Accenture Report: Cloud and Edge Can Foster Reinvention with AI

Organizations are interested in exploring cloud and edge to differentiate and compete, but there is still room for growth in this space, according to a recent report from Accenture.

Accenture surveyed 2,100 C-level executives in 18 industries from 16 countries, asking about their interest in edge and to what extent their organizations had adopted it. Though edge has its appeal, its adoption rate has not matched.

In the report’s findings, while some 83% of those surveyed saw edge computing as essential to competition in the future, 65% of organizations use edge currently. Accenture also found that companies that adopted edge as part of their cloud strategies saw better outcomes.

There might be pressure to accelerate edge adoption as the world rushes to embrace the capabilities of AI. The report indicates that 98% of executives around the world expect AI foundation models to be important to their organizations’ strategies in the coming three to five years. Edge may be a factor in those strategies because, as the report states, it enables real- or near-real-time data analysis, which can simplify training AI models and improve the performance of AI-driven apps.

Teresa Tung, Accenture’s cloud first chief technologist, spoke with InformationWeek about how cloud, edge, and AI are being wound together in IT strategy.

Related:3 Tips for CIOs: Pitching Edge to Your Organization

What is the equation that is playing out here? From the report, there is interest in edge and cloud, but then there is also this factor of the AI arms race -- does that have the potential to drive edge adoption?

I think edge has actually been around longer than cloud. A lot of the traditional compute and automation -- that’s where a lot of this data and computer has been born. In our first two types, you have initial type one, which is ad hoc, which was more from a very traditional compute like virtual machines, IoT type of setting and oftentimes those would be companies like energy companies, or some of your traditional retailers or hospitals that had a little box doing the compute.

But then you have the tactical set that was started by some of these package solutions to bring together the AI and the compute and some of the more modern capabilities. But because it was through these ISVs [independent software vendors], it’s the easiest way to scale, but also meant it wasn’t extensible. So things like your point-of-sale device or loss prevention in a retail setting. What you just hit on, this equation, the value for the company comes from being able to take that valuable data, whether it’s from the IoT devices or some of these point-of-sale devices and to be able to reuse them for other purposes and to be able to extend that case on their own -- we know how important data is.

Related:Why the Cloud Needs an Edge, and Vice-Versa

We started seeing that value comes when we’re thinking about the edge as a part of that extension of the cloud continuum. A lot of these people who are seeing bigger value are being able to take edge almost like that common platform. Just like how cloud, when you’re able to imagine an idea, try test and apply it in the digital space, now can we do the same thing within a physical space within a store or within a factory?

Some of the things that we’ve done for a large, global packaged goods company is exactly that. Extending so that in all their factories we have this common blueprint of getting edge. When you do edge analytics, in this case it was for product giveaway, they were giving away too much of their product. We pulled that data in, used the cloud to actually create the model, push the model to the edge, tested it in one of the factories. When it worked, we can easily see all the other factories that have similar types of engagements and deploy it.

That same sort of story that we had in the cloud and digital -- that A/B testing that we’re all familiar with in improving our experiences -- is now in that physical space. That’s super integrated; there’s a subset of that integrated that is really thinking about that extension, not just of cloud as we know it and containers and moving that around, but it has really differentiated AI. The automotive industry, just look at how that’s transformed every part. Your car and the electrification of the car and then the software in your car towards self-driving -- the way you build that new type of car is both a combination of new factories that they never had before. A lot of those factories, not only did they change like I mentioned in that package, good story, but they’re doing a lot of that commissioning of the factory all in software. You’re going to simulate it all before even rolling it out because there’s so much change, and if you do a physical change, it’s tens of millions or hundreds of millions of dollars impact. Instead, you could do this faster by doing all that change, simulating it as much as possible and then delaying some of that manifestation.

Related:Edge Computing Eats the Cloud?

There’s this new sort of economics that when people are thinking about the edge, they really need to think about this, not so much the infrastructure like the VMs [virtual machines], the racks, and stuff. That’s certainly important, but thinking about integrating their data, AI, and cloud strategies.

