Making Zendesk AI a tool not a toy

Hi @viachaslau11,

Thanks for taking the time to provide us with feedback. Hopefully you have found it already, but our team is working hard to build and enhance the ticket summarization feature which is in EAP. We'll take your feedback into consideration for future reach of that tool. If you haven't seen it yet, we also are developing some functionality to help agents discover similar tickets which you can read about here. Perhaps there is some combination in functionality that could best assist your agents?

Regarding article summaries - we are starting work on some generative knowledge base functionality as well as have been making improvements to Help Center search with Machine Learning (aka semantic search). Is there a location where your agents interact the most that you'd like to see these efforts surfaced? Do they use the Knowledge app? Other tools?

Finally with reporting, we do have fields which are populated with Zendesk AI using our intelligent triage enrichments. These can be reported on in Explore as well as compared to the values set by your agents as discussed here. Is your primary blocker/feedback that you would be able to pull information like this in realtime? How would you leverage this data? To see trends, historical reference, etc.?

Thanks again for posting!

Thanks @jake18

Yes of course I follow all updates related to ZD AI. 

1. Thanks for mentioning this EAP Announcing the similar tickets EAP – Zendesk help, I'll check

2. Generative AI for Knowledge Base could be helpful (but not for my feature request). As for semantic search — great tool, at the moment partly useful for us (we use 3 languages in our HC - English, Polish (planned by ZD) and Ukrainian (not planned). My feature request is to combine intent (existing feature) and HC articles summarization. 

3. You're right, ZD has intelligent triage. What kind of reports can be created? For example, can I create a report with the five most common reasons for contacting support? But if this problem can be solved (since agents indicate the reason of customer contact), then what's going in real-time is the biggest headache in support teams. In practice, when a fire occurs, our shift managers quickly read unassigned chats in an attempt to understand what triggered the spike in requests.

@viachaslau11

I'm glad you're keeping up to date on what we're building. 😀

I'll make sure my colleagues see your point about combining summary and intent.

For reporting - yes! There are some pretty easy reports you can construct using Explore. I'll give the criteria for your example and perhaps you can go from there:

1. In Explore, click the reports (line graph) icon.
2. In the Reports library, click New report.
3. On the Select a Dataset page, click Support > Support - Tickets, then click Start report. The report builder opens.
4. In the Metrics panel, click Add.
5. From the list of metrics, choose Tickets > Tickets, then click Apply.
6. In the Rows panel, add the Intent custom field. (Click on this entry and Exclude the NULL values, so you only see tickets with intents)
7. On the right side toolbar, select the Result manipulation (up and down arrows) icon.
8. Select the Top/Bottom option
9. Check the box next to Top and then use the field to select how many results you would like to show.

You can of course filter this report for the timeframe you want to target and add other attributes or metrics, but this would get you started with visualizing the types of requests your users are submitting. There are several additional Explore recipes listed here which you may find useful. As for real time reporting improvements, feel free to break out your ideas for the Explore team as well here. The Machine Learning team will be looking at better ways to report on AI powered features, but real time improvements could possibly improve functionality for non-AI features as well.

Hope this helps!