May 11, 2018
VARBERG, SWEDEN - With 21 billion Internet of Things (IoT) devices estimated to be in operation by 2020, user and consumer data lakes are predicted to skyrocket. Those ‘things’ must become autonomous and intelligent. Otherwise, the great promises that IoT holds will fall flat in the face of manual maintenance and supervision.
As businesses look for ways to best leverage IoT data - whether in industry or the consumer sphere - machine learning and edge computing start-ups are rising to the challenge of providing effective cognitive platforms to supercharge the IoT for industry.
Ekkono is a Swedish start-up specialising in 'making connected things smart'. Offering a horizontal edge intelligence solutions for any industry utilising connected devices, Ekkono aims to provide IoT-ready enterprises with advanced machine learning capabilities even on constrained hardware platforms. In practice, that means the ability to implement predictive maintenance, machine automation, optimized production, and self-improving products.
[caption id="attachment_11423" align="aligncenter" width="305"] Jon Lindén, CEO and Co-Founder @ Ekkono[/caption]
We sat down with Ekkono CEO Jon Lindén to find out more about how enterprises can effectively leverage AI-IoT technologies to the benefit of their business.
What should the key priorities be for someone starting out on their AI-IoT journey?
“Data, data, data. You will always need data. In my opinion, there are two kinds of companies – those who understand the value of data, and those who don’t. Every company in every industry must assess what data assets they have, and what to do with it. Next, you must decide what problem to solve. This is known as Business Understanding. Based on the problem, you can dive deeper into how the data you have can help solve the problem – this is called Data Understanding.”
“You must understand the value of data and expedite the first steps where you start aggregating and processing it. You will always need more data, better granularity, and sensors that you don’t have today. Think about it as a roadmap, and take quick steps to move forward and get first results. If possible, design your product so that you can update software, have some spare computation and memory capacity, and preferably even the possibility to retrofit additional sensors.”
“For piloting and proof-of-concept, you can either update the software on a subset of units, or attach a separate hardware to do data aggregation and edge analytics. The latter can be achieved using a standard IoT gateway like Dell’s, or you can even duct tape a Raspberry Pi or cell phone with standard sensors to your product.”
"To summarise: get started, think roadmap, and focus on the small wins and initial conclusions that can be used to align and enthuse all the stakeholders."
What are the challenges of implementing edge intelligence for IIoT applications?
“Cross-organizational understanding. You see this in the technicians who know the product inside out, who can put their ear to the machine and figure out if something’s wrong, but who see sensors as something that only shows when the machine is failing – rather than being a data source that can be combined with other sources to build a complex picture of what’s happening, 24/7.”
"The kind of horizontal, cross-organisational implementation that IoT demands is also a major challenge. The big benefits come from harmonizing solutions across all product lines, both because they can better one another, but also because the investment is significant. A lot of industrial companies are used to launching new products on a product line, but struggle when they are to provide a horizontal product spanning all their verticals."
“Legacy systems still remain an issue for many organisations, and typically it’s a gigantic project to extract data from them. Usually they are pre-big data, which means that they are designed to store a high-level overview based on blunt averages, rather than details. This is the opposite of what IoT and the possibilities of edge computing represent. Try and get what you can from the legacy systems but, if possible, build a parallel infrastructure using a modern cloud-based platform like AWS IoT, as well as edge computing.”
“Finally, I think that a lot of industrial companies are very much ‘product companies’. The service organization is almost a necessary evil which is all of a sudden placed on a pedestal and made central to a more service-centric business model.”
Which industry sectors stand to gain the most today from AI-IoT synergies?
"I think the industrial sector is closer to short-term monetary gains. Thanks to remote technologies, less people can support more products globally. It enables entirely new business models for them, which are easier to implement in one-to-one B2B relations than in a marketing-driven consumer space. It also enables better usage of labour resources, as only relevant events are automatically filtered for manual handling, as well as premium services and products that vendors can charge extra for. We have seen that industry, automotive, and energy verticals are pretty far along in this process, but of course, it's eventually more a question of companies rather than whole industries."
“Having said that, consumer markets will potentially see more long-term gains, i.e. when individual lessons within companies become crucial to customer loyalty and lock-in. The advancement cycles are longer in consumer IoT, since new features typically ship with the next generation of products.”
"Making connected consumer products smarter enables a lot of the same benefits that we are seeing in industry, allowing for premium editions of products, expert and remote support, less resource-intensive support requirements, and more service / data-driven business models."
As told to Ciarán Daly
Ekkono will join 185+ leading sponsors and exhibitors at The AI Summit London, June 13-14. Find out more about how you can get involved.
About the Author
You May Also Like