I am seriously losing my patience with the Agent Assist feature. Our IT department turned on the AI knowledge bot last month and it is driving my thirty agents crazy. The bot constantly suggests completely irrelevant knowledge base articles to the agents while they are talking to customers. It is distracting and ruins their handle times. Worse, there is no button for my agents to just tell the bot that the article is a bad suggestion. Why would you build an AI tool without a simple thumbs-down button for the end user to correct it? How do I stop this garbage from popping up on their screens without completely disabling the feature?
Hey there! I completely understand why your agents are frustrated, but please do not give up on Agent Assist just yet! It is an incredibly powerful tool once it learns your data! You are totally right that there is no native ‘thumbs down’ button in the agent workspace right now, but you can build a brilliant automated solution! Using the Platform API, I created a custom Script with a simple ‘Bad Suggestion’ button for the agents. When they click it, it fires a Data Action that logs the interaction ID and the exact knowledge query into a spreadsheet so our content managers can retrain the AI models every single day! It works beautifully!
Yeah, the API workaround is cool, but if you do not have developers to build that, you can leverage the native Quality Management tools. What we did for our agents was create a very quick gamification program. We just told them to flag the interaction with a specific wrap-up code like ‘AI Misfire’ whenever the bot hallucinates.
Then our quality team reviews those specific calls and updates the knowledge base keywords directly. We even reward the agents with gamification points for finding bad AI suggestions.
It turns a frustrating experience into a fun bug-hunt for the floor.