Advancing the service desk: how innovations in artificial intelligence and machine learning will change the way your help desk works
It seems like the internet can’t stop talking about AI and machine learning. Everything from practical applications to recent advancements and ethical discussions are becoming an almost daily occurrence. There’s no doubt machine learning will have a huge impact on AI and in turn the IT industry as a whole. It’s also pretty clear that this forward march of progress is coming very soon to a service desk near you.
Despite suffering what may ultimately prove to be a minor setback in 2017, AI research continues to progress at a rapid pace, with huge names in the tech industry getting involved in nearly every sector of business. Google, Amazon, IBM, Apple and countless others are either starting to get their feet wet in the pool of artificial intelligence or have already dived in head first to try and gain a market advantage years down the road.
Many of these research and development projects are barely beyond conception stages, with some less ambitious projects gleaning results here and there. These developments will eventually transition into very real projects that have the potential to revolutionise a number of industries. It’s not only possible but incredibly likely that the service desk software of the future will be entirely automated, from ticket reception to ticket resolution.
Artificial intelligence and machine learning snapshot
If you haven’t been keeping up on machine learning in the last year, here’s a brief snapshot of what the community has been working on. For those unfamiliar with how these concepts are interlinked, most experts believe that machine learning is going to pave the way for future artificial intelligence breakthroughs. Machine learning is a process where AI is placed through a series of iterative scenarios or stimuli, committing each to memory and gradually “learning” whatever the developer is attempting to teach it.
A very practical example of this would be facial recognition software, which faced a handful of setbacks this year when it was discovered that a poor set of sample images resulted in a “racist” AI that was unable to recognise the faces of people who weren’t Caucasian. This highlighted a slight flaw in the current machine learning process: that the aggregate sample data being used as the learning process for a given AI needed to be far more diverse than originally thought. Not only that, but it underscored the potential for abuse and misdirection when “teaching” an AI a given set of tasks.
Another powerful example of machine learning comes from the entertainment industry, with Unity Technologies releasing the Unity3D machine learning suite, giving game developers working with the Unity engine the opportunity to teach new AI mechanics using true machine learning while building their games. This approach has already found popular application in strategy games, from Starcraft to DoTA, where intelligent bots have been extremely popular and successful for several years.
Much of this machine learning comes from a process originally outlined by Geoffrey Hinton called backpropagation. Hinton has dubbed this method a failed experiment, stating that this method of iterative machine learning wouldn’t lead to the development of a true AI. The announcement shocked the AI community as many of the major developments in AI have come from the backprop method.
What does this mean for the service desk?
Looking at the practical applications for AI and machine learning, the service desk is an obvious choice. The work is often repetitive, time consuming and usually requires low-level service personnel to address. It’s a perfect candidate for an advanced AI to handle from start to finish once the technology has reached a sufficient level of complexity.
Companies have already started turning to artificial intelligence to control their service desks. Most of these operate at a very basic level, taking rudimentary keywords and routing requests to a knowledge base for the end-user to take advantage of. This results in what appear to be glorified automated responses, with little interaction and actual problem solving being utilised by the AI.
But the seeds of change are there. As AI learning continues to progress, the possibilities for almost complete automation within the service desk become closer to reality.
Machine learning gives a potential AI the chance to sit and analyse common service desk issues, learning from typical responses, and beginning to “reason” on its own what the best course of action would be to take in any given situation. This available data can be fed from many different sources, with multiple companies involved in machine learning providing cross-platform learning and repositories for AI to pull data from.
In the near future, it’s easy to see how this kind of collective information pool will enable programs to paint a comprehensive picture of service desk application. Even complex tasks could be worked through solely by the AI after enough sample data.
These autonomous programs wouldn’t have to respond with the same cold precision that they analyse data. Programming human-like responses into AI is already happening on a routine basis, and the end-user experience relies as much on courtesy as it does on actual information and problem resolution. Soon an AI service desk application will be able to mimic most of the functions of live support personnel.
This doesn’t completely remove the need for actual humans to respond to tickets, but it will drastically reduce the amount of personal attention service desk personnel will need to deal with. Only the most complex or unusual tickets will need an in-person response, with most of the heavy lifting done by the service desk application itself.
In the not-so-distant future…
Real applications of intelligent AI are coming very soon, and it won’t be long before developers will be taking these machine learning concepts and applying them to their own applications. It’s not a question of if but when the industry starts seeing AIs handling ticket requests from start to finish almost entirely on their own. While the overall impact of machine learning on the service desk remains to be seen, it’s inevitable that we will start seeing it in the near future.