> For the complete documentation index, see [llms.txt](https://docs.praiz.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.praiz.io/ai-agents/what-is-an-agent.md).

# What is an agent?

An agent reads a transcribed conversation and produces a structured output: a summary, a score or detected signals.

It takes a transcript as input. It always fills in the same fields, defined in advance: a summary, a score, a list of detected signals. Every call is filled in the same way. The data is therefore comparable from one call to the next.

Each output feeds your conversational database. There it can be filtered and analyzed. Agents apply to your sales calls, your customer calls and your internal meetings.

Praiz offers a [library of ready-made agents](/ai-agents/ai-agent-library.md). You can also [create a custom agent](/ai-agents/creating-a-custom-agent.md).

![AI Agents page with the workspace's agents](https://3785996597-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FU54oKzNb3EyinriyUP5P%2Fuploads%2FGsNUYDSqnxc59Neqyrn5%2F17-agents-page.png?alt=media)

## The three types of agents

| Type       | What they produce                                                                      | Example agents                                                                            |
| ---------- | -------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| Generation | Ready-to-use deliverables: summaries, follow-up emails, customer stories, testimonials | Sales Call Summary, Sales Follow-up Email, Customer Case Extractor, Testimonial Generator |
| Scoring    | An evaluation of the call against a method or your playbook                            | MEDDIC Scorecard, Cold Call Scorecard, Objection Scorecard                                |
| Tracking   | Structured signals: objections, competitors mentioned, churn risks, product feedback   | Competitors Tracker, Objection Tracker, Churn Risk Signals                                |

## Public agents and private agents

* A public agent can be used by the whole company and applies to every user's calls. Only admins can create or import one.
* A private agent applies only to its creator's calls and cannot sync to the CRM. A non-admin user can create one if the "Non-Admin Private Agents" setting is on.

## Where agent outputs go

Every agent output is stored in your conversational database. From there, it can be:

* delivered automatically to the right person, by email or on Slack. See [Smart Actions](/automate/smart-actions.md);
* synced to your CRM. See [CRM sync](/automate/crm-sync.md);
* aggregated with all of your conversations to surface trends. See [Insights](/query-and-analyze/insights.md).

For the difference between querying one call, analyzing all of them and querying from an external AI tool, see [Key concepts](/get-started/key-concepts.md).