Artificial intelligence is often described as a replacement for human interaction.
In the IT service management environment it is making service desk agents’ jobs easier – and in some cases even creating new roles.
There are many established service desk use cases for artificial intelligence including: automatic incident categorization, AI-assisted issue resolution, robotic process automation as well as virtual service desk agents.
Many our customers are using machine learning with different open source algorithms for automatic incident categorization, albeit with varying degrees of success. In some cases, these companies have reverted to manual categorization. Yet, as the ability for algorithms to handle a larger range of categories develops, we are confident that automatic incident categorization will become commonplace.
We are seeing real value generated in the use of AI for issue resolution. Here the technology helps the support agent by providing suggested results for handling email or chat conversations. Unlike automatic incident categorization, the aim is not to replace the agent, but rather augment them by providing responses that fit the conversation.
Robotic process automation can be categorized as the most basic form of AI, although a number of our customers have achieved significant benefits from being able to automate legacy processes. However, its lack of flexibility will see its influence decline in the coming years as advanced ITSM tools become more intelligent.
A typical agent may have half-a-dozen screens open, handling and searching information surrounding requests. A virtual agent can provide suggested responses in real time, allowing the agent to respond quicker and more accurately. AI not only improves response times – by automating routine tasks, it also frees up the service desk agent to handle more challenging conversations, leading to increased job satisfaction.
Training the bots
Chat bots are only as effective as the training they receive. This creates new opportunities for customer service agents to become ‘chat bot trainers’, helping bots to understand user needs and improve their responses.
Artificial Intelligence can be trained through supervised, unsupervised and reinforced learning.
An example of supervised learning in a service desk environment is, when a ticket with a certain issue is received, the system is directed to offer a specific solution from the knowledge base. In this scenario, the system learns from users input.
Unsupervised learning equips the system to identify patterns of incidents – which human agents may not be able to recognize. Reinforced learning combines elements from the first two models.
A key challenge of unsupervised learning is the quality of data. Many knowledge base articles and incident histories are not suitable for significant pattern recognition because the data quality is poor and/or the knowledge base is not properly maintained.
When to call in a human
Virtual service agents need to be able to handle actions (for example, ‘I need a new phone’), information (‘When will my phone arrive’) and knowledge tasks (‘What is my password’). They must also recognize questions they should not answer, ensuring inappropriate requests are not handed over to human agents, generating unnecessary additional costs for the client.
Moreover, AI needs to recognize when it is time to hand the request to a human agent. This requires the ability to use semantics to determine when the chat bot is no longer able to deal with the request. This is one area where we see opportunities for support agents to become chat bot trainers.
AI will save clients time and money by automating routine requests and providing faster resolution for repetitive issues, offering a better customer experience and improving service desk agents’ satisfaction.
To dig deeper into the topic of AI at Service Desks, you can view our recent webinar on Future of AI-powered Service Desks.
For more insight about AI in IT Service Management read our blog series about AI: the 4th Generation of Process Automation in ITSM or download our ebook on the topic.
We are always happy to discuss how AI could boost your service management. Request a meeting below.