I have an IT Help Desk agent. I've shared it with an #help-it Slack channel. I've also added Glean Search to the channel. This agent has been instructed to understand the context of the question and the persona of the person asking the question and it's not a simple ask a question and get an answer response.
I go to my channel and I ask a question, "How do I print?" Nothing happens. I thought it lurked but maybe not.
I use the Glean search and ask again. This time I get the :eyes: emoji and a response in thread. Great!
But this is a conversation bot and it's asking me to confirm a step. So I confirm it and nothing happens.
Let's try using Glean search and confirming. Ok, that did something but instead of continuing the conversation, it broke the fourth wall and mused what it should be doing (which it definitely has been instructed NOT to do) before it answered.
My question is - what's the expected behavior here? If I truly want to publish a functional Glean agent to a Slack channel am I going to get a first rate experience sharing my agent with a channel or should I use webhooks and the API to get the experience I want? Having to prepend Glean search to every conversation inserts a lot of friction for a channel meant to be driven by my agent. I swear this used to work differently before agents but I don't see any settings I'm missing in Glean.
What I'd like to see - Have Glean lurk in a channel without being invoked. I've built several Slack bots that receive every message in a channel, review the message, and reply if appropriate. I've primarily used AWS Gateway, Lambda, and Bedrock. It works well but I have to manage a Slack app and scaffold and write code and it's several hours worth of work.
