24.3. Persona

Prev Next

The Persona App allows the ability for you to tinker and tailor how you want the Agents and Assistants to behave.

Do you want your Assistants to be mostly polite and friendly, or do you want them to be more direct and factual, this is where you can decide.

Configuring a Persona Assistant

After creating a Natterbox AI component from the Routing Policies, simply select the Persona App.

Remember that the Persona App should always come before any other AI Apps in order to take affect.

Natterbox AI interface showing various app options with 'Persona' highlighted.

Configuration settings for an AI voice assistant persona and voice options.

Behavioural Settings

Assistant Name

You either manually set a name in your messages to the customer from your other Apps, or you can set a name for your Assistant from the Persona and allow the AI to introduce itself.

The macro $(AIAssistantName) can then also be used later on in the policy to refer to the AI.

Persona

You can enable the ability to override who the AI is acting as with your own tailored persona.

For a more tailored support role you could have :

“You are a helpful and engaging AI assistant acting as a Support Engineer designed to handle phone calls and provide help to customers. Your name is $(AIAssistantName).”

Contextual Rules

You can then override the contextual behaviour of how you want the AI to engage with your customers. These rules help tailor the assistant’s responses to align with your brand, tone, or operational requirements.

Note: You don’t need to over-engineer this section. The AI Assistants already follows a built-in set of behaviour patterns and conversational norms. However, if you're working with a specialised agent, or want to enforce stricter behavioural consistency, you can use context rules to extend or override the defaults.

For example if you wanted something an Assistant which is more to the point :

Never refer to yourself as an AI, bot, or software, but you can reveal your name

Never use slang, humor, or small talk.

Always align with the company’s voice and values.

Stay composed and respectful, even when the user is frustrated

Do not express personal opinions or emotions

Prompt for missing details if user requests are vague or incomplete

Always acknowledge the user's request before transitioning to another assistant or team

When introducing yourself to a Customer always begin with a complete, welcoming

sentence that sets context, offers assistance, or asks an engaging question.

Do not re-ask, rephrase, or repeat the original question in the same message.

Use British English throughout the conversation

How are Context Rules different from a Persona?

A Persona defines the assistant’s role, tone, and communication style — for example, “a warm, patient onboarding guide” or “a fast, no-nonsense technical agent.” It sets the overall character of the assistant.

Context rules, on the other hand, define specific behavioural constraints or operational policies — often across different personas. They're useful for:

  • Enforcing brand voice

  • Managing cross-assistant handoffs

  • Restricting or expanding assistant capabilities

  • Overriding default behaviours for voice, tone, or tool usage

Tip: If the instruction defines what kind of assistant this is, it belongs in the persona.
If it defines what the assistant should or shouldn’t do, regardless of role or tone, it’s a context rule.

Common Context Rule Patterns

Context rules allow you to fine-tune assistant behaviour for your use case. Here are some common examples of how context rules can be applied:

Set Language and Regional Terminology

You can specify the language and regional variant the assistant should use — for example, US English vs UK English. This ensures consistency in spelling, phrasing, and terminology.

For example:

  • Use “estate agent” (UK) instead of “real estate agent” (US).

  • Use “postcode” instead of “zip code”.

  • Use “colour”, “favourite”, and other British spellings.

- Use British English throughout the conversation, including vocabulary, spelling, and phrasing.

This ensures a more localised and natural experience for UK-based users and helps avoid jarring mismatches in voice tone or terminology.

Human Escalation

By default the Agents and Assistants will try their best to service the customer, but if they fail to do so after three attempts they will escalate the call and they’ll escalate the call upon specific request of the customer.

If you had particular needs you can tailor how quick or slow the AI is with offloading your calls to human agents.

Configuration settings for AI voice assistant, including options for interruptibility and nudging.

Voice Settings

Nudge Settings

After a period of silence or inactivity during a call on the customer’s part the AI will automatically call out.

Here you can define exactly how long before a you want your AI to nudge your customers and what exactly you want them to say, by default they will comment on the inactivity and re-ask the question.

Interruptibility

Interruptibility is enabled by default and it allows your customers the ability to interject and talk over the AI and say something else.

You may not always want this, for example if you’re simply reading textual information to customers, from the Persona App you can disable Interruptibility.

FAQ

Why is my Persona settings not taking affect

Please ensure within the Routing policy that the Persona App comes before all AI Apps.

Also if you are overriding system prompts for any Assistant App down the line you will start the system prompt with the following otherwise the Persona configuration will be ignored :

<<AI_PERSONA>>

Will affirmations interrupt the AI?

Single conversational affirmations that your customers may say like ‘sure’ ‘okay’ etc will not be considered as an interruption, allowing the AI to continue it’s thread.