Torrey Leonard
When will mid-call action be released?
I can’t deploy the AI agent without SMS support or this feature. Right now, I’ve got neither.
• SMS Issue: I can’t buy an AU mobile number, despite repeated attempts to get support. If I need to purchase one from Twilio and port it to Thoughly, I’m fine with that. But I need clarity:
- a.
Can Thoughly accept an AU mobile number?
- b.
What’s the cost once the number is ported? I want to avoid double charging by Twillio & Thoughtly.
This isn’t just my problem—many AU users face the same issue.
Agent Context Awareness
I’m also trying to understand why the AI agent:
- a.
Can’t retain context from previous conversations.
- b.
Repeats conversations unnecessarily.
- c.
Struggles with escalation handling.
I’ve posted these issues in the support channels, but no one is responding. Here’s the full post and the questions I’ve raised and tagged Will Del Principe:
I have rebuild the agent 2 more times after our chat. And, now facing issues with contextual awareness, escalation handling and repetition questions.
- a.
I have noticed that when users provide answers to specific questions early in the conversation, the agent sometimes asks redundant or related follow-up questions instead of progressing contextually. Is there a way to configure Thoughtly to better recognise and retain contextual information from earlier parts of the conversation to avoid this? I have added rules & custom prompts to avoid unnecessary repetitions but it does not work.
- b.
Thoughtly allows for custom prompts, additional rules, and conversation scripts. Can you clarify the best use cases for each? Specifically, where should I focus on injecting these elements to make the conversation flow more natural and context-aware?
- c.
I’d like to implement escalation to a human agent at any point in the conversation where the user’s query falls outside the AI’s capabilities. Given Thoughtly’s process-driven approach and structured nodes, how can escalation be managed effectively to allow it to occur dynamically, regardless of the conversation step?
- d.
Are there specific settings or features within Thoughtly that can help improve the agent’s adaptability to handle “off-script” responses without sounding robotic or redirecting unnecessarily to predefined questions?
- e.
For conversation design, what are the best practices to ensure the AI agent can handle layered intents (e.g., when a user provides multiple pieces of information in one response) without breaking flow or asking redundant clarifications?