IoT, AI and edge-computing projects face complex ecosystem


Companies keen to deploy technologies such as Internet of things (IoT), artificial technologies and edge computing to further automate their business processes have to contend with an increasingly complex ecosystem of myriad services providers and rapidly evolving technologies, which makes implementation of any project a challenge.

This was  one of the major takeaways during a panel discussion at the recent APAC IoT vSummit organised by FutureIoT.

Dubbed “Powering IoT in the Workplace with AI and Edge Computing”, the panel  was moderated by Pankaj Lunia, supply chain & B2B Collaboration Solutions Leader, IBM and composed of Rajan Upadhyay, Head of Digital Lab, Cyberjaya at DHL; Riza Alaudin Syah, CTO at Indonesia-based Eateroo; Manuel San Miguel, CEO, Ignatica ; and, Atul Babu, SVP & Head of International Business, PCCW Solutions.

Burgeoning ecosystem

Any IoT implementation – the panel unanimously agreed – has a number of moving pieces, the least of which are the connected devices and sensors that collect information. Device manufacturers, who are at the hear of the IoT ecosystem, come out with new devices while rolling out enhancements to existing ones. They now monitor and measure – in ways never seen before – details in minutiae that can be found in a work or industrial environment

Upadhyay of DHL noted that the billions of data the comes through these connected devices are essential in the quest for an intelligent enterprise.

“Everything you’ll see, especially in IoT, is a kind of enabler and they are the channel interface to bring some intelligent automation…. I think that is where you can bring the value. If you see this intelligence,” he said.

He added: “AI is bringing intelligence capabilities and billions of devices are pumping the data into our systems. These systems in the edge nodes are accumulating these data and pushing them to the cloud. What are we going to do with these zettabytes of information? Intelligence is the key and the intelligence you can only bring once you we can share these information. And based on this information you can bring intelligence and effectively use it into your business processes. We can make this world more connected and safer.”

Babu of PCCW Solutions pointed out that IoT devices measuring data is just a starting point in an expanding ecosystem.

“You have the entire ecosystem of edge computing  where telcos are playing a key role,  and they are also at a nascent stage. There is hardly any telco that can claim to have a very solid multi-edge compute. They are still in the initial parts,” he said. The same is true for cloud services providers. They have a role to play and they are also adapting with time. On one side while they may have a very mature public service cloud offerings, how can they modify their services to edge compute, multiple-edge compute, and at the same time new services as well.”

A lot of these new companies are coming together to not just look at what is the advancement, but also help choose – what is the right tool, what are the right IoT devices, what are the right ecosystem components that would come together and it’s changing every single day because of technology advancements and  new tasks being added.

“Being able to stitch together becoming much more cumbersome than it was in the past. Earlier, we used to talk about working with three four technology components and that’s that about that’s about it. Now, we are talking about 20 30 components on a regular basis – it is pretty much a norm.

By the time you finish implementing, you realise these four components are no good anymore and I need to look for new ones,” Babu said.

San Miguel of Ignatica echoed the challenge of putting different components together into a functioning whole.

“How can you actually have the enterprise architecture to stitch them all together with enough rigor and robustness to manage a business model that needs to survive more than a quarter. That is where we see a key challenge but also where a massive opportunity for certain platforms are starting to come up,” he said.

Putting AI to work

Once the connected devices and sensor delivers data, the challenge is sorting them out in a way that makes sense. This is where AI comes in.

For Ignatica, which provides a platform for digital insurance, AI is playing a big role in the development of new business models that are changing the insurance industry. The technology is behind use cases such as telematics, usage-based insurance and parametric insurance, which offer pre-specified pay-outs based upon a trigger event.  (An example of this is a farmers insurance, where if rainfall falls under certain threshold a month, the insured gets automatically compensated.)

“From an AI perspective, once you have access to all these billions and billions of data points coming in on a steady basis – how do you make heads or tails of them? And you start seeing now things like algorithmic underwriting. You have automated claims decision; you can now have dynamic pricing based on changing conditions for different product types,” San Miguel said.

Meanwhile, San Miguel sees more dynamic changes on medical devices and ecosystem place.

