Technical Support: Could AI Bots Replace Human Tech Support?

13 Apr

As I wasted many hours on the telephone the other day with the technical assistance team of a significant computer manufacturer, I was reminded just how much the current specialized support is anything but reassuring. From regurgitating scripts which have no similarity to the problem at hand to specialized staffers that lack any familiarity with the company's goods, the unhappy state of technical support now raises the question of whether AI robots must mainly replace fundamental frontline technical support staff?

A year ago when my old desktop was having problems, I phoned technical support to ask a replacement for a hard drive that had died in my RAID and also for any extra instructions I had to rebuild the RAID. The very first words out of the mouth of the support staffer were "is that the computer plugged in?" I replied that, yes, the computer was on, I'd the RAID diagnostic applications up on the display and that I was ready to browse off the malfunction codes. The answer? "Ok, that is fine, but let us confirm that your computer is plugged in, can you assess the wall plug." I again repeated the pc was on and I was ready to read off the RAID error codes from the screen. The next response? "Great, but let's make certain the other end of the cord is plugged directly into the back of the computer." And so it went for ten whole minutes as the staffer refused to budge out of his delegated script, even forcing me to answer again and again as he walked through every question related to if the computer was plugged in and powered on, while I kept reminding him that it was clearly plugged in and turned on if I had been operating the diagnostic utility on it and would we please just move on to the driveway error I was seeing.

Finally, after ten minutes prior to reassuring the tech the computer was on, the next step was supposed to pass me on to some more experienced support representative who specialized in storage. From there I was passed to some other representative who specialized in hard drives and ultimately to some RAID support technician. At each juncture I had been requested to re-explain the entire scenario from the start, read off each of the error codes, only to be told what I already knew, I had a deadly force and that they had to hand off me to a lot more specialized. Finally, after being passed from person to person to person, I was told they would finally send me a replacement drive.

When the drive finally came several days after, it was both the incorrect interface plus a small percent of the necessary size. Yet more, I needed to repeat the entire procedure of forcing them that the personal computer was plugged in and functioning, that I needed a replacement disk and that the drive they sent me was the incorrect one. Ultimately, after a week and a half of back and on, I finally had the appropriate replacement drive in hand.

Welcome to this world of technical assistance in 2018.

It appears truly ridiculous that in 2018 a household name computer maker would need me to invest ten minutes assuring them that my system is powered on if I am reading from the display or to move me from person to person and send me the wrong replacement drive. For something as simple as a dead hard disk, it ought to be fairly trivial to spend the error code from the company's own diagnostic software and the serial number of this machine and also verify that the drive needs to be replaced and appear the particular drive model that needs to be sent as its replacement.

In similar fashion, after a vital service ceased working in a major web hosting business, it required over three months of daily interactions with their senior support staff to first prove to them the service was broken and to help them know what the problem was and how to fix it. In essence, for users with any level of technical experience, your own interactions with technical support staff these days is more akin to you personally instructing them how to diagnose and repair the matter and compelling them to do so, as opposed to reporting a problem to the specialists who built the system to really repair it.

In just about any interaction I've had over the last few years with specialized assistance staff from any business, it's been amazing just how little real technical understanding or familiarity with their own products they appear to have. Fundamental technical concepts appear to elude them, they seem to have not seen, let alone touched, their own products and their mindless script regurgitation mean that even those with some level of insight are seldom able to leverage their knowledge to really assist a customer.

This raises the issue of why we're still using humans at all for frontline technical support. If they are just carrying out a canned script and reading answers from an email guidebook, wouldn't it make more sense to transition that the vast majority of these service over to AI bots? Such bots may easily reply the vast majority of regular support calls, mixing all available information regarding the client's account with thorough knowledgebases which are constantly updated. Any change to a product that offers or configuration requires just one update to this common knowledgebase to instantly be revealed in all subsequent support requirements. This would leave the human support team for the intricate cases and genuine unknown scenarios. After all, a request to get a replacement hard drive might be trivially addressed by a bot that just needs the device serial number and the RAID error code to determine whether the drive actually needs to be substituted, the warranty status of the machine and the proper replacement drive to send.

A frequent argument of the past has become the enormous complexity of building high-quality AI bots. Yet, the rise throughout the past couple of years of conversational bots, "smart speakers" and house assistants like Alexa, Google Home, and Siri has caused it to a wealth of tools, expertise and experience that may help companies quickly roll out state-of-the-art AI bots with minimal investment. Moreover, when building along with programs such as Google's Dialog flow, the underlying building blocks, in voice recognition to speech understanding, are constantly improving, meaning the resulting bots will get better and better over the years since Google enhances the services which feed to it.

Now's conversational AI techniques offer a Q&A experience that, while still quite simplistic and crude, is lightyears beyond the helpfulness of their normal human-based frontline technical service desk of several major companies. As such tools get smarter and better in synthesizing data and reasoning about novel scenarios, they will soon be able to transition out of frontline screening towards actually solving fundamental support problems and finally assisting expert support staff in narrowing down potential root causes and advocating solutions. In chosen scenarios, they might even have the ability to address intractable problems through pure brute force or invent creative solutions no human could have considered.

Putting this all together, we have reached a crucial crossroads when it comes to technical assistance in 2018 at which the current frontline support offerings at many companies have significantly diminished in grade, just as AI conversational robots have experienced an exponential increase in capability that's closely aligned with the needs of frontline support. Maybe in the future once we're having a technical difficulty, instead of spending hours on the phone trying to convince an individual staffer that our computer is plugged in, we will just say a few words to our smartphone and moments after get an instantaneous resolution, which makes it as easy to muster technical support since it's to request the most recent hit song on our Alexa or Google Home. Welcome to 2018 as it ought to be.

Share this post with your friends!