My colleagues and I have had a really good conversation going of late on the topic of self-learning systems. Specifically, can the IT operations systems being adopted by nearly every ITO provider (think Arago and IPSoft) actually learn, as service providers claim they can? And if they can learn, why does it matter to IT operations leaders?
Recently, I’ve begun pushing back on the term "self-learning," because I think it’s misleading. My sense from enterprise IT buyers is that they assume intelligent automation systems acquire knowledge by themselves – i.e., that they "learn." If this were the case, it would be the Holy Grail of IT operations: a system that learns how and when to solve problems would mean that service desk and L1/L2 staff would have to focus on only the really complicated stuff.
I think we’re beginning to see this kind of capability, but the term “learning” needs to be qualified. In these systems, the term is specific to how problems get solved, not the solution itself.
Think about the difference between vocabulary and grammar. Vocabulary is the set of words specific to your language that you know. Grammar is the set of rules that dictate how to put words together. To be effective at your job, you need a deep vocabulary and a broad understanding of grammar. Otherwise, you may have too few words to say what you mean, or the words you do say may be gibberish.
Think of solutions to IT problems as vocabulary. For example, clearing the temp directory of a Linux computer when the free disk space is less than five percent – this would be an example of a word in your vocabulary. Now think of all the other standard solutions you have to the problems you usually encounter in IT. This makes up your entire vocabulary – combined into a dictionary of solutions to common problems.
But how do these systems know when and how to use words from this dictionary? This is where grammar comes in, and where something akin to learning is taking place.
To solve problems, these systems either randomly pick a word out of the dictionary to see if it solves the problem, or they ask a human if they are using the right word from the dictionary. Either way, what the market calls a “self-learning” system eventually finds the right word or combination of words and remembers it for the next time. In this way, the system learns the grammar – sometimes by trial and error, sometimes with the guidance of humans – that it needs to put the right words together.
Why is this distinction important to IT leaders?
IT leaders who invest in these types of systems must understand that they are the teacher. And as the teacher, they need to educate these systems on the vocabulary they want them to learn – and this takes a lot of time. In some cases, service providers will bring a dictionary to the table, but a IT leaders will need to review each word in the dictionary to make sure its meaning is relevant to their specific environment. And they’ll need to teach the system new words as the environment changes.
In the case of a sourcing agreement, the service provider is the teacher. I think some providers are underestimating what it will take to teach these systems and provide the “continuous education” services they will need to stay current. When a service provider assumes an intelligent automation system will learn quickly and take over more and more work – or if it assumes the automation capabilities will eventually catch up to the service level it committed to – it can put customer contracts at risk.
And, while these new systems may be able to make response times faster and improve some service levels to some degree, providers will struggle to meet their internal profitability targets because they’ll need too many people to deliver on commitments, which will inevitability pass back to the customer in the form of diminished service quality. Then we’re back to the same old watermelon problem all over again: green on the outside, red on the inside.
So, as you’re thinking through your intelligent automation strategy for IT, remember you – or your provider – are the teacher. And, as everyone knows, teachers have a very, very hard job.