Don’t believe the myths surrounding scaling up your artificial intelligence. An expert tells us how to do it properly and get the most benefit from it.
TechRepublic’s Karen Roby talked with Greg Douglas of Accenture, an artificial intelligence (AI) company, about the myths surrounding scaling AI. The following is an edited transcript of their conversation.
Karen Roby: As the pandemic continues, many companies are looking to scale AI projects, and there are a lot of myths out there about how you actually take this technology to scale. You guys have put together a really interesting study that looked at a real cross section when it comes to AI. Let’s talk a little bit about the most important findings and what really jumped out to you from this study.
Greg Douglas: We interviewed 1,500 executives across 16 different industries around the world, so a really vast survey, to see where our clients and where companies were at in terms of scaling and deploying artificial intelligence. And that’s everything from machine learning (ML) to robotic process automation (RPA) to neural networks, the entire gamut of what we would today call artificial intelligence or applied intelligence. And we found a few interesting things.
SEE: TechRepublic Premium editorial calendar: IT policies, checklists, toolkits, and research for download (TechRepublic Premium)
The first one is that really only 20% of those companies had made any progress in truly scaling their AI initiatives. And by scaling, we mean moving out of the proof-of-concept stage into something where they were truly using applied intelligence to instrument and drive their business operations across more than one function. Only 20% of those companies surveyed had reached that level of maturity, and an even smaller number had really industrialized applied intelligence and artificial intelligence across the entire enterprise and were truly driving their business and making key financial business decisions based on the output of their artificial intelligence. That number was in the low single digits. What that tells us is that 80% of the companies, and many of the folks listening to this that work for those companies, still have a long way to go to achieve any kind of scale.
Karen Roby: Some of the myths that we’ve talked about and that you guys are dispelling in different ways, one being that doing it right doesn’t mean doing it fast. I think that that’s something that a lot of companies or people or C-suite members often think that that’s something that has to be done quickly, but that’s not really the case, right?
Greg Douglas: That’s exactly right. One of the myths is that it’s all about speed, and in reality it’s not about speed. What we found is that those companies that were very pragmatic and took the time to establish the right data foundation, to put the right processes in place, to put the right leaders with the right skills in place, that takes time. And they were willing to take the time to do that and establish a rock solid bedrock foundation of their artificial intelligence initiatives, and then they went to scale. While they moved through that fairly quickly, they didn’t rush to just deploy proofs of concept and quickly scale the organization. Those are the ones who failed [the ones who rushed]. There is something to be said about going fast by slowing down. And that’s what our strategic scalers have been able to do.
Karen Roby: And then the other big thing besides the time involved, the money. Spending more doesn’t always mean a better outcome.
Greg Douglas: That was perhaps the most interesting finding of the entire survey, was that you would be led to believe that those who had industrialized for growth, those very single-digit percentages, or those strategic scalers, the next 15% to 20%, who had clearly used this at a great scale inside their organization, you would believe that they had used a lot of funding to do that, right? They had invested a lot of human capital, a lot of financial capital to get there. The reality is, that our survey found, that they actually were spending less, that they were very strategic and focused and precise on what they wanted to spend money on in terms of organizing their data, hiring the right talent and deploying specific capabilities into their company than the 80% who hadn’t reached that level. The 80% who were still messing around and playing with proofs of concept were actually inefficiently using money and spending more than those had been successful. It’s a quite interesting finding.
Karen Roby: When we talk about the operating models, I think this is a big myth, that IT departments own the AI operating models. But again, not always the case.
SEE: Natural language processing: A cheat sheet (TechRepublic)
Greg Douglas: It’s been all the rage to hire a chief data officer. And I’m sure many folks that are listening to this may be chief data officers and have a chief analytics officer in your company. While there’s nothing wrong with having that position, assuming that that position will be the end-all, be-all and control point for all things applied intelligence, artificial intelligence in your organization, we’ve found that that actually doesn’t work. It’s not about just a single leader. You need excellent leaders like those people in your organization, but it’s about a multidisciplinary team. Those that were successful had a multidisciplinary team of business leaders, finance leaders, technology leaders, including data officers and analytics officers and CTOs and CIOs, and they worked in a collaborative manner, almost in a pod, if you will, to decide what functions should be utilizing artificial intelligence and exactly what tools should be deployed and how to organize their data. When they did it in a multidisciplinary fashion, those are the ones that ended up in the strategic scaling category at the top.
Karen Roby: How difficult is it finding talent, people who are well versed in AI?
Greg Douglas: Certainly many of the technology employees around the world are very keen to learn. What we’ve found is that there’s a hunger and a thirst inside of a lot of the organizations for their employees to really gain these new skills. When we look at the training curriculums that we roll out for our clients, the number one ask is digital capabilities in the umbrella, but underneath that, specifically, it’s around analytics capabilities. How do I deploy analytics capabilities and the tools like artificial intelligence, machine learning, RPA, neural nets, etc., around that? Number one, employees are thirsty to learn. Number two, there’s a set of great training curriculums being developed by a number of firms out there, and ours included, to help clients really scale up and deploy on that. Number three is that we’re seeing almost a wave of college graduates coming out who are learning these skills in the universities, whether that be undergrad or graduate programs. In fact, a number of universities are starting to offer degrees in analytics and data sciences now, and so we’re seeing a lot of the young, new graduates coming into these companies having skills that many of the more mature experienced employees don’t have. It’s really interesting to see the pivot that’s happening across the technology and business landscape in terms of skills.
Karen Roby: One of the big myths that I find really interesting is that big companies equal big AI pain. But again, it doesn’t always correlate.
Greg Douglas: No, it doesn’t. Big companies tend to have a lot of data, as an example, right? One of the things we learned is it’s not about having more data, because sometimes more data equals more pain, right? One of the things we often hear when we go in to start to deploy these types of capabilities is that the clients say, “We’re certainly not short of data. We have petabytes of data.” And actually what we’re really trying to get them to do is focus down on 100 to 200 key pieces of data that are going to drive and fuel their AI and machine learning capabilities. It’s not about more data that creates pain inside of the organization because people get lost in a sea of information, it becomes overwhelming to decide, it’s really about deciding what are the key pieces of information and key outcomes you want to get from that information, and focusing on that. When you don’t do that, you experience a lot of pain, a lot of slowness, a lot of bogged down, in terms of not being able to move at the right speed.
View original article here Source