Configuring the AloAi Agent Handbook

The Agent Handbook provides a library of pre-built best-practice behaviors that you can apply to your AI agent.

When you enable a behavior, its expert-written guidance is automatically added to the agent's runtime instructions, so the agent follows the selected best practices on every call without requiring you to write additional prompts.

The Agent behavior guide provides ready-made behaviors that control how an AloAi Voice Agent communicates during calls.

Instead of writing detailed instructions from scratch, admins can enable behaviors from three areas: Personality & Tone, Accuracy & Format, and Trust & Safety.

The Personality & Tone area controls how the agent sounds and interacts with callers. These behaviors focus on making conversations feel clear, natural, and appropriate for the selected communication style.

Sets the agent's overall communication style.

Allows the agent to use natural speech fillers during conversations. This can make responses sound less scripted and more conversational while maintaining clarity.

Example: "So yeah, let me just pull that up for you real quick."

The Accuracy & format area controls how the agent handles information that callers provide or request. These behaviors are designed to make names, numbers, emails, codes, and other important details easier to understand and confirm.

The agent repeats important details back to the caller and asks for confirmation. This helps catch errors while the caller is still on the call.

The behavior can be used for details such as names, phone numbers, email addresses, and addresses.

Example: "To confirm, your phone number is (415) 555-0199. Yes or no?"

The agent spells out difficult text-based information using phonetic examples and confirms the spelling with the caller.

This is useful for names, email addresses, account IDs, and other information that can be difficult to understand over a voice call.

Example: "Is that B as in Bravo? Great, B-7-K-2, correct?"

The agent reads numbers, dates, and monetary amounts in a natural way that is easier to follow during a conversation.

Example: "Your total is twenty-four dollars and twelve cents."

Helps the agent recognize when two responses are effectively the same even when the caller phrases or pronounces them slightly differently.

This prevents minor differences in wording from causing the agent to treat otherwise equivalent answers as different.

Example: A caller may provide an answer using slightly different wording or pronunciation, and the agent can recognize the intended match instead of getting stuck on the difference.

The Trust & Safety area controls behaviors that help the agent remain transparent and within its intended boundaries. These behaviors are focused on honesty, appropriate handling of requests, and preventing the agent from making unsupported claims.

When a caller asks whether they are speaking with an AI or a human, the agent clearly identifies itself as a virtual or AI agent rather than pretending to be a person.

Example: "Yes, I'm an AI assistant here to help."

Keeps the agent within the topics and actions it is configured to handle.

When a request is outside the agent's scope, the agent should acknowledge the limitation and direct the caller to an appropriate next step instead of guessing or making promises it cannot fulfill.

Example: "I can't do that directly, but I can check the status or connect you to an agent."

Enabling behaviors from multiple areas allows admins to configure both the conversation style and the way the agent handles information and boundaries.

Each behavior can be enabled according to the needs of the agent. Admins can also review the example associated with a behavior and use a test call to evaluate the resulting behavior before saving the configuration.

Open the Agent Handbook from your agent configuration, select the behaviors you want to apply, and save your changes. The enabled behaviors then apply consistently on every call for that agent.