In the report, you’ve got overall interest in edge versus actual current adoption. It’s a notable disparity. It’s not like it’s 80-20, but there’s some lag there. Is there any indication of what’s behind that? Is it just organizations taking their time with the adoption of something that is new to them, or is there anything with edge that has been a stumbling block for organizations?

I certainly think there’s been stumbling blocks and that’s part of why we talked about these different archetypes. Some already have a type of edge, especially in the ad hoc and the tactical. It might be owned by a particular line of business, maybe it’s in the manufacturing, or in a labs, or in a HPC [high-performance computing] type -- all of these would be different buying patterns, so it’s different parts of the business doing their own thing, which makes sense. We don’t want to stop that.

There’s a gap in adoption, but some have already adopted. There’s about 50% who are in that ad hoc and tactical category and there’s another 50% who are doing it in this integrated way. The way to do it in this integrated way, it’s not the ways of those traditional buying behaviors. It’s not the HPC team suddenly growing and it’s HPC everywhere. It’s actually a more natural extension of your cloud and your data AI strategies, which now talk about value. It’s thinking about, like I mentioned with the CPG company and that product giveaway. This was from us identifying ahead of time there was a billion-and-a-half-dollar opportunity in terms of this factory of the future program. We knew there were lots and lots of things you can do and rather get stuck in this paralysis, it was using this cloud type approach to edge that we could start. So, with that giveaway problem that was just one of those opportunities in that $1.5 billion. We were able to say, ‘Let’s use the cloud extension and extend that, use some of these cloud offerings to the edge and be able to pull some of the data that was coming off this production line, along with analytics that was now pushed out onto the container in that edge environment and then to try. Let’s just see, does it actually save anything?’

We were able to reduce the giveaway by 80% and it was tested only in the order of three months. We were able to say we’ve proven the value and now I can scale that to all the other factories. And so that savings from that single case allowed us to fund the rollout of this edge capability to these all these other factories around the world. That that speed sounds more like this digital sort of story. It’s not what HPC or self-checkout loss prevention is going to be able to do. This is a different type of behavior. It’s a different type of ownership.

With the concerted attention that AI is getting, are there battleground areas where AI is expected to flourish? How significant is the cloud as a space where AI might further spread its wings?

We call it the cloud continuum and that is not central cloud. It is literally a continuum into the network and then into our products in everyday spaces. I think the automotive example is really good one. There’s things that happen in AI directly in the car and that happens really low power and at a moment’s notice. Even when the car is off and I want to unlock the door via voice power, it can’t be running a big model all the time. It needs to be super passive and use something like neuromorphic compute and to be able to identify this is Teresa’s voice and it should open the door.

When the car is on you can do a lot of things with the self-driving car -- that can’t have even a round trip time within a network. It’s all in the car. Then you have things that will take all that data and learn from my car and your car and all these others. That still needs the cloud. You have Tesla even creating their own chips; it’s such a unique differentiated factor to be able to even train those really large models as well as those really small, low-power models. That ability to work across the continuum, this is why we’re calling this super integrated. Before, you might choose a cloud partner and that’s kind of in that integrated part; you find the cloud partner to work with. This continuum now goes all the way to the edge.

Are we coming to a point where edge adoption and cloud will be vital to competition? If you don’t try to get on board with this, will you be left way behind in the dust?

I think that’s not just us saying it. Our survey results find 81% of those respondents said they do believe that the failure to act is going to lock them out of the full benefits and some of it we’ve already seen this play out with the digital disruption with the companies that were able to embrace cloud-first and be able to tap that agility into their digital businesses. They were able to jump ahead. In the automotive industry, if you just look across the board all of them are adopting and even partnering with silicon providers in the same way that they were picking their cloud partners. They’re doing a lot of that again, commissioning factories in simulation, self-driving cars, embedding the AI into the car to create a differentiated experience. They’re doing that all across the board and in the same way, definitely seeing the same across the other industries.

And this is part of why we’re saying don’t get locked into that ad hoc or the tactical. Just like you have modernization with cloud, you’re going to have modernization at the edge.

You’re going to get savings just by modernizing that traditional edge environment.