple AI agents can create conflicting actions when their objectives, authority boundaries, or assumptions about the same business context are incompatible. The remedy is to define ownership and priority for overlapping decisions, detect conflicts before execution where possible, and route unresolved tradeoffs to an accountable decision-maker.
• QuestionHow do companies prevent stale AI proposals from overwriting newer changesCheck expected or base state before execution and reconsider the proposal when production has changed.
• GuideHow do companies limit the blast radius of AI-generated changesLimit blast radius with limits on batches, fields, stores, downstream effects, and escalation thresholds; authority and impact scope are different controls.
• QuestionWhat should an audit trail for AI-generated ecommerce changes containAudit should connect proposal, policy decision, approval, execution, production result, verification, and reconciliation or rollback.
• GuideHow do you verify that an automated ecommerce change was actually applied correctlyVerification compares intended and approved state with resulting production state. A successful write alone is not proof of a correct outcome.