Skip to content
Plain Help Center home
Plain Help Center home

AI Agent

AI Agents are only available to Scale customers.

With Plain you can deploy an AI agent to automatically help customers. It's a simple process to set up and doesn't involve any code.

You can

  • use it on all channels that Plain supports;

  • train it with your own internal knowledge base, documentation or other resources you have;

  • selectively pick which support requests it should handle (e.g. "Only answer chat support requests from Free Tier customers" or "Only from this Slack channel");

  • monitor its behavior and performance.

How it works

By default, the AI agent uses your configured knowledge sources to handle all assigned support requests. Its primary goal is to resolve customer questions automatically and efficiently.

When the agent can't help the customer directly, it will hand off to a human.

  • The agent is always transparent about handoff.
    In all handoff-to-human cases, the agent explains to the customer what is happening to reassure them the request is being handled and that they are not stuck.

  • The agent leaves private notes about gaps in documentation.
    If the agent fails to find relevant information in the knowledge sources provided, it will leave a note on the support request suggesting.

When the agent hands off to a human, it unassigns itself and sets the thread status to Todo.

Early handoff

In some cases, the agent will preemtively hand off to a human without attempting to help further. This happens when

  • the customer is angry;

  • the support request is about sales, pricing or similar;

  • the customer asks to talk to a human;

  • the report is security-sensitive;

  • the report is about downtime or an incident;

  • there are messages from one of your team members. (This is to avoid race conditions where your team decides to pick up support request directly in Slack, MS Teams, etc.)

Automatic priority increases

In addition, the agent will automatically increase the thread’s priority in certain cases:

  • Sets priority to High for security-related requests.

  • Sets priority to Urgent for incidents or downtime reports.

How to set it up

Follow these steps to get set up.

  1. Add your knowledge sources in Plain AI → Knowledge sources. You can find out more on how to do this in Knowledge sources.

  2. Create your agent in Plain AI → AI agent. All you have to provide is a name and avatar for your agent. The customer will always know when they're talking to an agent so you don't have to include that in the name.

  3. Go to any support request and assign the agent to it to have take over the conversation.

When you're comfortable with how the agent handles the request it's been assigned to, you can switch on automatic assignment using Workflow rules.

For example, your workflow rule could specify If channel is chat, and the tier is "Free", then assign AI agent.

Monitoring the agent

To see the support requests the agent has handled, you can either go to any queue and filter by assignee, or you can go to Plain AI → Activity which will show you the threads the AI agent has been involved in.

Powered by Plain