CIOs are starting to use different AI and machine learning capabilities to improve IT service management proce...
Use cases that incorporate AI and machine learning in ITSM typically involve natural language processing and AI-infused mining of ITSM data.
Natural language processing (NLP) is being used to automate user requests for IT services. As the underlying technology in virtual agents, NLP is often configured through integrations with chat services like Slack that employees are already using, giving users a familiar interface for communicating with IT service desks.
Applying machine learning to ITSM data gives IT practitioners a richer understanding of their infrastructure and processes. Applying machine learning to ITSM processes, in principle, makes a lot of sense due to the volume of data ITSM systems generate. The systems collect lots of data about what's being requested, along with information about when, why and by whom. This data also provides a glimpse into the IT assets and processes that are in place and can help identify who owns them, how they are used and whether they are still relevant.
"Data is the fuel that AI needs to deliver relevant and valuable insights," said John Peluso, CTO of AvePoint's Public Sector. Machine learning generates insights that can help IT organizations prioritize ITSM issues, take preventative action, improve time to resolution and ultimately boost employee productivity.
Ready to dig a little deeper? Here are 12 ways CIOs are using AI and machine learning in ITSM and changing how IT services are delivered.
1. Automatic categorization of incidents using chatbots
Chatbots integrated into ITSM infrastructure can be used to categorize the underlying problem in employee requests. For example, Genpact has added the BMC Chatbot to its BMC Helix SaaS suite, which automates connectivity with ITSM infrastructure across cloud and on-premises infrastructure. The chat interface simplified the ability to prioritize and automate level one and two service requests for 50,000 Genpact users.
2. Intelligent assignment for incoming requests
Service desk teams have different skill sets, and some team members are better at resolving certain types of IT requests than others. Incorporating AI in ITSM can automatically triage tickets to the correct support groups without having to have humans reading the content in the ticket to make a decision, said Milind Wagle, CIO at Equinix Inc.
3. Automatically fulfill basic requests through task automation
NLP can help chat agents handle categories of requests and incidents. These agents can help answer common questions using historical ticket data and an ITSM knowledge base.
But this application of AI in ITSM requires knowledge management experts to create a repository of proper documentation of past request history and relevant knowledge articles, said Kumaravel Ramakrishnan, evangelist at ManageEngine, an IT operations and service management provider under the Zoho Corp. umbrella.
This article is part of
4. Generating a solutions repository
Enterprises are starting to embed IT operations provisioning tools directly into the chat services used by developers. These DevOps bots help generate a unified repository that allows developers and operations teams to keep track of changes to infrastructure and how different types of incidents were successfully resolved. Later, when similar problems emerge, AI engines can mine this repository to help operations teams resolve the problem more quickly.
IT service desks are adopting the same types of tools to generate knowledge repositories for broader sets of IT service requests as well.
"IT professionals with experience could tell you that most of the issues they solve today are based on past experiences with solving issues," said Oded Moshe, vice president of products at SysAid Technologies Ltd., an IT help desk provider.
AI can also help mine IT service request data outside of the chat channel into repositories that can be used to solve current issues. This augmented repository speeds up the process for solving issues.
5. Event resolution guidance
A good ITSM repository can also guide incident resolution through the ITSM tool. AI-generated advice could be as simple as suggesting a related incident, solutions article or configuration item, thus reducing the time required to think about how to locate an item and then search for it.
Guided use of the ITSM tool is even better, said Matt Cox, director of solutions engineering at N-able, an ITSM consultancy. In this case, the AI creates best practices and automatically points IT service desk practitioners toward that behavior, rather than forcing them to identify the best use of an ITSM tool. One example is to create an algorithm to identify related incidents and their trends for problem management and suggest that agents open a problem record.
