AI Agents in IT Operations: Automating Incident Resolution and Monitoring

The rise of AI agents in IT, combined with the growing complexity of enterprise operations, which often take place in multi-cloud, hybrid environments with expanding amounts of data, is putting massive pressure on technology leaders within large organizations.

The welcome news is that properly orchestrated AI agents can meet rising challenges, such as increased incidents, longer resolution times, and overworked human agents. According to a recent McKinsey survey, IT function is a common focus area for AI solutions. “78 percent of respondents say their organizations use AI in at least one business function… The business function that saw the largest increase in AI use in the past six months is IT, where the share of respondents reporting AI use jumped from 27 percent to 36 percent.” [1]

This blog explores some of the ways that AI agents can assist across all stages of incident resolution, including a case study with European information and communications technology (ICT) services provider, Getronics.

Understanding the Role of AI Agents in IT Operations

AI agents are an emerging technology that use large language models (LLMs) to communicate with humans and to perform tasks in the real world. In the right technology ecosystem, AI agents can make decisions, collaborate to complete complex objectives, and solve problems autonomously (or semi-autonomously).

For many enterprises, pursuing Agentic AI in the IT function presents a vast array of opportunities for improvement. AI agents can be designed to play an integral role in automating incident triage and can monitor and analyze information within an IT function, acting autonomously around specific tasks or objectives. 

Because AI agents can communicate with both humans and machines, they can unify communications in hybrid settings with multi-cloud and multi-application requirements. AI agents for IT can also access information across spreadsheets and unstructured documents, making them more flexible than rigid approaches, such as robotic process automation (RPA) or business process automation (BPA).

Properly designed and orchestrated AI agents can create a more efficient system for triage that leads to faster response times with better experiences for customers and employees. Taking an Agentic AI approach can also reduce costs and make it easier to identify and track recurring issues. 

AI Agents Automating Incident Triage 

The term “triage” is a French word that means to separate out or sort. In IT, incident triage generally begins with a user providing an incident report, which is usually followed by these steps:

1. An initial assessment identifies the problem.
2. The incident is categorized by type and severity.
3. Using its severity rating and importance to operations, the incident is prioritized.
4. Once the incident is categorized and prioritized, it can be assigned to the right person.
5. Once the incident is resolved, a report is filed.

Figure 1 shows how AI agents can automate this process. A user submits an incident report, and the AI agents are able to parse the information, which might contain easily categorizable information (e.g., numerals in pre-defined form fields) as well as unstructured data (e.g., user descriptions of the incident). Once the AI agents have assessed, categorized, and prioritized the incident, they can assign it to the correct person in the IT function.

Figure 1: Automating incident triage

Once the IT function fixes the issue, an AI agent is notified and can let the user know that the issue has been resolved. This closes the incident report, but the AI agents’ work is ongoing. An ecosystem of AI agents can continuously track logs, metrics, and performance, providing tailored reports for different department leaders. AI agents can also use anomaly detection models to spot issues before they escalate.

These kinds of automation can make the IT department more efficient and dynamic, as incident reports stop falling through cracks in communication and are routed to the correct people or teams who can resolve issues. AI agents can play a critical role in reducing Mean Time to Resolution (MTTR) — a critical metric for incident resolution — while also improving the quality of the interactions humans have with the process.

Leading Technology Services Provider, Getronics, Automates ITSM Ticketing

OneReach.ai recently worked with Getronics, a leading European information and communications technology (ICT) services provider, to design, deploy, and evolve AI agents for IT service management (ITSM) ticket automation. Leveraging the open and flexible OneReach.ai Generative Studio X (GSX) Agentic AI Automation and Orchestration Platform, Gentronics has AI agent logging ServiceNow tickets for requests and incidents. These AI agents guide users through issue resolution and updates, including assigning tickets to human agents. Some of the key functionalities of their approach include:

• AI agents are available across channels — including Rich Web Chat (RWC), Microsoft Teams, and email — can communicate in multiple languages and can perform user authentication.
• AI agents integrate with multiple systems, including ServiceNow ITSM, Systrack Diagnostics, and a range of internal systems to automate common use cases.
• Human-in-the-Loop (HitL) functionality increases user support for more complicated transactions.

Taking a dynamic approach to Agentic AI, Getronics tracked massive improvements in efficiency and containment:

• 862 million  hours saved workflows automated by AI agents
• 4-8 weeks AI agent deployment time
• 85% automated ticket resolution
• 1 million+ tickets managed on an annual basis.

Challenges and Considerations 

In today’s multi-cloud, hybrid environments, enterprises seeking to effectively introduce Agentic AI into IT operations need to look for AI agent orchestration platforms that are capable of integrating with existing SaaS systems as well as multiple data types and sources. Platforms also need to be open to new technology so that organizations can leverage the best tools in the marketplace at any given moment. Given the rapid evolution of LLMs alone, businesses of all sizes need to prepare for disruptive innovations by keeping their technology ecosystem open.

The importance of an open ecosystem has also been emphasized by the sudden rise in interest surrounding Anthropic’s model context protocol (MCP). This protocol standardizes communication between AI agents and external tools. MCP provides two-way communication between AI agents, data sources, and tools. MCP servers live in an open-source repository, and Anthropic has shared pre-built servers for enterprise systems like Google Drive, Slack, and GitHub, Git, Postgres, and Puppeteer.

Legitimate AI agent orchestration platforms will also be able to handle unique security concerns and are able to put guardrails on the LLMs that power AI agents, so that they align the requirements of IT function and the broader organization as well. Human-in-the-Loop (HitL) is also a critical capability that allows humans to step in when AI agents need support. HitL can also create learning opportunities for AI agents to improve their ability to automate IT workflows.

Conclusion 

Gartner has pointed to a future where consolidated platforms will enable end-to-end business process orchestration and automation and offer a unified and cohesive experience for both business and IT users. [2]

As AI Agent orchestration platforms evolve, so will the capabilities of AI agents, including advances in reasoning and context window size. As flexible tools with the ability to communicate and execute tasks across systems and functions, there are numerous use cases for AI agents.

When it comes to how to use AI Agents in IT support, the answer will look different for every organization. By starting with processes that are highly compatible with agentic AI, such as  incident triage, businesses can achieve short-term gains on efficiency and user experience. 

Sources:
[1] “The state of AI: How organizations are rewiring to capture value,” McKinsey
[2] “Quick Answer: Beyond RPA, BPA and Low Code — The Future Is BOAT,” Gartner