Skip to main content

General Settings of your AI Agent

This is where you fine-tune how your bot acts, talks, and thinks. Navigate to AI Agents > Configure > Settings tab.

Updated yesterday

Things you can configure:

Agent Name

Whatever you name it shows up in chats. Keep it simple:

  • "Support Agent"

  • "Sales Assistant"

  • Your brand name + "AI"

Avoid weird names like "Bot3000" unless thats your vibe.

AI Model Selection

Pick your bot's brain. Each model has different strengths and costs.

Fast and Cheap Options (1 credit per request)

  • Gemini 2.5 Flash - Fast, cheap, good for simple queries

  • GPT-4o - Solid all-rounder, handles most tasks

Marginally Slower But Smarter Options (1 credits per request)

  • Gemini 2.5 Pro - Smarter, better with complex questions (takes time to think)

  • GPT-o3 mini - Top tier, employs reasoning for complex queries

Start with Gemini 2.5 Flash or GPT-4o.

Upgrade if customers need more sophisticated answers. Monitor your credit usage in analytics.

System Prompt (Your Bot's Personality)

Full guide on that here.

Reply Buttons

Those thumbs up/down buttons after bot responses.

What they do:

  • Collect feedback on bot performance

  • Show you what's working

  • Highlight problem areas

Setup:

  • Toggle on to enable

  • Shows "Did this help?"

  • Customers get to tell whether it did or not

Keep them on initially to gauge performance. Turn off once the bot's solid.

AI Rate Limiting

Stops one person from burning all your credits.

Why enable:

  • Prevents abuse

  • Controls costs

  • Maintains fair usage

Settings covered in detail here.

Quick version: Set max requests per user per hour. Start with 20, adjust based on usage.

Advanced Settings Tips

For Ecommerce:

  • Add product recommendation guidelines

  • Include size/fit help in prompts

  • Set up abandonded cart responses

  • Train on return policy heavily

For Service Businesses:

  • Focus prompts on appointment booking

  • Add location/hours prominently

  • Include service descriptions

  • Set escalation for quotes

For B2B:

  • Make bot more professional in tone

  • Add technical knowledge to data sources

  • Quick escalation to sales team

  • Include pricing tiers if public

Testing Your Settings

After any settings change:

  1. Test in playground first (right side panel in Data Sources)

  2. Try edge cases:

    • Angry customer simulation

    • Complex product question

    • Refund request

    • Multiple questions at once

  3. Check constraints work:

    • Ask for a discount

    • Request something bot shouldn't do

    • Try to break its rules

  4. Verify tone matches brand:

    • Should sound like your company

    • Consistent across responses

    • Natural, not robotic

Common Settings Mistakes

Bot too salesy?

  • Dial back sales language in prompt

  • Add "be helpful not pushy" constraint

  • Focus guidelines on assistance first

Bot giving wrong info?

  • Check data sources first

  • Then verify prompt isn't contradicting training

  • Model might be hallucinating - if you're using a fast model, upgrade to a thinking model (Gemini 2.5 Pro or GPT o3)

Bot too expensive?

  • Switch to a cheaper model

  • Enable rate limiting

  • Check if complex prompt is causing multiple reads

  • Simplify constraints

Bot not following rules?

  • Constraints must be numbered and clear

  • Check for conflicting instructions

  • Some models follow instructions better (GPT-4.1 is the best at this)

Remember: Settings is where you make the bot YOURS. Default templates work fine, but custom prompts make it shine.

Spend time here - it pays off when bot handles queries exactly how you want.

Did this answer your question?