• The easiest way to add sub-accounts to your agency is using your custom menu link in the Launchpad. Copy the link, and paste it under your agency settings in custom menu links. Then navigate to the account that you want to create and just click the custom menu link.

• Another way to add sub-accounts is through a direct connection in the Sub-accounts tab. Click on Create Account in the top right hand corner, then Direct Connection, and click on the sub-account you want to create in the GoHighLevel modal.

When connected Assistable.ai to GoHighLevel, you will get some actions installed in your sub-accounts by default. You can access these actions in a workflow by clicking on the 'Plus' icon in the workflow editor.

If you scroll to the bottom of the action menu, you will see some actions from Create Assistants. Just click on the Make AI Call action to add it to your workflow.

All you need is your assistant ID which you can get underneath your assistants name in the portal.

Of course! The only different is the way you connect Assistable.ai and GoHighLevel. All capabilities, actions, and use is the exact same other than the connection type.

Instead of using the Agency connection in the Launchpad, you will have to use the Direct Connection in the Sub-account tab. This is because GHL doesn't allow agency-level connections on the $97/mo plan of GHL. Instead of clicking on the agency (which will result in an error), click on your sub-accounts individually to connect / import them.

So by using the direct connection, you can connect to your sub-accounts but you have to do them individually.

The easiest answer is to tell the assistant to avoid doing 'X'. Most of these issues come down to prompting and instruction formatting with the assistant.

LLMs (like OpenAI or Claude) are programmatically built to follow instructions. So most verbiage, formatting, and language can be fixed / mitigated via instructions.

Quick Tip: AI models are like humans in one way - the Pink Elephant Paradox. If I tell you to not think about a pink elephant, whats the first thing you do? Think about a pink elephant... AI models are no different. Avoid telling the AI NOT to do something and rather tell the AI to do what you want it to do.

The number 1 reason for booking troubles tend to come from user prompts. Our AI models are pre-configured and prompt-injected to follow proper booking protocol. Most of the time, users tend to prompt all scenarios for booking which may cause confusion in the model's injection. Simply your prompt to just state:

"Get your availability and book an appointment with the user by conversing your available appointment start times, and then booking the time if it is available and preferred by the user."

There are 3 systematic reasons your AI is incapable of booking appointments:

• Access Token / Oauth Connection: If your access token or oauth connection has failed or doesn't exist, we systematically would have the access to book appointments to your calendar. A simple refresh of your integration will fix this. Symptoms include: AI is able to pull calendar availability correctly, but unable to book. There is a red banner in the top of your sub-account portal reading 'No Active CRM Connection'. Conversations in the Inbox tab are showing IDs instead of names.

• No Calendar Management Tools Added: Even if your calendar is connected, we still need to give the AI the ability to manage it. This can be added in the Tools & Abilities section when configuring the AI. To properly book an appointment we need two tools specifically: Book Appointment and Check Availability. With those two tools, you have all the tooling you need to book. Symptoms include: You have a calendar connect it but the AI is not reading calendar times properly or able to book an appointment. The conversation around booking sounds like AI rather than real data.

• No Calendar Connected / Added To The Assistant: Even with a great booking prompt and proper tools - it is still not possible to book without a calendar. Simply navigate to the Calendars section in the assistant configuration and add a calendar using the modal.

The main things to check for an outbound call not going through:

1. Your access token / integration with GoHighLevel & Assistable.ai is current. Without this information, we are unable to check information relating to the call. Reset your oauth connection.

2. The contact doesn't have a phone number / valid phone number. This would cause the telephony to reject the call because there is either no number to call, or the number is invalid.

3. Your assistant doesn't have a number assigned phone number. Simply add a platform or imported number to the assistant and re-try the call.

4. Rate limiting. If you are requesting a large amount of calls at once we will reject your requests on a roll 60 second basis. The default limit start at 15 calls-per-minute and scales to 60 calls-per-minute as long as the account's reputation is good. This limit can be requested to be increased based on your needs - however, this does not mitigate number reputation as a concern.

This usually comes up in 3 scenarios:

1. You are using Labs. Labs do not have access to tools, knowledge bases, or additional configurations - they are purely meant to test conversation ability of the assistant before deploying to a live channel to test main functionality.

2. You are using Inbound / Outbound / Web Voice. As of writing, voice doesn't have access, or utilize the knowledge base system to save on latency that comes with accessing knowledge. This is on the roadmap to add in the future.

3. Your knowledge base isn't refined or has bad information. Knowledge works by taking what the user (contact) said / asked and running it against your knowledge base to find a potential answer. It's given a match score - if the match score is too low, we filter those responses out. This can be solved by using the knowledge lab, or adding in more specific information.