by Jennifer O'Brien

How to innovate with analytics and deliver real business value

Jun 21, 2018
Artificial IntelligenceBig DataDigital Transformation

Adopting advanced analytics and the road to artificial intelligence and machine learning is top of mind for many organisations looking to discover true business value, according to SAS vice-president of global tech and industry practices, Bob Messier.

And that’s because analytics not only delivers increased automation and speeds up processes, but also enables the identification of new models, the discovery of hidden relationships, and can unlock a new level of intelligence, Messier told CIO Australia while visiting from the US.

“Companies can now take advantage of different data and are seeing customers in a different light, and are able to create better customer experiences.”

CIO Australia sat down with Messier to discuss the future of the technology, how organisations are capitalising on AI and machine learning and what’s needed in order to jump into the game.

Messier, who has 21 years experience in the industry, is all too familiar with the range of customer pain points and challenges, and various hindrances to adoption.

But he’s on a mission to help lighten the burden, reveal the true business value of advanced analytics, and help companies transform their world of data into a world of intelligence.

Understanding the trio of elements

Certainly, advanced intelligence – and the innovative push towards AI and machine learning – is so important to business and to the world, according to Messier.

But while there’s a ton of buzz with respect to AI, in particular, businesses are overwhelmed with information.

“They are struggling to really understand what it is and we need to simplify it for a lot of our customers. In many ways, what we’ve tried to do in terms of messaging is break it down and explain there are three components: computer vision; machine learning and deep learning; and natural language.

“Much of what we see, if we look at these three things, is that there are typically three different types of data. We have images and video for computer vision; for natural language processing you are typically converting audio to text; and machine learning we see it as unstructured and semi structured data. With a lot of the machine learning and deep learning we see on web sites we see different types of customer attribution.”

In that vein, he said customers need to understand the three different elements and then recognise within those elements that there are different mathematical routines applied to the three different types of data.

“Actually, there are different algorithms that are applied to the different types of data. So even as I say that, you have a matrix of three different types of data, different types of algorithms, and different types of use cases. It gets complicated for the customers to understand what it is,” he said.

“We’re moving from traditional batch analytics to near real-time instantaneous analytics. And if you want to get closer to a transaction and what’s happening, you’re going to have to move to things like images and video.”

But taking a step back, he said many companies are looking to simply automate – and are actually on the roadmap and trajectory to doing AI.

“Today, lots of analytics are still done in batch. But as you get closer to automating, you get closer to real-time, you get closer to the point of transaction, and you’re actually moving on a continuum towards AI.

“It’s a journey and I think the big part of AI that’s overlooked is automation. AI is as much about artificial intelligence as it is about analytics and automating analytics.”

Building a use case

Asked why companies are jumping into the advanced analytics space, he said the reasons are many and varied, and are growing in scale and intensity across all verticals from retail to health to financial services, as companies seek new and innovative ways to compete and improve the customer experience.

“From operational efficiency, to mitigating risk, mitigating fraud (closer to when the transaction occurs), to using different techniques to do cancer screening and cancer detection, to using AI to figure out how we provide the next best offer to a customer. There are lots of different use cases.”

From a customer point of view, he said companies can utilise the technology to vastly improve the customer experience, and engage the customer better.

“Can we identify, even through image or video, the stature of a customer coming into a store and then be able to provide them a better customer experience. Or even provide a better layout of the merchandise in the store if we know the size and shape profile of people coming into that store. You can reduce your markdowns. So there’s lots of different use cases.”

And while companies are all grappling with how and where to jump into the advanced analytics game, or simply continue on their journey, Messier urged companies to just get started and keep the ball rolling. Looking at current company use cases – and determining how to enhance them with text or image data, or with semi-structured data, for example – is a sound strategy.

“In a lot of today’s use cases – using AI and these new algorithms and the new data we can look at – they are actually enhancing existing use cases. For instance, take a look at online fraud. Online fraud was done a certain way a couple of years ago, but now you can embed machine learning algorithms on the transactional data that will help you identify it faster, and see different patterns more rapidly.

“If we think about risk, banks manage risk, but now if you think about AI and leveraging unstructured data like text, I can actually score emails and score dialogues that people are having with customers and use that to predict risk in a different way than we predicted it in the past. So it is the same use case, but now we’re applying different data and different algorithms to those use cases.”

