NewNew: The enterprise guide to Agentic AI — 24 min read.

Read →
Case study · Enterprise IT · AI for IT

58% ticket deflection on the IT service desk at a global manufacturer

A global manufacturer's IT service desk was buried under repetitive password, access and how-do-I tickets across 34 countries. pronix.ai launched an agentic AI service desk on Kore.ai integrated with ServiceNow and Entra ID — 58% of tickets resolved without a human, MTTR on Tier-1 issues down from 42 minutes to under 4, and a 24/7 multilingual experience inside Microsoft Teams.

Client
Global industrial manufacturer, 84,000 employees
Industry
Enterprise
Platform
Kore.ai Agent Platform · ServiceNow · Microsoft Entra ID · Microsoft Teams
By pronix.ai Strategy Practice8 min readQ1 2026
58%
Ticket deflection (Tier-1)
42m → 4m
MTTR on password / access
$6.8M
Annualized service-desk savings
24/7
Multilingual coverage across 34 countries

*Representative outcome; results vary by client, scope and platform configuration.

The challenge

The service desk handled 1.2M tickets/year — 61% were password resets, MFA lockouts, access requests, VPN issues and knowledge lookups. Average speed-to-answer was 42 minutes on Tier-1 and 6 hours on Tier-2, with a $28/ticket fully loaded cost. Prior chatbot attempts topped out at 12% deflection because they couldn't actually complete transactions in ServiceNow or Entra.

Our approach

Step 01

Bounded agent per intent, not a mega-bot

Six intent-bounded agents — password reset, MFA unlock, access request, VPN triage, software install, knowledge lookup — each with a clean escalation path. That scoping is what took deflection from 12% to 58%.

Step 02

Real tool actions on ServiceNow and Entra

Idempotent, audited tool calls that actually reset passwords, unlocked MFA, provisioned SailPoint access requests, opened and updated ServiceNow tickets — not just deflection screens telling the user to file a ticket themselves.

Step 03

Teams-native conversational surface

Deployed inside Microsoft Teams with SSO, so employees never left their flow. Voice and email channels routed through the same agent layer with the same guardrails.

Step 04

RAG over the IT knowledge base

Grounded answers on the ServiceNow KB and SharePoint runbooks with citation, freshness scoring and a 'no confident answer → route to human' rule. Zero hallucinated procedures in production.

Step 05

HITL for anything privileged

Access grants above a risk threshold, production-system changes and anything touching finance systems required a human approver in ServiceNow — the agent prepared the request, the human approved.

Step 06

Evaluation and drift monitoring

A golden set of 900 historical tickets ran on every prompt or model change, plus weekly production sampling by the service-desk lead. Anything degrading > 3% blocked promotion.

We stopped measuring the bot's containment rate and started measuring how many employees never had to open a ticket at all. That is the number that moved.

Head of Global IT Operations
For CIOFor Head of IT OperationsFor VP Digital WorkplaceFor Head of Employee Experience

Illustrative case study. Scenarios, metrics, quotes and client details are representative composites based on Pronix engagements and industry benchmarks unless a named client is shown with written consent. Outcomes vary by client, scope, data quality and platform configuration. Nothing on this page is a guarantee, warranty or professional advice. See our Terms of Use for the full disclaimer.

Trademarks.

AWS, Salesforce, Microsoft, NICE, Genesys, Kore.ai, Google, ServiceNow, Amazon Connect and logos referenced on this site are trademarks of their respective owners. References are for descriptive purposes only and do not imply endorsement, sponsorship or partnership beyond stated partner relationships. See our Disclosures.

Bring this to your program

Talk to the team that shipped it.

Book a working session with a pronix.ai lead on this practice — we'll map the approach to your platform, industry and constraints.