AIOps - why the key word here is ‘intelligence’

In April, diginomica's Jon Reed gave Avantra a warm and much appreciated welcome, as we made our entry into the ranks of diginomica partners. At the same time, and in his own inimitable style, Jon threw down a challenge, relating to our use of the term ‘AIOps'. As he put it:

One of my running jokes at diginomica — if I make fun of a buzzword too often, invariably, I'll end up writing about its virtues. That holds true for Avantra's approach — which, in a phrase, is AIOps for ERP. Or, if you'd prefer, automating IT operations. Despite the potshots, I have no objections to a phrase like AIOps — as long as Avantra can convince me that AIOps is genuinely different from the ERP automation tools that have come before.

To which I'd like to respond — challenge accepted!

A matter of definition

The main problem here, as I see it, is that both terms — ‘ERP automation' and ‘AIOps' — are open to misinterpretation. So a great place for me to start might be to explain how we at Avantra define them.

Let's take ‘ERP automation' first. This is problematic, because while Jon is 100% correct in saying that what Avantra provides are tools to automate IT operations, and more specifically in the context of enterprise SAP landscapes, others might see that term and come to an entirely different (and incorrect) conclusion about what we do.

For them, ERP automation may well refer to the automation of business processes — the handing over to a computer system of steps in a business workflow that might otherwise be handled by human employees. Examples might be invoice matching or credit management. There are plenty of tech companies out there that do this, and do it well — but that is not how I would define AIOps.

So now let's turn to the term ‘AIOps'. For me, the big drawback with this term is that AIOps focuses too much on the ‘how' (the use of AI) and not enough on the ‘why' (the outcome for companies that adopt it.)

When I talk about these outcomes, I'm thinking of costs avoided, crises averted and skills diverted to more valuable work. Those responsible for keeping complex, business-critical ERP landscapes up and running know all too well that this work demands specific skills, which are difficult and expensive to acquire, and many hours of work spent on routine tasks just to keep the lights on and put the fires out. As I see it, what AIOps offers in an ERP context is the chance to automate routine maintenance tasks in a way that is faster, more cost-effective and more accurate than human intervention. 

Just one part of the picture

But AI is just one part of this automation picture. The key word here is ‘intelligence'. You can use AI to automate tasks, certainly - but in some cases, it might be overkill. What I'm basically saying here is that the ‘how' of AIOps is a whole mix of things.

It could be configuration — the stuff you have to do to inform the system of who is responsible for resolving an issue.

It could be simple rules relating to best practice — for example, is this particular security setting turned on?

It could be a basic forecasting algorithm — how many hours before this disk or table runs out of space? 

It could be machine learning — how do I automatically adjust thresholds, based on experience and training, so that they are set at appropriate levels for this environment? 

It might be natural language processing, to parse logs and notifications, to understand what actions need to be taken.

Perhaps, in light of this, a better moniker might be ‘intelligent operations', but Gartner coined the term ‘AIOps' some time back, and that's the term that the broader market seems to have adopted.

One of the biggest challenges you face when you launch a new business is figuring out how that business can describe itself in a way that helps people easily understand who you are and what you do. We use the term ‘AIOps' because it serves that purpose for us — even if ‘AI' is just one method that we deploy in pursuit of a wider goal.

But on that wider goal, we can be very clear — it's about delivering smooth-running ERP landscapes, by using a range of methods (including AI), to automate critical operations tasks intelligently. If those tasks are left undone, you get systems that are insecure, prone to failure and major contributors to wider business disruption. In other words, you get the kind of chaos that I believe every ERP  customer can agree they are adamantly keen to avoid.