This feature first appeared in the Summer 2023 issue of Certification Magazine. Click here to get your own print or digital copy.
Our world is experiencing an artificial intelligence boom that has sparked excitement and existential dread over the role AI might end up playing in our personal and professional lives. The current AI gold rush is being spearheaded by IT industry giants like Microsoft and Google, but there is no shortage of companies working to catch a ride on the AI bandwagon.
The business case for the rapid rollout of AI-based products is simple: Companies hope to gain a competitive advantage by baking AI into as many products as possible. And the stakes are high: Some analysts predict that Microsoft could see AI-related revenue gains of $100 billion by 2027.
The AI explosion isn’t without its detractors. Numerous organizations and individuals have expressed worries over the possible negative consequences of rapid, unregulated AI development. While some of the AI doomsday scenarios are a little more James Cameron-inspired than reality-based, there is a growing concern over what the impact of AI taking over jobs normally done by human workers will be.
Which leads to a question that’s on many IT workers’ minds: Will ChatGPT, or something like it, end up replacing the tech people who perform IT support? To fully answer this question, we need to start with some basic facts about the current situation, and put it into some recent historical context.
AI? AI? Oh!
You may have noticed that many of the recent articles about AI (including this one) have titles that end with question marks. This is a clear sign that the near and distant future of AI is highly speculative — and that there is no shortage of speculators.
Even educated guesses about AI’s impact on our world, however, are exactly that: guesses. We don’t know for certain what the current headlong rush into an AI-driven world is leading to.
If you were involved in the IT industry during the dot-com boom of the late 1990s and early 2000s, then you’ve likely noticed some similarities between that era and the current AI craze. There is no shortage of hyperbole in the mainstream media and tech press when it comes to AI’s potential to change our lives.
Side note: If you have seen the term “AI bubble” in recent months, there’s a good reason. IT industry hype is typically somewhat based in reality, with a large side dish of marketing and venture capital funding frenzy. Let’s just say that if the AI revolution ends up fizzling out in the next two or three years, then the survivors of the dot-com bubble era will not be overly surprised.
The problem is that the term "AI" is being used (and somewhat abused) by journalists and industry experts as a catch-all technology category that’s a thousand miles wide and twenty fathoms deep. Some of this is due to the baseline definition of AI itself: intelligent behavior displayed by a machine, rather than a person.
This is a decent enough starting point for considering the subject, but it’s also a shove onto one of the slipperiest slopes out there. What constitutes intelligence and intelligent behavior?
For example, consider autocomplete, a feature found in nearly every modern word processor and text message app. Autocomplete is able to predict which words will follow other words and make practical suggestions based on these predictions.
Does this constitute intelligence? Is autocomplete an artificial intelligence?
Autocomplete makes its predictions based on two key sources of information:
1) The language training it received when it was being created
2) The history of your usage with the program it runs in
When autocomplete works well, it can save time whether an individual is writing a business case or texting a friend. There’s a reason, however, why there are so very many autocomplete memes littering the internet. When autocomplete lays an egg, the results range from annoying and/or embarrassing to flat-out hilarious.
And so autocomplete is generally viewed (by those who don’t disable it altogether in their apps) as a useful function that needs to be monitored for its small-to-outrageous missteps.
Which brings us to ChatGPT, tech support, and a possible future.
Troubleshooting (and missing)
As far back as Windows 7 (and possibly earlier), Microsoft's signature operating system included self-help tech support applets called Windows Troubleshooters. If your printer was giving you a hard time, you could launch a related Windows Troubleshooter that would attempt to identify the problem based on your feedback, and suggest a number of common fixes.
Most Windows Troubleshooters would only dig so deep into a technical issue before waving a white flag. They were designed to help users diagnose and fix the most common problems that occur with network connectivity, peripheral usage, and so on.
I don’t think many technology enthusiasts would equate a Windows Troubleshooter with an AI. A Windows Troubleshooter is more like an expert system — a piece of software that follows a pre-configured logic tree to assist a user with solving a problem or making a decision.
Another example of an expert system is commonly found on e-commerce websites. A customer can use a product selector system to answer a series of questions based on their requirements or preferences, which leads to the site recommending one or more related products. Vehicle sales sites typically offer this functionality to help you find the ideal make and model of car based on your needs and likes.
Internet service providers have been using chatbots to perform tier-one tech support for home customers for several years now. These are essentially online versions of Windows Troubleshooters, taking in the user’s descriptions and feedback and making suggestions for possible fixes. Once the chatbot has exhausted its built-in knowledge, the customer is typically transferred to a human for additional troubleshooting and possible scheduling of a technician.
Let’s take everything we’ve discussed so far in this article, and offer some educated guesses about the future of tech support.
AI and human techs will work side-by-side
Businesses that provide tech support for their products and services will employ both AI and human tech staff, in a combination that results in the best value from both. Companies always keep an eye on payroll numbers, as staff reductions are an easy way to increase revenues. But owners and executives are also aware that customer service is a critical component of financial success.
The public’s level of comfort and satisfaction with AI-based tech support is still under development. The frustrated customer’s shout of “I want to speak to a human being!” is still very much in play. Modern businesses can use advanced analytics (some of them powered by AI, naturally) to determine the exact proportion of AI/ human costs compared to customer satisfaction/repeat business revenues that results in the desired gains.
Short version: Will AI replace tech support workers? Yes, some of them, probably mostly tier-one workers. There will still need to be human techies to keep an eye on the AI’s performance (remember autocomplete?) and intervene as necessary to maintain the level of what businesses refer to as “successful customer outcomes.”
As AI systems improve enough over time, the number of human workers required to ride herd on them will likely diminish. But for now, you can expect the relationship between AI systems and human workers to be a partnership designed to achieve the best business results.
And here’s an interesting side thought: Since it is likely that most AI products will exist as cloud-based services, they will probably be dominated by the majority market share holders in the cloud computing industry (think Amazon, Microsoft, and Google). Businesses will need to carefully consider the value proposition of putting all of their tech support eggs into one cloud service’s AI-generated basket.
Down the road
Finally, there is this question: Is there a potential future in which AI systems become powerful and intelligent enough to perform all of the human aspects of IT support? Will AI systems be placed in charge of network administration and security response? At some point, will companies phase out the CIO/CTO role in favor of an AI information technology overseer?
"Speculation, this is," as Yoda might put it.
But there is this: Ten years ago, we were all being told that self-driving vehicles were going to fill the nation’s roadways over the next decade. There would be no taxi or Uber drivers, no long haul truckers, no courier or delivery vehicles with a human behind the wheel.
Cities would have to rezone into driverless sectors and human-driver permitted areas. New incredible traffic efficiencies would be achieved and maintained through AI powered vehicles taking over the streets.
Today, what we actually have are mainstream consumer vehicles with some effective camera- and sensor-powered driver-assist features. Automated emergency braking, adaptive cruise control, and automatic lane centering have all made their way into most modern vehicles.
And yes, there are some self-driving vehicle implementations out there, mostly limited to specific geographies and environmental conditions. Most cars on the road today, however, are conventional vehicles driven by humans ... with some (intelligent?) assistance.
In other words, it might be best to stay clear of both the overhype and doomsday ends of the AI conversation. Remember, progress takes time. That, and not even autocomplete gets every prediction right.