AI Agents in Action: Small Wins That Make Big Impact — What Will You Build?

[Nick Derbawka]

I was on a call with a customer who told me that, in some months, their HR team receives over 100,000 cases by email, funneled through more than 200 shared inboxes. The volume alone sounds overwhelming, but the real problem is how those cases are created. The Inbound Action rule they have in place takes the email subject line and uses it as the case’s Short Description.

Think about subject lines like “Re: Fwd: Payroll question from last week” or “URGENT!!!”. Not very helpful. The result is that every case enters the system with a label that doesn’t actually describe the issue. At this customer, their Tier 1 HR team has to waste countless hours every week clicking into each one just to figure out what it’s about, slowing response time and adding unnecessary frustration.

That got me thinking: how could AI Agents help here?

So I built a simple one to test the idea. Instead of relying on the subject line, the AI Agent I created reviews the email content that gets stored in the case Description, summarizes the intent into a single line of 100 characters or less, and populates the Short Description with something clear and actionable.

Now, instead of nonsensical subject lines in the Short Descriptions, the case list looks like this:

• “Overtime pay missing from paycheck”

• “Request to update benefits after marriage”

• “Question about incorrect tax withholdings”

It’s a small change, but the impact is immediate. Cases are easier to scan, easier to triage, and faster to resolve. Multiply that by thousands of cases a month, and you start to see real-time savings.

The bigger point is that AI Agents don’t have to be huge, sweeping projects. I built this in under an hour. You don’t need to start with an end-to-end transformation. Look for the small, tactical wins where people are re-typing, re-summarizing, or cleaning up messy data. Those are the perfect entry points to build trust and prove value.

That’s my example. Now I’d love to hear yours. Where in your processes do you see small inefficiencies an AI Agent could fix? Share your ideas, and I’ll pick a few to build out and post as demo videos for everyone to see.

[Thomas Bohan]

Practical real world examples like this are great, thanks Nick

[James Van Sickl]

FYI, this is a very wasteful design from licensing consumption aspect. Each use of an AI Agent to rebuild the ticket's short description would cost 25 assists. In an org that gets 100k emails, that would be 2.5 million assists just rebuild short description.

A better design would be to use Now Assist Skill Kit to create a skill that rebuilds short description based off email subject and body, then update the email's inbound action to call that skill via script and returning the new short description. Skills cost 1 assist. So, assist usage for the scenario is reduced from 2.5 million to 100,000.

[Nick Derbawka]

Thanks for your feedback @James Van Sickl -- this was something I built pretty quickly to accommodate a hypothetical use case. Of course, with ServiceNow, there are many ways to get to the same result. Considering licensing is always something I recommend as I work with my customers to build out "production-grade" use cases.