Indeed, the use cases are endless, he said, explaining the predictive analytics piece is so exciting. “Looking at videos, if we can see what’s happening on a street in video, you may be able to predict or stop a crime before it happens. And that’s a very different way of thinking.

“Or if I listen to and I convert customer service rep conversations into text, and I analyse it, I could predict loans that are at risk, before they are at risk. So if I look at these unstructured data types, it’s actually moving the analytics closer to when things are happening.”

Getting the data strategy right

But before companies can reap the benefits of the advanced intelligence, there are steps to follow.

When determining best practises and setting up an analytics strategy, Messier said businesses – and heads of technology – need to be “really mindful” when cleaning the data and determining the quality of the data.

“Sometimes if companies apply too much data quality and rigour, they actually lose some of the value of the natural data. So that’s one thing I would watch out for. The other thing you want to rustle with is how frequently do you want the models to be updated. So it’s not just about the data and getting it in the model, it’s the frequency that you want the model updated.”

He explained the bigger problem – “more than getting your arms around the data” – is actually the deployment of the model once it’s built.

“This is where we see companies struggling. And it could be for regulatory reasons or it could be that they build a model in one language and then they rewrite it into another language to deploy it. I’ve seen banks, believe it or not, with a 270 day lag between when they build a model, and when they have it deployed. By the time you build a model and deploy it, consumer behaviour could have changed and you’ve lost the opportunity.”

Getting the skills house in order

Messier said one of the positive things to come out of AI and the overall movement towards advanced analytics is the fact business and IT are now working more closely together than ever before.

“Going back six or seven years ago, and sitting with business in the room to discuss analytics was a difficult conversation. IT is responsible for running operations and the analysts are actually responsible for what I call ‘discovery’. And the discovery process is an iterative process, where IT is not as iterative, it’s a production process. And it seemed like those two worlds would collide.

“But now I see with more agile development, I see a closer connection between the statisticians and the architects. But I also think advances in hardware, and customers moving to the cloud, and advances in software, have allowed those two parties to work more closely together. I’m seeing progress on that front.”

But where he’s seeing gaps are in the ability to manage change and drive an AI or an analytics strategy across the organisation.

“I see a group like marketing or the fraud department or the risk department driving a strategy across the department, but I still don’t see, at most organisations, a pan-organisational adoption of running the business on analytics. I still think we’re away from there. I actually think artificial intelligence and this kind of market movement will help us get closer.”

The common phrase, ‘we have executive buy-in’, Messier said simply means someone signed the cheque.

“That’s good and that’s required, but what we lack and where there’s room for improvement is, are the executives willing to make decisions based on what the analytics tell them. Right now, the move towards AI is around automation, automation, automation. But are people talking about the decisions that need to be made from the output of AI or analytics.

“Culturally, the challenge with making decisions on analytics is often times the results of analytics, or the outputs, can be counter-intuitive.” Companies are typically used to measuring themselves in a certain way, by a certain yardstick, and are now having to adapt and change and ask, ‘what metrics really matter?’

“Every company has a dashboard and probably a scrollbar on the right hand side of some internal web page with all of the measures on it, but are they measuring the right things?

“When we start getting executives asking those questions, and using analytics to solve that question, that’s when more broad adoption of analytics and not just the math, but making decisions based on the math, will start to happen. I would say we’re five to seven years away from that.”

Future moves

Looking ahead on the technology advancement front, he said it will be hard to disconnect AI, IoT and cloud in the future.

“The data coming from IoT chips and sensors is going to be streamed and there’s so much data coming off of them, you are going to be looking to store it, either on site or in the cloud. If you are streaming the data in, you will be applying most likely a machine learning algorithm to that data as it is coming in.

“AI and IoT are so intricately linked,” he said, explaining the SAS vision is to enable customers to leverage the cloud, in whatever deployment and service model option is appropriate (private, public, hybrid or community), and a decision based on their use case.

“We have built an architecture that will allow us to deploy the technology, as our customers see fit, and it will depend on their use case on how they want the technology deployed. With IoT, we typically see a cloud component. Almost every single IoT use case we see a cloud component.

“With AI we don’t always see a cloud component. A lot of times that’s happening behind the firewall. So it depends on the use case, but being able to support and work in various clouds has been a big part of our strategy.”