Integrations · Azure AI Foundry

Azure AI Foundry + ServiceNow: AI agents that resolve tickets before humans touch them

Azure OpenAI connected to ServiceNow via REST API enables AI agents that triage incidents, suggest resolutions from the knowledge base, and auto-close repeat issues — without modifying your ServiceNow instance or installing a plugin.

74%
Tickets auto-resolved
4.2 min
Avg resolution time (was 2.1 hrs)
91%
First-contact resolution
60%
Analyst time saved

What's Possible

Six ITSM capabilities AI agents can handle autonomously.

Incident Triage & Routing AI

Incoming ServiceNow incidents are classified by the AI agent — category, subcategory, priority, and affected CI — before a human analyst sees the ticket. Routing rules assign to the correct team instantly, eliminating manual triage queues.

Self-Service Resolution Agent

Azure AI Search indexes the ServiceNow knowledge base (KB articles, resolved tickets, runbooks). When a new incident arrives, the agent retrieves the most relevant resolution steps and attempts autonomous resolution for known issue patterns.

Change Advisory Board Automation

Change requests are assessed by the AI agent against change history, affected CI blast radius, and current change freeze windows. Low-risk standard changes are auto-approved; complex changes receive an AI risk summary for the CAB.

Problem Management Intelligence

The agent correlates recurring incidents by symptom patterns, affected CIs, and error codes — identifying candidate problems automatically. Problem records are created with an AI-generated root cause hypothesis and linked incident list.

SLA Breach Prediction

Azure Monitor tracks ticket age, assignment group queue depth, and historical resolution times. The agent proactively re-prioritises or escalates tickets predicted to breach SLA before the deadline — surfaced as Teams alerts to service managers.

Service Catalog AI Assistant

Employees request IT services through a conversational Teams bot rather than navigating the ServiceNow catalog. The agent identifies the correct catalog item, pre-fills request fields from employee context, and submits to ServiceNow via REST API.

How We Connect It

Azure AI to ServiceNow — no instance modifications required.

01

Architecture & Auth

Azure AI agents connect to ServiceNow via the Table API and Scripted REST APIs — no ServiceNow plugin installation required. OAuth 2.0 with a ServiceNow OAuth provider profile handles authentication; Azure API Management adds rate limiting and request normalisation.

  • ServiceNow Table API endpoints registered in Azure API Management with OpenAPI schema for Semantic Kernel plugin generation
  • OAuth 2.0 client credentials flow to ServiceNow OAuth provider — token refresh handled by API Management
  • ServiceNow integration user with scoped role assignments — read/write limited to incident, problem, change, and knowledge tables
02

Data Pipeline

Azure AI Search indexes the ServiceNow knowledge base via HTTP push connector, receiving new and updated KB articles via ServiceNow Business Rules. Resolved incident text is periodically indexed to build a resolution pattern corpus.

  • Azure AI Search HTTP push indexer receiving ServiceNow KB article payloads via Business Rule webhook
  • Vector embeddings on KB article content + resolution notes — hybrid search (BM25 + semantic) for retrieval
  • Azure Monitor receives ServiceNow incident event stream via webhook for SLA prediction model input
03

Agent Surface

Semantic Kernel orchestrates the incident lifecycle — calling ServiceNow APIs to read ticket context, Azure AI Search to retrieve KB resolutions, and ServiceNow Table API to post work notes and close tickets. Human escalation surfaces via Teams adaptive cards.

  • Semantic Kernel agent with ServiceNow read/write plugins for incident, change, and knowledge operations
  • Teams adaptive card escalation for incidents the agent cannot resolve — full context pre-loaded for analyst
  • Azure Monitor alert rules trigger the SLA breach prediction agent on queue depth and age thresholds

Use Cases

Real outcomes from Azure AI + ServiceNow deployments.

Password Reset & Account Unlock

Scenario: Employee messages the IT bot in Teams: 'My account is locked.' The agent authenticates the request via Entra ID MFA, calls ServiceNow Table API to create an incident, invokes the Entra ID API to unlock the account, and resolves the ticket — all within 90 seconds.

Outcome: Password and account lockout incidents represent 25-35% of helpdesk volume at most enterprises. Automating these fully frees analyst capacity for complex incidents while improving employee experience.

P1 Incident Auto-Triage

Scenario: A monitoring alert fires and creates a P1 incident in ServiceNow via REST API. The AI agent reads the incident, queries the Azure Monitor log stream for correlated errors, retrieves the relevant runbook from Azure AI Search, and posts a structured triage summary as a work note within 60 seconds.

Outcome: Incident commanders begin response with full diagnostic context rather than spending the first 15-20 minutes gathering information. MTTR for major incidents reduces by 35%.

Proactive SLA Management

Scenario: The SLA prediction agent monitors queue depth and ticket age every 5 minutes. When a ticket is predicted to breach SLA based on assignment group historical resolution times, it alerts the service manager via Teams and optionally re-assigns to a shorter-queue group.

Outcome: SLA breach rate drops significantly as proactive re-prioritisation replaces reactive escalation. Service managers have predictive visibility rather than reacting to breaches after they occur.

Problem Identification from Incident Clusters

Scenario: The problem management agent runs nightly, clustering incidents by affected CI, error codes, and resolution category. When three or more incidents match a pattern within 30 days, it creates a Problem record with AI-drafted root cause hypotheses and links all contributing incidents.

Outcome: Problem management team identifies systemic issues weeks earlier than manual review would allow. Problem resolution rate improves as root cause hypotheses are better supported with linked evidence.

Built With

Azure technology stack

Azure OpenAIServiceNow REST APIAzure API ManagementAzure AI Search (KB indexer)Semantic KernelAzure Monitor

74% of your tickets could resolve themselves. Let's find them.

Book a call and we'll audit your ServiceNow ticket categories to identify which ones are candidates for full AI resolution.