The current state of AI in service and support is underwhelming. Promises of bot-powered agentless support, problems that automatically correct themselves, and preemptive problem resolution—the holy grail of service and support—have gone largely unfulfilled.
But there are bright spots on the landscape. I have seen more than 100 demonstrations of AI tools in the past few years. Only recently have these demos convinced me that we are at an inflection point with AI. There are now credible demonstrations of AI that not only exceed the capability of human experts, but are getting smarter over time. They exhibit true machine learning.
Here's a recap of where we are, and what's still on the horizon.
A brief history of AI in service and support
AI was declared a fantasy as early as 1985, when IBM programmed a supercomputer known as Deep Blue to play chess. Deep Blue lost its first match against world champion Gary Kasparov and would lose others before finally beating Kasparov in 1997.
The story of Deep Blue is the story of AI more broadly. Initially, IBM hired chess grandmasters to program Deep Blue. Since none of the grandmasters played at Kasparov's level, Deep Blue was unable to beat Kasparov. It wasn’t until IBM pursued a different strategy—machine learning, whereby Deep Blue would learn from its mistakes and improve its game over time—that Deep Blue finally won a match against Kasparov.
By the same token, much of what passes for AI today is simply human knowledge dressed up to look like AI. Some of the best-known AI tools in the industry do nothing more than read knowledge articles and then tell you what they have read. That's not AI, and it’s certainly not machine learning!
The true test of any AI tool for service and support is this: Without human intervention, will the tool reduce ticket volumes, resolve problems more quickly, decrease total cost of ownership (TCO), and improve the customer experience?
If it checks all these boxes—and gets better over time—then it's true AI, powered by machine learning.
The inflection point for AI
A legitimate proof of concept is one indication that we are at an inflection point. We now know that the promise of AI in service and support is technically feasible. The second criterion is rapid adoption and acceleration.
Here again, the evidence supports the case for widespread, rapid adoption of AI in service and support. A year ago, there were many AI trials going on in the Global 2000, but very few bona-fide AI success stories. Today, many of those trials have moved into production and are yielding the promised dividends of lower cost, preemptive ticket resolution, quicker resolution times, and machine learning.
What changed? What precipitated the inflection? Two phenomena are behind the dramatic growth of AI in service and support. The first was behavioral in nature, and the second was technical in nature.
The behavioral change was quite unexpected, but well documented. Front-line support professionals—support technicians, if you will—no longer fear AI. Moreover, at the top of their wish list is a desire to see more automation in IT support. This behavioral adaptation, illustrated in figures 1 and 2 below, removed one of the last cultural barriers (fear of losing one's job) to the widespread adoption of AI.
Figure 1. Front-line workers no longer fear AI. Source: ITSM Intelligence Report, MetricNet
Figure 2. AI and machine learning top the front-line worker wish list. Source: ITSM Intelligence Report, MetricNet
The second barrier that was overcome, thereby paving the way for rapid and widespread adoption of AI, was the application of data science to the terabytes of data generated by support organizations worldwide.
You have undoubtedly heard the expression "connect the dots." That may be possible when the dataset is limited in size. But when you have millions or even trillions of data points, some in structured data fields and some in unstructured, free-form text, manually connecting the dots is not only unrealistic, it's impossible!
Massive computing power, combined with data science, has allowed for mining insights that would otherwise escape even the most experienced IT support professional. All the data that has been sitting, largely unused, for years or even decades in ITSM systems, knowledge bases, phone systems, remote control tools, and other technologies can now be unlocked by data science to produce actionable insights that yield the long-awaited benefits of AI.
The future of AI in service and support
The beauty of machine learning is that it will take us places we never dreamed possible. Nevertheless, we can look at the repeatable success stories in the industry and make some reasonable predictions about where AI is going. These demonstrated successes include the following.
