Option to enable automatic AI suggestions within Ticket Automations
I would like to be able to enable AI suggested replies and solutions for all tickets received (perhaps within ticket automation rules). Ideally, it would look something like this
Ticket comes in > AI suggest both a reply and a solution > user can deploy the suggested reply with one click, or override it and write their own reply
They can manually attempt the solution if it seems viable, or disregard this.
If it could do this with each subsequent reply in the conversation that would be fab
the Ai assistance is actually incredibly helpful for us and generates good replies most of the time, but it's a manual process to use it on each ticket currently. Just smoothing out the workflow a little would be very valuable. However, this would have to be an optional feature as I think many users would not like it. Being able to enable it within ticket automation would be even better as then we could assign it to specific customers or contacts tickets only
D Chamberlain
shared this idea
-
D Chamberlain
commented
Hi Stephen,
You have misread my suggestion.
"Ticket comes in > AI suggest both a reply and a solution > user can deploy the suggested reply with one click, or override it and write their own reply
They can manually attempt the solution if it seems viable, or disregard this."
Most of the value in the AI thus far is in suggesting replies, not solutions. This is what we would use it for, and we have tested it extensively in this regard.
That being said, the suggested solutions are quite useful as a jumping off point, which is why we'd leave those on. But as mentioned in my suggestion, we would only ever action these manually, and only if they were viable.
If you train your users well in ticket hygiene, then it translates quite well into an AI-readable prompt. If they continue to struggle with this, you can create an intermediary form in MS Forms that does a little bit of ticket shaping that can improve things (by making them specify particular apps, hardware, types of issue etc)
-
Stephen Schillinger
commented
Neat idea, but I have 3 questions.
1. Do you trust the AI to suggest a working solution that wont break something?
2. Do you trust the user to be descriptive enough for the AI to get a good idea of the problem?
3. Do you trust the user enough to approve the execution of said AI script?I don't remotely trust enough on all 3 counts.
It'll get there some day, but it is a ways off still.