#2648-Keyword Not Working with AI Agent for All Queries.

Ricardo Szemat
OPEN
#1
Ricardo Szemat

 Ricardo Szemat

I am stuck in a configuration loop where I can only achieve one or the other, resulting in a broken user experience.

Scenario A: Intent Detection = ON
  • The Good: High-intent keywords successfully trigger the visual Appointment Flow.
  • The Bad: The fallback routing (Default Reply Node) loses all native contextual memory. Every open-ended question is treated as a standalone event, forcing the AI to repeat itself.
Conversation Flow Example:
  • User: Good morning. What is the price of the consultation?
  • AI: To give you pricing, first I need to know which branch you want to visit.
  • User: Ohio.
  • AI: The consultation in Ohio is $100. What is a good day for your visit?
  • User: I would like to know the price of [Service].
  • AI: To give you pricing, first I need to know which branch you want to visit. (Memory Loss: Forgot user already said "Ohio")
Scenario B: Intent Detection = OFF (AI Agent for All Queries = ON)
  • The Good: Seamless, flawless contextual memory. The AI naturally holds the conversation and retains details provided by the user.
  • The Bad: Global keyword triggers are completely bypassed. When a user states they want to book, BotSailor fails to launch the interactive visual flow, leaving the user stuck in an endless text loop.
Conversation Flow Example:
  • User: Good morning. What is the price of the consultation?
  • AI: Good morning, to give you pricing, first I need to know which branch you want to visit.
  • User: Ohio.
  • AI: The consultation in Ohio is $100. What is a good day for your visit?
  • User: I would like to know the price of [Service].
  • AI: [Service] in Ohio is $300. What day works for you? (Perfect Memory: Remembers "Ohio")
  • User: Monday 4th.
  • AI: Thank you for your cooperation, please provide your details... (Failure: Fails to trigger the visual Appointment Flow to map the calendar automation)
The Question

How can I configure BotSailor so that the AI Agent manages the open-ended dialogue natively (with memory), but still allows the platform's keyword/intent engine to intercept the conversation and launch a specific visual flow?

Is there a specific webhook, session variables ({{subscriber_id}}), or a Function Calling/Tools feature I should be using to bridge this gap?

#2
Kamrul Hasan

 Kamrul Hasan

Dear Ricardo Szemat,


Thank you very much for explaining your issue so clearly. I now fully understand your situation, and I sincerely apologize for the inconvenience this may have caused.

Regarding the AI agent's Intent Detection being ON or OFF, the Contextual Memory feature should handle this scenario. However, you didn't mention whether you have enabled it.


In your AI Agent settings, there is an option called "Enable Contextual Memory." Please make sure it is enabled and specify how many previous conversations the AI should consider when generating its next response.


If you haven't enabled this feature yet, please try it and see if it resolves the issue.

However, if Contextual Memory is already enabled and you're still experiencing the same problem, please open a support ticket and include your account credentials. Our technical team will review your AI campaign and investigate the root cause of the issue.


You can open a support ticket using the link below:

@botsailor.com.../login

Thank you for being with BotSailor.


Best regards,

Team BotSailor

Related Threads