Why your ITSM agents are closing tickets fast—and teaching your AI nothing. | Value unlocked ITSM

[KelliA080304624]

Why your ITSM agents are closing tickets fast—and teaching your AI nothing

Hello value. How to turn your service desk's worst habits into your AI's biggest strengths

Welcome to the AI Center of Excellence (CoE) team at ServiceNow. I'm Kelli Abbott, a former data scientist now focused on AI strategy and helping customers unlock the full potential of their AI investments.

Here's the irony of my career shift: I spent years as a data scientist finding and organizing good data to build mathematical models. Now? We have incredible models—but organizations are starving them of the data they need to deliver value.

The models are ready. Your data practices need to catch up. Let me show you how in this deeper dive into ITSM. 

Your organization has invested in Now Assist and AI Agents. The technology works brilliantly. But adoption remains low because AI needs rich data to learn from, and most organizations are still capturing data like it's 2015—sparse descriptions, missing resolution notes, minimal context.

Here's a typical conversation I have: 'Kelli, we enabled Now Assist for ITSM six months ago. Our agents love the technology, but adoption is lower than expected.' I ask to see their incidents.

Incident INC0047392:

Short description: 'Printer broken'

Description: 'Fixed it'

Resolution notes: [empty]

Your AI is ready to learn. Your data just needs to catch up.

The Now Assist Lifecycle (When You Feed It Right)

ServiceNow's AI has this beautiful lifecycle that multiplies human expertise:

Step 1: Agent documents incident with detailed context (200+ characters)

Step 2: Incident Summarization generates useful summary for stakeholders

Step 3: Resolution Notes Generation captures the solution approach

Step 4: Knowledge Article Generation creates searchable, reusable content

Step 5: Future agents get Suggested Steps based on this collective wisdom

Each step amplifies the next. One well-documented incident doesn't just help that agent—it creates learning opportunities for every future agent who encounters similar issues.

Four Configurations That Unlock Value

1. Minimum Description Length

What: UI Policy requiring 200+ characters in incident description field

Where: System Policy > UI Policies > Create New

Why: Incident Summarization needs substantial context to generate useful content. Brief descriptions produce generic AI output that agents ignore.

Value: Organizations see 300-400% increase in usable AI-generated knowledge articles. Supervisors get better visibility without reading every incident detail.

Change Message: 'Provide detailed context to enable AI-assisted resolution for similar future incidents.' Show agents examples where their descriptions became helpful summaries and knowledge articles.

2. Mandatory Resolution Notes

What: Business Rule or UI Policy blocking incident closure without resolution notes

Where: Business Rules or UI Policy on incident form

Why: Resolution Notes Generation uses documented solutions as the foundation for Suggested Steps. Without resolution notes, future agents get no AI guidance on similar incidents.

Value: Resolution time for common issues drops 25-40%. Junior agents benefit from senior agents' documented expertise without constant interruptions.

Change Message: 'You're training AI to help your teammates solve this faster next time—document what worked!' Celebrate agents whose notes become AI suggestions that save colleagues 30+ minutes.

3. Structured Categorization

What: Mandatory Category, Subcategory, and Assignment Group before closure

Where: Data Policy or UI Policy on incident form

Why: Triage Cases and Classify Tasks AI Agents depend on consistent categorization to learn routing patterns and identify recurring problems.

Value: Incidents route correctly on first assignment, reducing frustrating reassignments. AI identifies recurring issues and recommends knowledge creation.

Change Message: Show agents how proper categorization helps them avoid irrelevant tickets. Frame as 'helping everyone work on what they're actually good at.'

4. Complete Problem-Solution Context

What: Require Description, Resolution notes, and Close notes fields

Where: Data Policy or UI Policy

Why: Complete problem-solution context enables Knowledge Article Generation. Partial data produces incomplete knowledge that agents don't trust.

Value: Duplicate incidents decrease 20-30% in first quarter. AI-generated knowledge becomes actually useful instead of generic filler.

Change Message: Position as 'building knowledge that helps your whole team resolve faster.' Highlight agents whose complete documentation eliminated recurring questions.

From Tracking to Training

One organization launched a 'Feed the Brain' campaign. Instead of mandating documentation, they celebrated examples where detailed notes helped other agents resolve incidents in under 10 minutes. They showed agents their documented solutions becoming AI-powered suggestions. Adoption went from 23% to 78% in six weeks—through showing value, not mandates.

Your One-Week Action Plan

Monday: Pull 50 recent incidents. Calculate average description length. That's your baseline.

Tuesday: Share the Now Assist lifecycle with your team. Ask: 'What if our documentation became AI help for everyone?'

Wednesday: Implement the 200-character minimum with positive messaging.

Thursday: Monitor impact. How many incidents meet the threshold?

Friday: Sample 10 new incidents. Compare to Monday. Celebrate improvement.

Post your before/after stats in the comments.

The AI-Human Partnership

AI handles pattern recognition, knowledge retrieval, and routine troubleshooting. Humans focus on complex problem-solving, creative solutions, and situations requiring empathy and judgment. You're not replacing expertise—you're amplifying it. Your experienced agents tackle challenging incidents while AI helps junior agents using the team's collective knowledge.

The Results

Organizations implementing these four configurations see knowledge article generation increase 300-500%, similar incident resolution time drop 25-40%, agent productivity improve, Now Assist adoption jump 40-60%, and first-time resolution rates improve.

Your ITSM AI is ready. Start feeding the brain today.

What's your average incident description length? Comment below. Let's compare notes and build momentum together.

Which workflow should we cover next? CSM? HRSD? ITOM? Your input shapes this series.

PS - Views expressed are my own and do not represent ServiceNow, my team, partners, or customers.