We are configuring SmartAssist for our Live Agent Transfer project and have imported our UAT bot into SmartAssist and published it.
We were just able to get our published Sharepoint UAT site connecting to SmartAssist and I did some playing around with it tonight. I am concerned with the behavior I’m seeing in SmartAssist and am not sure if there is a configuration we can apply to fix the problems I’m seeing, or if we need to submit a bunch of product fix and enhancement requests.
The two biggest issues I found tonight are:
When connecting to SmartAssist Via a published channel, the NLP is not coming through into the NLP reports. The NLP utterances are only coming through in the NLP reports when I’m testing directly in botbuilder and not the published channel. As soon as I talk to the published channel, no utterances are coming into the NLP analysis.
The conversation history/logs in SmartAssist are not accurate. They are capturing every time someone utters anything into the published bot, even when the bot alone answers it. Using SmartAssist we expect and want only actual Agent to Employee conversations to be captured in the conversation logs. Further, the conversations don’t start from the start time of the conversation - they pull info from before the user connects with a live agent. Also - when I try to download the conversation history by clicking the download conversation option it gives me NOT just that once conversation but multiple days worth of conversations. It should only give the conversation that happened between the time the person connected to the agent and the time they disconnected.
We want the bot task NLP to continue to come through the botbuilder NLP reports when someone did not connect to a live agent. The conversation logs in SmartAssist should only contain the conversation history from the time the person connected with a live agent until the time the conversation ended.
We don’t want to capture conversation history in SmartAssist at all - unless an agent has connected with the person chatting. We basically want all of our bot tasks to run as normal when users trigger them and not go to SmartAssist at all, unless we ask it to transer to a live agent via our custom dialog task we called “Connect to a Live Agent”. This will enable us to clearly distinguish between when someone was able to successfully use the bot tasks we’ve developed, vs when someone had to deflect to a live agent and have a live chat. Our metrics will be much cleaner.
Please help if you have any insight. I will definitely be contacting product support but wanted to post here first as it’s a weekend.