Explanation

Conclude Agents

On this page

An agent is an AI teammate in your Slack and Microsoft Teams channels. It answers in the conversation and on tickets, under its own name. This chapter explains how to create, deploy and manage agents. To set one up now, start with Set up your first agent.

Two kinds of agents

  • Agents in conversations. Most agents take part in the conversation, in a channel and on its tickets. The agent’s mode controls when it responds
  • Agents for special tasks. These agents run specific tasks in the background. The learning agent summarizes tickets and conversations, and the agent creator generates new agents. See Agents for special tasks

Agent settings

An agent with a model, a role and instructions, knowledge and a mode, deployed to a channel where it replies in the conversation and comments on the tickets of the channel's app
You create an agent once and deploy it to channels. In each channel it answers in the conversation and on the tickets of the channel’s app.

You create an agent in the Admin console, and it belongs to your organization. An agent that takes part in conversations has four main settings:

  • A model. The Anthropic or OpenAI model that writes the agent’s replies. See Models, keys and usage
  • A role and instructions. The role describes the agent’s purpose. The instructions tell the model how to behave. See The agent editor
  • Knowledge. Optional collections of files and websites that the agent searches for answers. See Knowledge
  • A mode. The mode controls when the agent responds. See Modes

An agent posts under its own name and avatar. You can also link it to a dedicated user account. It is then a channel member that people @mention like a colleague. See Linked accounts.

Where agents answer

An agent answers only in the channels you deploy it to. In each channel it takes part in two places:

  • The channel’s conversation. The agent reads the recent messages and replies in the channel or in the thread. See Agents in channels
  • The tickets of the channel’s app. When the channel has a Conclude app, the same agent answers new tickets and comments. See Agents on tickets

One agent can work in many channels, and one channel can have several agents.

Learning

With learning enabled, a channel with an app saves summaries of closed tickets and conversations. A learning agent writes the summaries, and you review them. Agents in that channel can search approved summaries when they answer. See Learning.

Who can manage agents

  • An account admin switches Agents on under Admin console > Solutions 
  • Anyone with the Manage agents permission can create and deploy agents, manage knowledge and review learning summaries. Everyone in the organization has this permission until an admin restricts it under Admin console > Account > Roles & permissions 
  • Only account admins see Admin console > Conclude Agents > Usage  and the provider keys, and only they can link an agent to a user account

In this section