I recently attended an industry event where I was asked to be a panellist to comment on the impact on artificial intelligence (AI) and machine learning (ML) on IT operations and the Service Desk.

I’m always excited about the prospect of new technology transforming the work of IT support staff on the Service Desk, and in the second and third line support teams elsewhere in IT.  However, many new technologies don’t deliver all they promise, but why is this?  Are they over-hyped, under-specified, or do we just fail to maximise their true benefits?

In the case of AI and ML, I believe it comes down to preparedness, as is also the case for robotic process automation (RPA).

In order for a machine to be able to select a course of action, for example, the diagnosis of an incident or the application of a fix, it needs to have a basic understanding of the constraints within which it is working.  It needs to understand the environment, its users, the services they are consuming, the symptoms they might experience and the potential fixes which could be applied.

Armed with this information, it could then potentially capture the correct information, understand the symptoms and recommend a course of action.

This brings to mind a recent story on the BBC website about an Artificial Intelligence “doctor” who not only beat real doctors in a medical examination, but was now practicing medicine on real patients with a great success rate than a human doctor.  So, why can’t we have automated support staff and service desk agents?

The big difference here between my two examples of medical and technology professionals is the availability of source data, to provide that basic level of orientation I refer to above.  In the medical world, there is no shortage of definitive and authoritative works of literature which can be “learnt” by a machine.  In IT, we don’t have this luxury.  Our systems, whilst built upon recognisable standard operating systems, are then configured to meet the needs of a specific business need.  There is seldom documentation to support it, and often a lack of knowledge data to aid support of it.

To bring us down to earth, how positively can you answer these questions:

  • I have an accurate knowledge base containing data which my end users regularly consult
  • I have an accurate knowledge base containing technical data which my first, second and third line support teams regularly consult
  • I have a process which is routinely followed for the creation, update and deletion of knowledge data
  • I have a process which measures the frequency with which knowledge articles are consulted
  • I have a process which captures feedback on knowledge articles which enables me to improve them

I’d be surprised if you got more than 1 or 2 of the above.  If you did, congratulations, you’re in the minority and you’re more prepared than most to move to AI based IT support!

If you didn’t, don’t panic.  I believe that if organisations are to truly maximise the benefit of new technology, they must first develop the knowledge content, in the same way as the medical sphere have done.  Without this, we’ll have lots of technology which fails to maximise its potential and won’t live up to the hype.  Again!

Some might say, “we don’t need knowledge data.  We only need past ticket data.  We can analyse this and use it as the basis for machine learning”.  To this I say, yes, that’s a great idea in principle.  However, it assumes that you always categorise your incidents and requests consistently every time.  It assumes you always record the resolution fully every time.  And it assumes your fix always works, first time.  How realistic are these assumptions?

Please feel free to get contact us to discuss your specific issues and challenges, not only with your desire to improve and optimise generally, but how you can position your organisation to truly benefit from the new technology on the horizon.