“Think about the smart refrigerator where you keep your insulin. It measures when and how much people are actually going in there to actively manage a particular disease or condition. And based on disease management, you’re able now to reduce the premiums or to provide wellness treatments”

He added: “You are going to see a shift in insurance from protection and kind of compensation after you saw a loss event into actively managed prevention. And  that’s going to happen – not only with humans as we start wearing more and more smart clothing that tells us more about what’s happening with our bodies and what we can change – but also across PNCs. You start seeing preventive maintenance regimes across ships. You see it across freight supply lines, where the cost and the optimisation on and therefore the insurance around the supply line for how they’re using all these resources that are consumed are drastically impacted by sensor data coming in from IoT; and the ability to have a validated well-managed full provenanced data store that can now drive the machine learning algorithms.

“The challenge that insurers have is not just in getting access to the consistent stream of sensor data for triggering activities, but really having the right validation and full management of the business model,” San Miguel said. “Because one of the challenges that we’ve seen is having the consistent provenance and well-managed data from a data quality perspective and from a security perspective that you can embed into an intelligently automated business model.”

Smart tech  levels playing field for SMEs and large companies

Syah of Eateroo sees SMEs benefiting from IoT adoption. In Indonesia, where the food-tech startup is based, Syah sees customers deploy face recognition and AI-powered recommendation system.

When I was in Bukalapak ecommerce marketplace, the AI that was first implemented was a recommendation system that was deployed to suggest further purchases.  The AI and ML of the recommendation system were developed inside the app,” he recalled.

He also cited the example of deploying MLs from embedded devices to scan faces for “mass detection”.

“It can be used for example in a commercial kitchen to ensure that every chef is following protocols [in food preparation],” he said.

About simplifying the process of implementing technologies such as IoT, AI and edge computing Babu of PCCW Solutions said all companies face the same scenario no matter their size.

“I think this whole notion of big versus a small is going for a toss. It’s all about fast versus slow. That’s why you know you’ve got a 50-people company coming out of nowhere – such as Whatsapp that takes over 50 billion dollars of revenues of telcos globally. That is the power of speed over being big so whatever we are talking about is true for everyone.”

About ensuring a smooth implementation process, Babu said it eventually comes down to human skills.

“Independent of how advanced the machines you get, eventually somebody’s got to put it in the context of business that they are in,” he said. “The contextualisation of that requires some human skills. It’s still not as automated as one would want to believe. It cannot be done by one person, it’s a combination of skills cross functions and business domain. So, if you’re talking about retail, it would certainly require some functional expertise about how the technology can be implemented – something  before sensor and something  after sensor,  something before AI and something after AI.”

He stressed: “It requires huge architectural expertise as well I mentioned earlier. It’s about putting things together. I can be given 100 different components, but if I don’t know how to put them

together and make the best use of it, it’s no use.”

He also pointed out that cost plays a big role. It has to be cost-effective.

“It has to make business sense. If I am investing 10 dollars, I should  be able to get the return of investment. People are not talking about five to seven years of return of investment, which was norm earlier on, people are talking about if I am investing ten dollars,  can I get it back in three months’ time?  The whole speed context has completely changed.”

Prioritising technologies in an IoT implementation

Asked about how to prioritise from multiple IoT components of a project, Upadhyay said it is not easy since one is faced with a whole ecosystem – of which sensors and devices form a small part.

“Perhaps I can classify this into two spaces: industrial IoT and consumer IoT. It varies a lot from each other. In consumer IoT, we are talking about some household items with basic sensors. With industrial automation, there are many different ways to look into it.”

Overall, what’s important is not only the device, Upadhyay said companies have to look into the network they have established, which is critical. The platform chosen is also critical because it enhances the interface with the channel and bring the effectiveness in their business processes.

“Consider all these three factors. The first thing to identify is what sort of IoT the use case is whether it’s consumer or industrial IoT. Then, the second stage is what network are  you using and how you do computing into it. The third stage will be how to accommodate the information, and how it can bring effectiveness to your business processes.”


Show More

Related Articles

Back to top button