Automated problem detection and resolution
We all know that the best ticket is the ticket that never happens. Effective AI makes this a reality by preemptively detecting and correcting problems before they become tickets.
With AI data mining algorithms, it is now possible to operate at Level -2 on the shift-left spectrum (illustrated in Figure 3 below). This "search and destroy" capability, not possible before the advent of AI, reduces TCO by lessening the workload on IT service and support, and it saves enterprises time and money by preventing many tickets altogether.
Figure 3. The power of shift-left. Source: MetricNet
Ticket categorization using NLP
Natural-language processor (NLP)-based AI engines automatically identify and group similar tickets based on formatted data fields as well as the free text data contained in most tickets. This resolves the decades-old problem of incorrect ticket categorization and eliminates the "other" ticket category altogether.
Automated problem management
Problem management is one of the most important ITIL practices, yet few IT support organizations spend much time, if any, nurturing this crucial discipline. By automating problem management, effective AI identifies and eliminates the root causes of incidents and service requests, thereby reducing ticket volumes and saving customers time and money.
Evergreen experience and expertise
IT service and support is a high-turnover industry. For enterprise-managed service desks, the average annual agent turnover is more than 40%, and for MSPs the number approaches 100%!
This has a devastating effect on the collective knowledge and experience of a support organization. The most effective AI tools keep an organization's knowledge evergreen by systematically capturing, cataloging, and reusing years or even decades of combined agent experience, knowledge, and expertise.
Improved knowledge and ticket quality
Knowledge and ticket quality go hand in hand. A knowledge base is only as good as the quality of tickets that feed it. Effective AI automatically measures and grades ticket quality and pinpoints specific actions to improve the quality of tickets. Simultaneously, these tools add to the knowledge base by creating new knowledge articles, updating existing articles, and purging articles that are obsolete.
Greater visibility and accountability
Effective AI provides unaltered visibility into people, process, and performance. It identifies top performers and those who are underperforming. It uncovers process inefficiencies, and automatically corrects them.
And with unmatched visibility comes unmatched accountability. Effective AI drives accountability by identifying the strongest performers while providing targeted coaching for those who are underperforming.
The implications for service and support pros
This article is no advertisement for AI. I am as skeptical as anyone when it comes to the more far-fetched claims that are made by some AI vendors. But the above successes are not hypothetical, and they are not "one-offs." I have highlighted these specific examples because they have been demonstrated time and again and because they have produced quantifiable benefits.
I have spent 30 years in the IT service and support industry, as a consultant, speaker, and author. When I began my career, front-line help desk agents were mere message takers. Today, they are some of the most highly skilled professionals in IT. This transformation was necessary because the demands of the job required ever more skilled professionals to meet the needs of an increasingly complex IT environment.
AI has, and will continue to be, a disruptor in the industry. Initially it will eliminate the need for agent-based commodity support—think Microsoft Office, Windows, password resets, and other easily resolved problems.
But as machine learning makes each deployment of AI progressively more intelligent, even the most complex support provided by today's customer-facing agents will be replaced by more intelligent bots. Even so, the majority of agents see AI and automation as a good thing because they believe the technology will transform careers for the better.
Much as today's auto industry assembly-line workers are engineers monitoring computer screens while robots actually build the cars, the support technician of the future will become a support engineer who monitors, coordinates, and directs the efforts of the AI bots. For the best and most talented in the industry, the future of IT service and support has never been brighter.
Keep learning
Choose the right ESM tool for your needs. Get up to speed with the our Buyer's Guide to Enterprise Service Management Tools
What will the next generation of enterprise service management tools look like? TechBeacon's Guide to Optimizing Enterprise Service Management offers the insights.
Discover more about IT Operations Monitoring with TechBeacon's Guide.
What's the best way to get your robotic process automation project off the ground? Find out how to choose the right tools—and the right project.
Ready to advance up the IT career ladder? TechBeacon's Careers Topic Center provides expert advice you need to prepare for your next move.
