Your AI Agent needs more than a personality to perform well. It needs to know what it's working toward. The Goal and instructions section sets out what the agent works toward and the rules it always follows.
This is separate from Configuring Your AI Agent's Personality and Tone, which covers how the agent sounds. Goals and instructions cover what the agent is trying to do.
To open it, click AI Agent in the sidebar, then Goal and instructions. The first time you open it, Inrō analyzes your account and pre-fills a suggested goal and instructions, which you can then edit. Who you are and what you offer goes in About, in Voice and personality.
Goal
What your agent works toward, in up to 500 characters. For example: "Answer questions, then ask for the size and number of pieces to send a quote."
Be direct and specific. "Get people to book a discovery call" works better than "encourage engagement". If a conversation could end several ways, the agent leans toward the goal you've named.
The goal applies when the agent replies on its own and in AI steps that have no task of their own. When an automation gives the agent a task, for example in a Hand over to AI Agent step or an AI Agent message with instructions, that task replaces the goal for that conversation. Public comment replies leave the goal out.
Instructions
Rules the agent follows everywhere, including in automations and comment replies. Click Add an instruction for each one. An instruction is a single sentence of up to 300 characters, and you can add up to 50.
Good things to put here:
Things the agent should never do or say: "Never promise a delivery date."
How to handle requests you can't fulfill.
Specific steps to follow for common situations.
Any quirks of your business the agent should know.
When your knowledge access is set to Limit to instructions, the agent replies only when it can find an answer in your instructions and knowledge. See Knowledge Base: Teaching Your AI Agent What to Know for that setting.
⚠️ Keep instructions clear and direct. Vague or contradictory instructions produce inconsistent replies. If you're unsure whether a rule belongs here or in voice rules: voice rules are about how the agent writes, instructions are about what it does or avoids.
Saving your changes
Changes save when you close the panel or navigate away. To see the effect straight away, click Chat with Agent in the top right of the AI Agent page and send a few test messages.
🐾 Netsuke's Tips
Write the goal as a specific action, not a vague outcome. "Book a free consultation via the Calendly link" beats "drive conversions" because the agent knows exactly what to push toward.
If the agent keeps giving off-brand answers, the About text in Voice and personality is the first place to look. Adding context about your positioning and who you talk to often fixes it without touching anything else.
Use instructions for your most common edge cases first. Think about the last few conversations you had to take over by hand, and cover those.
What's next?
With goals set, give the agent the knowledge to back them up. Go to Knowledge Base: Teaching Your AI Agent What to Know to add the content the agent draws on when it answers questions.