AI Employee Types

Every AI Employee (AIE) is defined by a workflow — a directed acyclic graph (DAG) of typed nodes that the workflow engine runs. The workflow is the logic. The interaction type is how end users reach that logic: the interface they use, the shape of the input it accepts, and the shape of the output it returns.

You choose the interaction type when you create the AI Employee, and it shapes the rest of the build — which tabs appear in the AI Employee builder, which starter workflow you begin from, and how the AI Employee is invoked once it's published.

The types you can create

When you start a new AI Employee from scratch, the create dialog offers three interaction types: Chat, Dashboard, and Voice. (Templates may also use these types — the badge on each template card shows which.)

Chat

The default and most common type. A conversational AI Employee that takes a natural-language message and streams back a text response.

• Best for: internal knowledge assistants, policy and HR question-and-answer, IT and support help desks, and any general-purpose assistant your team talks to.
• How it works: a person types a message; the workflow runs — typically retrieving from a knowledge base, calling Tools, and reasoning over the results — and returns a written answer with source citations.
• Where it lives: Chat AI Employees expose a Chat tab in the AI Employee builder, where you and your team can hold a live conversation with the published AI Employee. They can also be embedded in a chat widget or driven over the API.
• In the create dialog: described as "Best for open-ended conversations and Q&A."

Dashboard

A bulk, table-oriented AI Employee. Instead of a back-and-forth conversation, it processes many structured inputs — rows of data or batches of documents — and returns structured results in a table.

• Best for: analytical work over large volumes of unstructured data, bulk document processing, and extraction or rule validation at scale.
• How it works: you provide a set of inputs (for example, a batch of contracts or a list of records); the workflow runs once per input and writes a row of structured output. Each result is traceable back to its source, and the run history is available per row.
• Where it lives: Dashboard AI Employees open to a Dashboard tab in the AI Employee builder that shows the table of results. Because each row carries its own execution history inline, Dashboard AI Employees don't have a separate Audit tab.
• In the create dialog: described as "Best for monitoring metrics and viewing structured data."

Voice

A spoken AI Employee that handles phone calls in real time using speech-to-text and text-to-speech.

• Best for: phone-based customer support, replacing or augmenting an IVR, and other voice-first assistants.
• How it works: a caller speaks; the AI Employee transcribes, runs the workflow, and replies with synthesized speech, taking natural conversational turns.
• Where it lives: when you create a Voice AI Employee, Ema provisions the voice agent and takes you straight to the AI Employee's detail page. Voice AI Employees expose a Voice Configuration tab — for voice characteristics (language, voice model), conversation behavior, call settings (including call forwarding), and phone-number assignment — alongside the shared Configuration tab.
• In the create dialog: described as "Best for handling phone calls with real-time AI conversations."

What every type shares

Regardless of interaction type, every AI Employee shares the same set of builder tabs and capabilities:

• Configuration — the workflow editor, model configuration, and publish controls. The workflow logic is independent of the interface, so you build it the same way for any type.
• Details — name, description, group, icon, and the interaction type (shown read-only after creation).
• Permissions — who can view, run, and edit the AI Employee.
• Knowledge bases — the data the AI Employee retrieves from.
• Tags, Metrics, and Activity — organization, usage analytics, and a change log.
• Audit — a per-run history of executions. Chat and Voice AI Employees have a dedicated Audit tab; Dashboard AI Employees show the same history inline per row, so they don't have a separate one.

All types are backed by the same workflow engine and can be invoked over the API, no matter which interface is their primary surface. See the API Reference.

Choosing a type

A note on other types

You may see other interaction-type labels in the platform — for example Form (used by some structured-intake templates) and Recruiter. These aren't offered in the standard create-from-scratch dialog:

• Form AI Employees are created from templates that collect structured input through a form rather than a free-text chat.
• Recruiter is a distinct product with its own onboarding and trial experience; see Recruiter.

What's next

• Ready to build? Walk through Create Your First AI Employee.
• Not sure what to build yet? Use Effective Discovery to scope a high-value use case first.