Persona

Give the Informly chat assistant a name, role, tone, and target audience so its replies sound like part of your brand.

Persona controls how the AI presents itself in conversation. It doesn't change what the AI knows — that's a function of the documents attached to the project — but it changes how the AI introduces itself, the words it picks, and the level of formality it uses. A well-tuned persona makes the widget feel like a member of your team instead of a generic bot.

You'll configure this once per widget. Two widgets in the same project can have different personas, which is useful when you have a casual marketing chat and a more formal in-app support surface drawing from the same content.

Where to find it

Go to Widgets → (your widget) → Persona.

What you configure

Field	Example
Assistant name	"Maya", "Ask Acme", "Support Bot"
Role	"Acme's support assistant", "marketing concierge"
Tone	`friendly`, `formal`, `casual`, `professional`
Areas of expertise	"billing, refunds, account setup"
Target audience	"small-business owners new to Acme"
Response language	The language the AI replies in

Configuring the persona

Name the assistant

Pick something short and brandable. "Ask Maya" works well for a brand called Maya; "Acme Assist" reads cleanly for a company called Acme. The name appears in the chat header and in the assistant's first message.

Set the role

Describe what the assistant is there to do in a single short phrase. This becomes part of the system prompt, so be concrete — "help customers troubleshoot their Acme router" is more useful than "help users".

Choose a tone

Pick from the four presets. friendly and casual are good for consumer brands; formal and professional suit B2B, finance, and healthcare.

List areas of expertise

Comma-separated topics the assistant should focus on. This nudges the AI to stay on-domain and politely redirect off-topic questions.

Describe the audience

A one-line description of who the assistant is talking to. The AI uses this to pick vocabulary and depth — "developers integrating our API" gets a different reply than "first-time customers shopping for a plan".

Set the response language

Pick the language the AI should reply in. If you serve multiple languages, run one widget per language rather than relying on auto-detection.

Save

Click Save persona. The change applies to new conversations immediately; in-flight conversations finish under the previous persona.

What persona doesn't do

Persona only changes how answers are framed. It doesn't add or remove information from the knowledge base — that's controlled by the documents assigned to the project. If the AI doesn't know something, tuning the persona won't fix it; you need to add or improve a document.

If the AI is giving wrong answers, look at the documents, not the persona. If the AI is giving correct but stiff or off-brand answers, the persona is the right knob.

Examples

A few persona setups that work in practice:

Consumer brand — name "Ask Maya", role "Maya's shopping concierge", tone friendly, audience "shoppers browsing the site"
B2B SaaS — name "Acme Support", role "Acme platform support assistant", tone professional, audience "engineers and admins integrating Acme"
Healthcare — name "Caremind", role "patient information assistant", tone formal, audience "patients looking up appointment and prescription info"