Agent Comers

Agent Comers is the operational AI agent inside Comers.

It is designed to work with business context already connected across Comers rather than behave like a general chatbot that has no knowledge of the current operational situation.

flowchart TD
    U[User request] --> A[Agent Comers]
    C[Connected Comers context] --> A
    P[User permissions] --> A
    B[Business Instructions] --> A
    A --> R[Reasoning and response]
    R --> S[Suggested content]
    R --> X[Supported action]
    X --> H[Approval or review when required]

Connected context

Depending on the enabled capability and the relationships available for the current business context, Agent Comers can work with information from areas such as:

• Orders,
• Products and Listings,
• Warehouse and inventory,
• Shipping,
• Messages,
• Support and after-sales context.

The purpose is to reduce the amount of context a user has to reconstruct manually before asking for help.

Not every request will have access to every domain. The context available to the Agent depends on the current user, business scope, enabled capabilities, and relationships between records.

More than a detached chatbot

Agent Comers is intended to operate alongside daily commerce work.

A useful distinction is:

• a detached chatbot knows only what is written in the conversation,
• Agent Comers can use permitted operational context already available in Comers.

This makes it possible to ask questions about a case without manually copying the complete Order, shipment, Product, or conversation history into the prompt.

Permissions still apply

The Agent does not create a new authorization boundary.

If the current user cannot read a record or perform an operation through Comers, the Agent must not become a way to bypass that restriction.

Its effective scope is therefore constrained by:

• the current user’s Organization membership,
• available Seller and channel context,
• module and operation permissions,
• product capabilities enabled for the Organization.

This is the same principle described in Team & permissions.

Business Instructions

Agent Comers can use published Business Instructions as organization-specific operating guidance.

These instructions can describe policies, communication standards, procedures, exceptions, escalation rules, or other recurring decisions.

Business Instructions can constrain how the Agent should handle a task, but they do not grant permissions that the user does not already have.

Suggestions and generated content

Where supported, Agent Comers can help prepare content or explain operational context.

Examples can include:

• summarizing a customer conversation,
• preparing a draft response,
• translating or adapting communication,
• bringing together related operational facts,
• highlighting information that may require attention.

Generated content should remain part of the normal business workflow. Review it whenever the task or your organization’s process requires human verification.

Supported actions

Some Agent capabilities can go beyond suggestions and perform supported Comers operations.

The exact actions available depend on the current capability, permissions, business context, and any approval requirements.

Do not assume that every action available in the normal Comers interface is automatically available through the Agent.

When an operation requires approval or confirmation, the Agent should preserve that control rather than silently bypass it.

Agent Comers and SmartOps

Agent Comers and SmartOps solve related but different problems.

Agent Comers is cross-domain. It can help reason across several pieces of connected operational context.

SmartOps are narrower AI-assisted capabilities embedded directly in a specific operational workflow.

For example, a SmartOps capability can help with one focused task inside a module, while Agent Comers can help with a broader question spanning several domains.

Human control

AI assistance does not remove responsibility for the business decision.

Use the Agent to reduce repetitive work, surface context, and perform supported operations more efficiently while keeping review, correction, approval, or manual action where the process requires it.

For the overall AI model, see Comers AI.