Your AI Agent can sound like you, or like a carefully shaped version of your brand. This article covers everything in the Voice and personality section: the role, message length, style rules, and message examples. Getting this right is what makes the difference between an agent that feels human and one that feels generic.
You can set this during the setup wizard (Step 1 of 6), or come back to it anytime from the AI Agent dashboard. To edit it later, click AI Agent in the sidebar, then Voice and personality.
Let Inrō draft a personality for you
The first time you open this section, Inrō analyzes your account and your recent sent messages and fills in a suggested role, message length, style rules, and a few real message examples. You'll see an "Analyzing personality" message, then a set of suggestions you can review and adjust.
This is a starting point, not a final answer. Treat the suggestions as a first draft and edit anything that doesn't sound like you.
Role and behaviour
The role sets the main job and attitude of your agent. It shapes what the agent leads with and how it prioritises replies.
There are six roles to choose from:
💬 Community Manager: Engages followers and builds loyalty. Good for accounts focused on growing an audience and keeping people involved. This is the general-purpose default if you don't pick another role.
💰 Sales Closer: Turns chats into sales. Best for e-commerce or direct-offer accounts where the main job is getting people to buy.
🛠️ Support Agent: Answers questions and solves issues. Works well if you get a lot of inbound questions about your product or service.
📣 Content Promoter: Shares posts and boosts reach. Suited to creators promoting content, events, or launches.
🌱 Lead Nurturer: Follows up and warms prospects. Good for service businesses, coaches, or anyone working longer sales cycles.
🎩 Brand Concierge: Gives personalised guidance. Good for premium brands or anyone offering a high-touch experience.
Pick the one that matches the main job you want the agent to do. Click Edit selection to change it later.
Typical message length
This controls how long the agent's replies run. There are three options:
Short messages: Around 10 words, concise and to the point. This is the fallback if you don't choose.
Normal messages: Around 25 words, balanced and informative.
Long messages: Around 50 words, more detailed.
Match this to how you actually write. If your real replies are two short sentences, pick Short. If you tend to add context and next steps, pick Long.
The length you set here applies to the standalone agent. When you call the agent from inside a flow, each AI action can set its own length for that step.
Style rules
This is the most useful field in the section. Style rules are plain-language instructions that guide how the agent behaves in every conversation. Think of it as the brief you'd write for a new team member on their first day. You have 640 characters.
Use them to cover things like:
Language: "English only", or "match the language the contact uses".
Tone and opening: "Always open with a greeting and introduce yourself by name", or "Keep it casual".
Emoji: "Use emojis sparingly to add warmth", or "No emojis". You set emoji behaviour here, in your own words, rather than from a separate control.
What to avoid: Topics, promises, or phrasing the agent should never use.
Recurring patterns: "Always invite a follow-up question", or "End support replies with a next step".
Write in plain language, not prompt syntax. "Be friendly and concise" works fine. Be specific where it matters: "never promise refunds" is more useful than "be careful about refunds".
Auto-like positive messages
This toggle lets the agent heart-react to incoming messages it reads as positive, like compliments or thanks. It's a small touch that can lift engagement with no extra effort.
Each auto-like uses one AI credit, the same as a reply. See AI Credits.
Message examples
This is the feature most people underuse. Message examples are real messages you've sent before, pasted in as references. The agent reads them and learns the texture of your writing: sentence structure, vocabulary, how much detail you give, how you handle different situations.
Style rules describe how you write. Message examples show the agent. Together they produce much better results than either alone.
Aim for four to six examples from different contexts: a warm welcome to a new follower, a reply about pricing, a reply to a complaint or tricky question, a follow-up to someone who went quiet, and a reply to someone ready to buy. Click Add a message example, paste the message text, and save. Repeat for each one.
⚠️ Paste only the message text itself, not the full conversation. The agent learns from how you wrote the reply, not from the context around it.
Saving your changes
In the setup wizard, click Continue to move on. From the dashboard, your changes save when you close the panel or navigate away. To hear the effect straight away, click Chat with Agent in the top right of the AI Agent page.
🐾 Netsuke's Tips
Don't let your style rules and message examples contradict each other. When they disagree, the agent gets confused and leans unpredictably toward one or the other. Make sure your examples actually show the tone you describe.
Don't set the message length longer than you really write. If your messages are short and punchy, picking Long creates a mismatch contacts will notice.
Refresh your message examples every few months. As your voice shifts, so should what the agent learns from. A few new examples can noticeably change the output.
If the agent keeps drifting into a tone you dislike, add a style rule that names the exact behaviour to avoid, like "Never use stiff phrasing such as 'I would be happy to assist you.'" Specific beats vague.
What's next?
With personality set, decide when the agent should speak. Go to Activation Rules: Controlling When Your AI Agent Replies. To tell it what to work toward, see Goals and Instructions: Telling Your AI Agent What to Do.