Using Agentic AI for Quality Monitoring on outbound emails within cases

[FaizanKhan]

Hi all,

I’m currently working on a requirement to introduce agentic AI-driven quality monitoring for outbound emails sent from cases in ServiceNow, and I’d appreciate some guidance on the best approach.

We want the system to automatically scan outbound emails  and identify 

• Spelling and grammar mistakes
• Emails that are logically confusing or unclear

What we want to achieve

• Automatically evaluate outbound emails when they are sent from sys_email
• Flag low-quality emails for review
• Create a task/record containing:
  • Email content
  • Identified issues
  • User who sent the email

-----

What I've tried so far is using AI Agent studio to create a Agentic Workflow on the sys_email table however that is not working. 

Iam not sure how best to approach this entire problem, anyone who has experience in AI agents in servicenow and using AI agent studio do you have any tips and the best way to implement this problem is thank you.

[NivethaJ]

Hi faizan,

Did u find any solution?

[Vamsi_Krishna07]

Hello @NivethaJ

For this use case, I would recommend using a Now Assist Skill rather than an Agent, since an Agentic Workflow/Agent could be more expensive, with a minimum consumption of around 25 Assist credits per transaction. You can create a Now Assist Skill that takes the email body from "sys_email" as a dynamic input, evaluates it against the quality guidelines provided in the prompt, and returns the identified issues or an appropriate evaluation result. You can then trigger the Skill from a Flow Designer flow when the outbound email is sent, evaluate the Skill output, and if the email does not meet the defined guidelines, automatically create a task/record containing the email content, identified issues, and the user who sent the email. This approach should be simpler and more cost-effective for a straightforward evaluation use case where you don't need autonomous agentic reasoning or multiple tool interactions.

Thanks!

[Travis Rogers]

Moreover, I'm not even sure monitoring the emails themselves is the best path here. If the goal is to clean up user-facing content, they really could just be analyzing the Comments stream on relevant records. This will address the content at the source, rather than introducing the complexity of parsing out thousands of "a new ticket in your queue" type of emails. 

I have a customer who did just this by having skills run at the closure of each Incident. The skill would look at various elements (process adherence, professionalism, friendly tone, etc) to score the activity of that record, and raise outliers up for a human-led quality review. 

[rpriyadarshy]

@FaizanKhan  from solutioning perspective i see 2 points here ..

(1) Proactively send good QUALITY email for cases using - Service-Now Email reply recommendation for CSM Module. This Now Assist skill automatically drafts  contextual email response for Case Agents.

(2) If still you need to analyze All Quality Parameters for your Organization so technical solution is given in below response by VAMSI

My Point here is to FIRST exploit OOTB for which we bought the paid product.

Regards

RP