Skip to main content
All CollectionsAI Agent
Best practices for setting up your AI agent
Best practices for setting up your AI agent

Optimize your AI Agent’s configuration with these targeted tips to deliver personalized, on-brand, and engaging Instagram interactions.

Updated today

Ready to refine your AI Agent setup? Below are the top best practices for leveraging key features—surveys, data training, persona settings, scenario actions, and more—to create a seamless, intelligent, and interactive experience for your audience.

A. Choosing Your Activation Method

When to Use Global Activation

  • Ideal for a brand or dedicated persona account where every message should receive an automated response.

  • Example: A 24/7 virtual storefront that greets and assists all incoming users.

When to Use Scenario-Specific Activation

  • Useful for controlled engagements where the AI only responds if certain triggers or keywords appear.

  • Example: A lead-qualification flow that prompts the AI to engage only after a user requests a quote.


B. Harnessing Surveys for Personalization

  1. Design Targeted Surveys & Save Custom Properties

    • Create focused questions that uncover a user’s industry, challenges, or goals.

    • Store unique responses (e.g., budget range, business size) as custom properties to tailor future interactions.

  2. Enable Custom Properties in AI Agent Settings

    • In the Settings tab, check the boxes for the custom properties you want the AI to reference.

    • Ensures the AI uses these details to personalize its replies.

  3. Collect Email & Phone for Multi-Channel Follow-ups

    • Surveys are a prime opportunity to gather contact details.

    • Share these details with your CRM or email marketing tool for cross-channel engagement.


C. Providing the Right Training Data

What Works Best

  • Product/Service FAQs: Covers common user questions, ensuring the AI can address them accurately.

  • Unique Selling Points (USPs): Highlights what sets you apart, allowing the AI to emphasize your brand’s strengths.

  • Brand Guidelines & Style: Ensures consistent tone and messaging across all interactions.

  • User Testimonials & Case Studies: Adds social proof and relevant examples for the AI to reference.

What to Avoid

  • Vague or Irrelevant Data: Unrelated snippets can confuse the AI and dilute your brand message.

  • Outdated Information: Ensure all data is current, especially if you frequently update products or policies.

Why It Matters

  • A well-structured dataset ensures the AI can confidently address user queries with accurate, brand-aligned responses. The fewer ambiguities you include, the smoother the AI’s performance.


D. Adding Relevant URLs

Including links to product pages, FAQs, or booking calendars allows the AI to direct users to the right resource at the right time. This is especially helpful for e-commerce queries or appointment scheduling

Best Practice:

  • Insert URLs in your training data alongside context.

  • For instance, “For detailed shipping info, visit [YourShippingLink].”

  • This helps the AI suggest or reference these links seamlessly during conversations.


E. Generating & Refining Training Data

Generate Training Data

  • Kickstart your AI setup by auto-creating data from your Instagram profile (bio, recent posts, etc.).

  • Great for testing and experimenting with a baseline of brand-relevant content.

Improve Training Data & Remove Formatting

  • Use these quick actions to optimize text for clarity and coherence.

  • This ensures both you and the AI can easily manage and interpret the data.


F. Testing the AI Agent Thoroughly

Multiple Angles of Testing

  • Check brand voice by asking casual questions vs. formal inquiries.

  • Probe for factual accuracy on product details, policies, or pricing.

  • Use edge-case queries (e.g., “Do you deliver to Antarctica?”) to see how the AI handles unexpected questions.

Chat as a Specific Contact

  • Simulate user details like name, location, or custom properties.

  • Ensures the AI references personal data accurately and maintains a personalized tone.

Concrete Example

  • Ask, “What is the return policy?” to confirm the AI’s ability to provide a correct link or info.

  • Then chat as a specific contact with a custom property (e.g., “VIP Customer”) to see if the AI tailors the response accordingly.


G. Setting a Unique Persona

Why Persona Matters

A distinct persona creates a memorable user experience and aligns with your brand identity.

Key Considerations

  • Define the AI’s role (e.g., friendly guide, expert advisor) and tone (formal, casual, playful)

  • If relevant, give the AI a backstory to make interactions feel more authentic and engaging.


H. Limiting (or Not Limiting) Response Data

When to Limit Answers to Training Data

  • Ideal for factual or support-focused AI Agents that must provide accurate info.

  • Example: Technical support or customer service scenarios where consistency is paramount.

When to Leave It Unrestricted

  • Suited for persona-based or gamified experiences that benefit from broader, creative responses.

  • Example: A fictional character or storytelling AI that can riff on user inputs more freely.


I. Coordinating with Human Intervention

Human Delay

  • Set at least a 1-minute delay to avoid clashes with other automations or live human support.

  • This ensures the AI only responds if no human or other automation has taken over the conversation.

Folder Organization

  • Group different human intervention scenarios into specific folders (e.g., “Escalations,” “Specialized Topics”)

  • Makes it easier to manage when the AI hands off a conversation and provides better context for you to manage a conversation.

Fallback & Unknown Answers

  • Configure a default response if the AI can’t handle a query or if the user explicitly asks for human help.

  • Helps maintain user trust by providing a clear path to human support.


J. Choosing the Right Scenario Actions

  1. AI Agent Auto-Reply

    • When to Use: Great for initiating or continuing conversations automatically.

    • Example: Sending a quick greeting after someone likes or comments on a post or providing a follow-up message after a user’s first DM.

  2. AI Agent Message

    • When to Use: This is best for targeted actions or segments of a conversation where the AI must perform a specific role.

    • Example: Mid-chat, instructing the AI to collect detailed user info (e.g., “Please ask the user about their budget range.”) or offering a curated recommendation.


Feedback


We’d love to hear how these best practices work for you! Please let us know if you have any questions, suggestions, or success stories about setting up your AI Agent. Your insights help us continue to enhance the Inrō experience for everyone.

Did this answer your question?