In this article, we’ll cover:

• Image of the result.

• Directiveness in instructions.

• Behavior triggers.

• Starting the dialogue.

• Ending the dialogue.

Image of the result

Before writing the bot’s legend, it’s important to understand who it is and what role it plays. Without this, it will give irrelevant answers, overstep its competencies, and mislead the participant.

Define the character’s behavior model: client, employee, or doctor. Also determine which questions it answers and which it doesn’t, whose voice it speaks in, what it knows, and what it doesn’t know.

Directiveness in instructions

Directiveness means strict, direct instructions that set clear behavioral boundaries for the bot. They leave no room for interpretation and help it play its role accurately.

Whether you should demand specificity from the bot depends on the goals of the scenario. In one case, you can set strict boundaries for it. In another, you can skip directives so that the participant directs the bot on their own.

Examples of phrases you can use in a prompt:

• You must
• You are forbidden to
• Never ask questions

🤓Examples of using directives:

• You are forbidden to ask clarifying questions. You answer only the questions the participant asked directly.
• You must give short answers, strictly within the information you have. If you don’t have the information, say “I don’t have data on this matter.”

Behavior triggers

Behavior triggers are conditions under which the bot changes its communication style, tone, or dialogue strategy. For example, it shifts from skepticism to trust or from short answers to detailed ones.

To make the bot behave naturally, define these transitions.

🤓Example of a behavior trigger:

• If the employee asks three open-ended clarifying questions, the client reveals their needs and shifts from skepticism to trust.
• If the employee provides a specific example or case from their practice, the client shifts from a formal tone to a more lively and engaged one.

Starting the dialogue

The bot always starts the dialogue first. You can decide the exact phrase it will use to begin the conversation. To do this, specify this condition and write out its first phrase.

🤓Example:
You start the dialogue first and say “Good afternoon!” or “I’m listening.”

Ending the dialogue

To prevent the bot from endlessly maintaining the dialogue and asking follow-up questions when the conversation is already exhausted, define the conditions under which it ends the conversation.

🤓Example:

• If the bot encounters rudeness, for example, hears the phrase “You’re talking nonsense,” it falls silent and ends the dialogue.
• After three questions, the client exits the dialogue.

Need Help 

Write to us in the chat — we're always happy to help 💬