Skip to main content
Open Settings → Agents to view the agents available to your team. Agents are separated into Calls and Messages because voice and messaging channels use different configuration.
The agents and editing tabs you see depend on your team’s products, agent type, and permissions.

Configure a call agent

Select an agent on the Calls tab. Its editor can include:
  • Voice Settings: Voice, customer-facing agent name, new-customer greeting, existing-customer greeting, and internal name and description
  • Agent Settings: The agent’s operational behavior and enabled capabilities
  • Agent Logic: Additional logic, when configured for the team
  • Agent Flowchart: A visual representation of the agent flow, when available
Voice Settings can also show whether the agent inherits the team’s Knowledge Base or uses an explicit override. Prefer the shared Knowledge Base when the information applies across agents; use an override only when that agent needs different instructions.

Configure a message agent

Select an agent on the Messages tab. Depending on the agent type, its editor can include:
  • Agent Settings for escalation, links, follow-up behavior, and channel-specific options
  • Objective for the goal, prompt, assistant name, first message, and available tools
  • Setup Instructions for channel installation or behavior
Message-agent types can include Outbound, Speed-to-Lead, Call Follow-up, Webchat, Google Local Services Ads, Yelp, New Customer Sign-Up, and Genesys SMS. Some campaign agents expose only the settings relevant to their objective.

Choose escalation reasons deliberately

Under Agent Settings → Escalation Settings, enable only the reasons that a team member is prepared to own. Common reasons are disabled until you select them, and workflow editors can expose the same choices for workflow-specific behavior. Two available reasons need special operational handling:
  • The customer asks to stop receiving messages: an exact STOP is handled as an automatic text opt-out. Other wording, such as “take me off the list,” escalates only when this reason is enabled; the receiving team member must complete the opt-out.
  • The customer asks to be contacted again at a later time: enable this when a person should take over requests such as “call me next week.” This escalation does not schedule an automatic future follow-up.
After enabling a reason, confirm the customer-facing escalation message, the internal alert recipients, and who resolves the conversation. Test one matching phrase and one phrase that should remain with the AI.

Configure Call Follow-up messages and attribution

For a Call Follow-up agent, expand First Message → Advanced settings to choose how the first text is written:
  • Leave Contextual first message off to send the configured first-message template.
  • Turn it on to generate a short opener from the broad service issue stated in the call transcript. The configured template remains the fallback when the transcript is unavailable, generation fails, or the generated message does not pass the safety checks.
The same advanced area can include Attribute follow-up bookings to the original call. This setting changes ServiceTitan reporting for a job booked in the follow-up text conversation:
  • Off (default): use the dedicated AI Text Rehash campaign and do not attach the original call.
  • On: retain the original call’s ServiceTitan campaign and call reference.
Changing the first-message mode changes live customer copy. Changing call attribution changes which campaign and call receive credit for later text bookings. Test the message and verify a controlled ServiceTitan booking before relying on either setting in production.

Enterprise-managed message agents

A message agent based on an enterprise template shows a notice that later template updates can modify the agent. The team editor exposes only Objective because Agent Settings and Setup Instructions are managed by the enterprise template. When an enterprise template is pushed or re-pushed to a team, the managed agent is enabled with AI responses on. The phone number remains a team-specific binding and must still be reviewed separately. Before an enterprise push, confirm the target teams, customer-facing objective, escalation behavior, tools, and phone-number readiness; then test one conversation on each affected use case.
An enterprise template push can adopt an existing team-owned agent and immediately make the resulting managed agent live with AI responses enabled. Treat adoption and re-push actions as production changes.

Replace a Yelp objective across an enterprise

When the enterprise has the required feature and you have permission to manage enterprise agent settings, open the enterprise workspace and go to Agents → Messages. The Yelp objective section can replace the channel objective for selected teams in one action.
  1. Review the team selector. Teams without a Yelp agent and teams managed by a template are labeled and skipped.
  2. Select only the intended writable teams.
  3. Paste the complete replacement objective and compare the will be updated and skipped counts with your intended scope.
  4. Choose Apply to n teams, review the confirmation, and replace the objective.
  5. Open each updated team’s Yelp agent and run an approved conversation test.
This action changes only the Yelp objective. Each team’s universal messaging prompt continues to apply before it. The enterprise page does not keep objective version history, so record the current objective outside the form before replacing it.
Replacing an enterprise Yelp objective overwrites live customer-facing instructions on every selected writable team and cannot be undone from that page. Treat the team count, objective text, and confirmation as a production change.

Make a safe change

  1. Open the correct team and agent.
  2. Record the behavior you expect to change.
  3. Update the smallest relevant setting.
  4. Save the change.
  5. Test a realistic customer scenario before relying on it in production.
Greetings, booking behavior, transfer rules, and message objectives are customer-facing. Review the result as a complete conversation, not only as an isolated field.