The automation of AI and analytic systems was a hot topic at the AI and Big Data Innovation Summit, which recently took place in London. Google Cloud's Robert Saxby even shared with delegates details about how Google used an autonomous AI to substantially reduce its energy consumption over the course of 2018.
DATAx sat down with Lee Greene, AI analytics and business intelligence specialist an Anodot, a real-time analytics and automated anomaly detection company which provides business intelligence to its customers, to discuss why companies appear to be gravitating toward automated analytics.
DATAx: First things first, what exactly is autonomous analytics?
Lee Greene: No one has actually ever asked me that question before! I would say it's taking the analytics process and condensing it into a system that enables it to move seamlessly and automatically. I guess that would be my immediate response to the question.
DATAx: But that is the textbook answer, is it not? What if you were explaining it to someone outside of the tech space?
LG: Well, I do have to do that a lot. I have family members who ask me what I do who have no idea what AI or algorithms are. It's basically very smart math that analyzes the behavior of users and applications, and lets people in the business understand when they are not operating profitably.
DATAx: That sounds pretty straightforward…
LG: One of the things that I really take pride myself in, is to take these really complex ideas and repackage them into something digestible for anyone.
DATAx: So, in your opinion, why should everybody be using automated analytics and why are they not already?
LG: I think everyone has a need for it. I explain it to people this way: Some companies need the Ford and some companies need the Ferrari. A Ford is like an open-source tool which you can build yourself and apply to a few metrics that are important to the business. It works well for smaller organizations, ones with not a lot of exposure to risk and with smaller platforms. They don't need what we [Anodot] are providing – the ability to analyze hundreds of thousands of metrics at once at minute or hourly granularity, giving visibility into the entire business and beyond like operational infrastructure. All of those important KPIs that are moving at rapid speeds and our models are able to understand them.
So, for companies that have scaled to the point where they are now at millions of data points a day, what we're able to do is condense that down into a couple severe alerts that they will need to know about in a week. We're talking about making the insane amount of information the decision-makers in business need to know about clearly. We call this "root cause analysis", the ability to an answer why it's happening, not, "it's happening. Why?"
DATAx: So, in a hypothetical world where every company used autonomous analytics, what kind of differences would you envision in the world?
LG: Really, what we're looking at are facts or time it takes to reach the resolution of an issue. Companies struggle to identify when something is happening or when something has gone wrong. Ten years ago, what people needed was some form of visualization reporting tool which they would ask to tell them about this or that, and the tool would add a nice visualization and provide some insights. But if you don't know what question to ask, then its pointless.
However, when you can have a set of algorithms analyzing every single permutation all the time and telling you when there's a problem, then the data has a voice and you don't need your dashboards any more. We're looking at the evolution of the dashboard analytics space, understanding your business. AI in general is meant to automate repetitive tasks, modelling time series data is a repetitive task which you don't want to put data science resources into but is necessary to do. You have to understand your business metrics all the time, to prevent loss or failures with customer service.
We're able to do this so accurately and at such speed that it doesn't make sense for it to be done any other way – and that's AI in general, that's why it exists. And that's where we're at with it now. I can't tell you what's going to happen in the future; people try but it's hard to model that as the math isn't quite clear. But in the present, we're pretty good at it; and by present, I mean hourly or less than a day.
DATAx: What is the biggest hurdle for any company trying to incorporate this kind of technology into their workings?
LG: Most of the hurdles that exist rely on current infrastructure and road maps in those companies. A lot of organizations set down the path of "we're going to do these kinds of things in our data strategy" and they don't necessarily always make sense or aren't always necessarily future or forward thinking. They are trying to catch up on the years of neglect they have shown their data resources, their warehouses, their [data] lakes. All these disparate, siloed systems that exist in so many different areas and they have no idea how to bring them together again. It's like Humpty Dumpty has fallen off the wall and they can't put it back together again.
So, for our technology to work, we need to get data from somewhere and push it to our cloud where we can run our algorithms. Companies are not always able to do that or don't know where the data is because of their antiquated structures. It can limit their ability to actually understand what is going on in their business because the data is locked away somewhere, they are not able to access it, so you hear, "we're not able to give it to you to analyze, sorry".
I think for enterprises, real enterprises, that's a massive issue. For younger companies, they are not structured that way. They are willing and able to engage in these kinds of activities because there is a value proposition. But if we can detect four or five incidents before you can and each of those incidents cost you tens of thousands of dollars, then we can save you money immediately.
You're never going to be able to catch them all and you can't build this yourself. Well, you could, but it would take an incredible amount of resources, so that is where we are trying to sit.
To hear more insights from AI and big data experts, visit next year's DATAx London. Or, if you're based in North America, check out our DATAx festival taking place in New York on December 12–13, 2018.
Find out more HERE