Conversation AI

When a user replies directly within the native Instagram or Facebook Messenger app, the Conversational AI should automatically pause for that conversation. This prevents the AI and the human from replying at the same time and allows the conversation to transition naturally to the business owner. The AI could remain paused until either: the conversation is manually re-enabled, or a configurable inactivity period has elapsed (e.g. 24 hours). This would create a much smoother handover between AI and human and is particularly important for businesses using AI for enquiry handling rather than full automation. Why this matters: Prevents duplicate/conflicting replies. Gives the business owner complete control once they step in. Creates a seamless AI → human handover. Particularly valuable for Instagram and Facebook DM workflows.

In Conversation AI, add the option to use different models (not just open.ai ), just like in voice agents. Anthropic particularly.

Currently, the conversational ai bot would go to sleep (if we activate the option) if we send a message or through a workflow. The problem comes when we send the message through the whatsapp business app: clients that use coexistence can talk through the whatsapp app, but those messages are not taken into account, and the bot don't go to sleep (which leads to duplicated messages and a lot of chaos). POTENTIAL SOLUTION: take into account inbound messages sent through the whatsapp business app so that the bot also goes to sleep when we send them.

Conversation AI can save information into a custom field, but it cannot update or overwrite a value once one has already been saved. This creates two compounding issues. Issue A: No Overwrite Function Exists There is no mechanism within Conversation AI to update a custom field once a value has been written. Even if the bot is explicitly instructed to update a field based on new information provided by the contact, the capability does not exist at the platform level. Issue B: Saved Values Are Permanent Regardless Even setting aside the missing function, the underlying behaviour treats the first saved value as final. If a contact provides updated information mid-conversation, the bot cannot reflect that change in the CRM. The outdated value remains locked in. Example: A contact places an order inquiry and tells the bot they want Product A. The bot saves Product A into the custom field interested_product. Mid-conversation, the contact changes their mind and says they actually want Product B instead. The bot acknowledges the change, but the interested_product field still shows Product A in the CRM. Any follow-up automation, order confirmations, product-specific sequences, and sales team notifications will fire against the wrong product. The CRM is wrong. The automation is wrong. And there is no way to fix it from within the bot.

Voice AI already lets us call an external API, pass structured customer data, receive a response, and continue the conversation. Conversational AI needs the same capability. The ideal solution is custom remote MCP support. At minimum, please add a Custom Action 2.0 equivalent to Voice AI Custom Actions for SMS, Web Chat, WhatsApp, Facebook, and Instagram. Businesses need to be able to: Trigger an action only when the customer’s intent requires it Collect structured information across multiple messages Send authenticated API requests with dynamic contact and conversation data Preserve session state throughout the conversation Map API responses back into the AI’s next reply Continue the same natural conversation after the action completes Prevent duplicate or accidental actions with session and idempotency controls Our real-world use case is booking appointments through a custom booking platform. Voice AI already does this successfully. Conversational AI currently requires fragile workflows, temporary fields, bot pausing, loops, and dozens of branches—and still cannot provide a reliable continuous experience. This would unlock far more than bookings: quotes, orders, inventory, account support, ticketing, payment status, and other custom business systems. Please give Conversational AI full external-action parity with Voice AI—through custom MCP servers, Custom Action 2.0, or ideally both. If your business would benefit from this, please upvote and comment with your use case so HighLevel can see the demand.

Please add support for Custom Actions in Conversation AI, allowing AI agents to connect to and interact with external APIs. This would enable developers to fetch data, trigger workflows, send requests, and integrate third-party services directly from Conversation AI, significantly expanding automation and use cases.

I would like a final message to go from the Conversational AI bot to the contact when the maximum number of replies is reached. The AI assistant should not be having a two-way dialog with someone and then suddenly go silent. In the Conversation AI Advanced Settings for Auto-pilot mode there should be at least 2 fields for what happens when the max number of messages is reached: 1 - What do you want the bot to say before it goes to sleep? (Example answer: "Thank you for providing this information. It will help us determine how to best serve you. I will hand this conversation over to a human who will get back to you shortly.") 2 - What happens next? Possible options: Trigger a workflow. Notify the assigned user. Nothing (leave conversation marked unread). Mark the conversation as read.

In AI Agents, within Bot Goals, the top horizontal menu takes up an excessive amount of vertical space, leaving the main working area very small on laptops. This makes sections like intent behind conversation and personality harder to review and edit. On large screens it’s less noticeable, but on laptops it feels like an unnecessary layout issue rather than a preference. Reducing the top menu height or making it collapsible would greatly improve usability.