1. Home
  2. Helpie
  3. AI Chat
  4. AI Persona

AI Persona

Out of the box, Helpie’s assistant is competent but anonymous. The persona settings are where it becomes yours — a name, a tone of voice, an opening line, and your own house rules for how it handles the odd stuff. None of it changes what the assistant knows (that’s the knowledge base); it changes how the assistant sounds and behaves.

Everything on this page lives under Helpie > Settings > AI, in the Assistant persona and Escalation voice sections.

One thing worth knowing before you customise: the assistant always speaks in the first person as part of your business — “we”, “our”, “us” — using your site name and tagline. It never refers to your business in the third person, and it never pretends to be a human. It’s openly an AI assistant that’s openly on your team.

Assistant Name

The name the assistant introduces itself with. The default is “Helpie”; you can rename it to anything up to 40 characters, or leave the field blank for an unnamed assistant that simply says it’s your site’s assistant.

A name shows up in the auto-composed greeting and in how the assistant refers to itself. If your brand has a mascot or a distinct character, this is where they get a job.

Voice

The Voice dropdown sets the tone the assistant writes in. Five options:

  • Calm — measured and reassuring
  • Friendly — warm and approachable (the default)
  • Chatty — casual and upbeat
  • Professional — polished and precise
  • Playful — light humour and personality

The facts in an answer don’t change — the delivery does. The same shipping question, two ways:

> Professional: “Standard shipping within Australia takes 3–5 business days. Express options are available at checkout — details here: [link]”

> Chatty: “Good news — standard shipping’s 3–5 business days here in Australia, and there’s an express option at checkout if you’re in a hurry! Here’s the full rundown: [link]”

Pick the one that sounds like your emails already sound. If your customers would be startled to receive a playful reply from you, they’ll be startled to get one from your assistant too.

Opening Greeting

The first message a visitor sees when they open the chat. It’s static text — no AI call, no cost, instant — so it’s also the one message you have word-for-word control over.

Leave it blank and Helpie auto-composes one from your assistant name and site name: “Hi! I’m Helpie, Your Store’s assistant — ask me anything and I’ll find it fast, or I can pass your question to the team.” Write your own to match your voice:

> Before (auto): “Hi! I’m Helpie, Wool & Co’s assistant — ask me anything and I’ll find it fast, or I can pass your question to the team.”

> After (yours): “G’day! Sizing questions, shipping times, care instructions — fire away. Anything I can’t answer goes straight to the humans.”

Behaviour & Boundaries

This is your house-rules field: owner-authored guidance the assistant follows in its own voice. It’s for personality and things to deflect — not facts. Facts belong in your content (and route rules), so there’s only ever one place to update them when they change.

Good uses:

> “If asked for passwords, logins, or security details, refuse with light humour — you’d lose your job.”

> “If someone asks about wholesale, don’t quote prices — warmly point them to the wholesale enquiry page instead.”

Not-so-good uses: “Our shipping is $9.95 flat rate.” That’s a fact — put it on your shipping page, where the assistant will find it, cite it, and where you’ll actually remember to update it when it changes.

The built-in safety rules (declining legal, medical, and financial advice and routing to a human) always apply underneath whatever you write here — your boundaries add to them, they can’t remove them. See Connecting the AI for how those refusals work.

Escalation Voice: Who’s the Human?

When the assistant can’t help and routes to a person, who is that person? The Escalation voice section lets it say so warmly. After several off-topic questions in a row, the assistant shifts from “let me open a ticket” to a more human “you might just want to email them directly” tone, using these details:

  • Their name — first name is plenty. Leave blank and the assistant uses generic “our team” framing.
  • Their role — how the assistant introduces them. “Sarah, the owner” reads better than just “Sarah” when handing someone over.
  • Their pronouns — she/her, he/him, or they/them, so the assistant writes “send her an email” naturally rather than dodging the sentence.

The difference in feel:

> Without: “I’ll pass this to our team — fill in the form below and someone will get back to you.”

> With: “This one’s beyond me, honestly — Sarah, the owner, is the person you want. Pop your details in below and she’ll get back to you.”

Small thing, big warmth. Escalation is the moment a visitor is most frustrated; a named human on the other end helps.

Common Questions

Do persona changes cost anything or require re-indexing?
No. They take effect on the next message — no re-index, and the settings themselves are free to change. (The answers the assistant generates still cost the usual per-message amount; see Connecting the AI.)

Can I test a new voice before customers see it?
Yes — the AI Tester (Helpie > Settings > AI Tester) uses your live persona settings, so you can rewrite, test, and tweak privately even while the public widget is on.

Will the assistant ever claim to be a real person?
No. It represents your business directly and speaks as part of the team, but it’s instructed never to pretend to be human, regardless of voice or boundaries settings.

How can we help?