There is a lot we try to keep out of your hands so you have to do as little work as possible. However, we don't have control over a couple of things:
Agent Suite calendars and calendar configurations - Your AI will pull true-to-time availability from your Agent Suite calendar via API v2 (oauth) & Agenet Suite's very own availability endpoint on scheduling pages. Meaning we get the most accurate times from Agent Suite based on your current settings. If you find a discrepancy, please check your calendar settings in your sub-account as well have no control over configurations at the calendar level. Just getting times and booking based on the calendar's unique identifier.
GoHighLevel API v2 (oauth) initial connection - If done correctly, this only ever has to be done once over the life of your account. We were early users to API v2 so we've mastered the logistics of the system - however, one thing we cannot control is what you click via initial connection. Improper configuration can lead to have to reset it in the future for new sub-accounts and loss of connection.
Response issues and LLM error handling - We have 99.9% of LLM and AI vendor error handling in place to make sure there is little downtime as possible. However, when it comes to building AI & prompting - there is a lot of variables that are impossible to account for until we come across them. For example - improper tool calling hard-prompted. As we come across unique break-cases and bugs, we fix them but we will share some common things we have ran across.
"Off the rails" or not following instruction - Unfortunately, we dont control the correlation of response-to-query percentages based off of instruction (if so, we wouldn't have to work). So, if you ever see your AI model go off the rails or not properly follow instructions that can be linked back to prompting. We generate torture tested prompts for you on assistants creation - With deviation from these instructions, we cannot control how you instruct your AI to do things (whether you do it well or not). But we will share some best practices with you.