6. Learning process flows optimized with machine learning: Use case
Many IT requests, like employee onboarding, require human agents to perform a complex set of steps to fulfill the requests. Enterprises are now using machine learning models to watch how humans execute these processes so that future requests can be more automated. In the case of employee onboarding , machine learning models learn from a historical database of requests that cover a range of actions taken based on type of employment, role and department of the new employee. The trained models then assign new requests to the right technicians. By recognizing patterns in the employee onboarding request database, machine learning-based models can also suggest what hardware or software an employee needs right when the onboarding request is created, said ManageEngine's Ramakrishnan.
7. Proactive problem resolution improved by big data analytics
Advances in big data and analytics are improving the predictive and correlative capabilities for ITSM. Based on analysis of the repository and user behavior patterns, AI and machine learning tools can help reduce the number of IT issues end users experience, or forecast and fulfill users' requests before they even know they have a problem.
"Issues ranging from IT outages to individual user hardware malfunctions can be predicted, and solutions automatically applied or at least suggested with an increasingly higher rate of success as the system learns from past experiences," said Ambarish Kayastha, regional CTO at Broadcom Inc.
AI enables better, faster, proactive and automated resolution of problems introduced through changes in the environment, changes in end-user behavior or changes within your apps and services, said Andreas Grabner, developer advocate at Keptn, a service delivery automation platform.
For example, Citrix is using ITSM integrations across Dynatrace, ServiceNow and AWS. When the Dynatrace engine predicts a problem, it can do root cause analysis before customers are impacted.
8. Anomaly detection by flagging unusual repeat incidents
The signs of some IT incidents may not be apparent through traditional ITSM monitoring tools. AI machine learning models are being trained to detect anomalous behavior that may occur across multiple IT systems. These models can help alert IT staff to a problem before an incident has occurred.
9. Using predictive analytics to flag requests that could violate SLAs
IT service requests can result in spinning up software and hardware configurations that degrade an application's performance or, worse, break the app entirely. Predictive analytics is being used to mine performance data within and even across enterprises to identify potential problems. This insight can provide guidance to users or the IT service desk on alternative approaches for fulfilling a request that meets service-level agreements.
10. Identifying security vulnerabilities
Security researchers are constantly identifying vulnerabilities with commonly used IT infrastructure applications and configurations. This can include libraries for developing applications as well as infrastructure. AI tools can interpret new reports and prioritize issues for security teams to address before a recently discovered vulnerability can be used by attackers.
11. Improve data quality
Organizations often have dozens of cloud or on-premises applications that generate large amounts of data that change daily. IT sits in the middle, working to provide solutions to frustrated users when these applications have data health problems -- e.g., human input errors that propagate through integrations, inconsistent data across applications or issues arising from application upgrades.
Egidio Terra, senior data product manager, machine learning at Talend, said an AI and machine learning substrate could help improve data quality. For example, AI and machine learning can monitor data and provide fixes or recommendations to improve data health. This can happen at any point -- for instance, using machine learning to check data at the data warehouse, or as the data flows from the sales application to a marketing application, he said.
These tools can also be configured to detect data problems and alert owners of downstream applications and reports. For instance, if the digital marketing team decides to use a new tag on its webpage, which then breaks an existing business intelligence (BI) report, the tool should detect the data drift and alert that team. In this case, the problem can be fixed before the BI report is generated and thus prevent a new service request from being created.
12. Automating access to systems
Companies use ITSM to architect, design, build, deliver, operate and contract access to services and systems for their stakeholders. Using AI to automate access to systems can drive huge efficiencies, shortening the cycle for granting access from days to minutes, said Terri Sage, CTO at 1010data, a provider of analytical intelligence to the financial, retail and consumer markets.
"When machine learning is brought into these processes, it becomes possible to automate these tasks," she explained.
Sage suggested that teams start these initiatives by gathering enough technical knowledge to provide adequate data for training and testing the ITSM system. Good sources include runbooks and incident systems that contain sufficient documentation of methods and steps to convert into a knowledge base for use in the ITSM system.
It is also important to introduce these automations in phases, with proper oversight by an IT specialist. "An improperly trained and tested machine learning ITSM could cause additional incidents, such as mishandling of data or improperly granting access," Sage said.