Conversation AI bot explained
Conversation AI is an AI-powered agent that understands incoming messages, uses the information it has been trained on, and responds according to the behavior, goals, and instructions you configure. It handles frequently asked questions, responding to new leads, appointment booking, collecting contact information, and keeping conversations moving when team members are unavailable. Depending on how you configure it, Conversation AI either drafts replies for a human to review or responds automatically in supported channels.
Where to find Conversation AI
Open AI Agents > Conversation AI to manage every aspect of your agents. From there you can view the agent list, create a new agent, open an existing agent, configure settings, training, goals, and instructions, and review reporting.
Response modes
Each agent runs in one of three response modes, which decides whether and how Conversation AI replies to incoming messages.
Off: Conversation AI does not respond.
Suggestive: drafts replies for incoming conversations but does not send them automatically. A human reviews each draft, can edit it, and must manually send the message. Use this mode for AI assistance without full automation, when human approval is required before a message reaches the customer, or when testing a new bot before switching to full automation.
Autopilot: replies automatically based on the agent's training, settings, goals, and instructions.
Primary and non-primary agents
A primary agent handles general inbound conversations for the channels assigned to it. Non-primary agents are typically used inside workflows and respond only when invoked in a workflow.
You can configure different agents for different purposes — general support, appointment booking, lead qualification, or workflow-specific messaging.
Bot configuration areas
Five areas control how an agent behaves: Bot Settings, Bot Training, Brand Voice, Bot Goals, and Additional Instructions.
Bot Settings
Bot Settings holds the agent name, its bot status or response mode, the supported conversation channels assigned to the agent, and primary bot behavior. Advanced behavior controls cover response delays, message limits, sleep/reactivation behavior, and attachment handling. The response style can be set to Concise, Balanced, or Detailed.
Bot Training
Bot Training provides the information the agent uses for accurate responses. Training sources include knowledge bases, documents, URLs, FAQs, and other supported training sources. Open it under AI Agents > Conversation AI > Agent List > Bot Training.
Training is recommended for accurate, business-specific responses. How your data is used in AI features covers training sources in depth, including the 4,000-URL limit for a single knowledge base.
Brand Voice
Brand Voice aligns responses with your organization's preferred tone, language, and style — friendly, professional, concise, or casual — plus brand-specific language rules. It controls the tone and wording of the agent's replies.
Bot Goals
Bot Goals define what the agent should accomplish in a conversation, such as booking an appointment, collecting required details, or triggering a workflow after a successful interaction. Goals can include appointment-booking behavior, questions the bot should ask, information to collect, post-booking behavior, workflow triggers, cancellation and rescheduling options, and notification preferences.
With the correct goals and appointment settings configured, Conversation AI can book appointments on your behalf.
Additional Instructions
Additional Instructions adds extra behavioral rules — conversation guidelines, tone expectations, initial messaging, and specific do-or-don't rules. Examples include maintaining a concise tone, mirroring the customer's language, avoiding emojis, asking one question at a time, and following compliance-friendly response rules.
Using Conversation AI in workflows
Conversation AI can run inside a workflow to generate AI-powered replies, wait for contact responses, and continue the automation based on the contact's reply. Workflow-based use targets a specific automation path rather than general inbound conversations from assigned channels, which is why non-primary agents are typically used this way.
In the workflow builder, the Conversation AI action appears under Communication Actions in the workflow actions catalog, and the Conversation AI Trigger sits under Events triggers in the workflow triggers catalog.
Monitoring your agents
The agent dashboard shows unique contacts, actions triggered, appointments booked, time saved, and past AI interactions. For a closer look at what happened in individual conversations, review activity through the agent logs.