Element.AI CEO Jean-François Gagné: “Big companies could be hurt by unwise AI implementation’’.

AI remains a sexy field for investments, especially if a startup includes some of the top people in the industry. Jean-François Gagné, an experienced serial entrepreneur, and his co-founder, a prominent AI scientist, Yoshua Bengio, launched Element AI last year. The company’s AI as a service (AIaaS) and applications platform aims to be the go-to place for enterprises looking to apply AI solutions to their businesses. The company closed a $102 million Series A round in June 2017, counting among its investors not only VCs but also industrial and financial institutions like the National Bank of Canada and South Korea. Within months, the company has grown from 5 employees to 200, and continues to expand.

Jean-François Gagné, CEO of the company, interviewed with buZZrobot to discuss the challenges of AI implementation in enterprises and the future of the AI industry.

Could you explain how your product is applied in the enterprise business process?

Well, we are building a platform for enterprises that enables them to run on much smaller datasets, making it easier to integrate AI solutions into their business practices. We extract a high-level representation of different companies’ concepts of data sets and use that as a starting point. Right now, we’re debating and formulating touch points, so the process is dynamic and needs to be standardized.

What use cases does it aim to tackle?

I can list among our use cases: manufacturing projects (improving the quality of the process of a chemical plant; applying deep learning to the fine-tuning of a chemical process); leveraging IoT; supply chain; AI as service partnerships with financial institutions.

We are growing the platform with applications for each domain, translating the advantages that we created at lower-level growth of big applications. We are targeting the ability of AI to scale everywhere to reach Fortune 2000 organizations.

What do enterprises not understand about AI technology?

AI technology is still in the educational and implementation stage of an actual solution. For a lot of businesses, it’s difficult to distinguish a difference between rule-based systems and real AI systems. On one end of the spectrum: if they don’t use AI properly, it creates friction, so there are no expectations. Another group might think it’s magic: They say, “Oh, yay! Let’s just do this through AI.” It doesn’t work that way either.

The market has just started to understand how this works. The education process continues, not just in the technical aspect, but more in terms of what it all means: interacting with different types of systems, harnessing people’s strengths, learning what to avoid, and learning what the downsides to training are.

How do you convince enterprises to adopt AI technology?

Well, sometimes they experience fear of missing out. We show them a chart from the McKinsey report called ‘Artificial intelligence: The Next Digital Frontier’, the achievements of different industries, and who has adopted AI and what they gained. They see a curve trend, and those who adopted AI are leading in the chart. We are trying to be specific, provide sample data, proof of concept, and actually show the real numbers.

What are the risks of AI technology?

In my opinion, it’s unmonitored fragility and there’s a need to introduce critical processes because AI is stupid and narrow.

Also, I think bias in AI is a risk and there’s a lack of transparency. I’m afraid we’ll screw up this very powerful technology in the wrong way. It feels like launching rockets into space with no direction. The consequences of unwise AI technology implementation could be big system failures: Big companies could be hurt.

But as much as it is bad, it can also be very good. When you automate certain processes that are critical for your business, you expect them to run. It’s very risky to end up in the situation where you create weak points and you have to monitor them. So it’s all hitched around putting the proper processes in place, watching what AI is doing, and scaling.

What will the industry look like in 5 years?

We will go through full replacements of systems in enterprises. Not all of them hear that, but it will happen. Most systems will become obsolete. The technology will take care of things by being human driven, it will take care of automation, and enterprises will no longer need the current systems that maintain their processes. AI hacks will grow and everybody will be right in the middle of it. They’ll say, “Oh, holy shit, we should do something with this now.” Now they are in the beginning and don’t fully comprehend what is happening.

But AI will not move as fast as people think, it takes time to create the right applications. Deep RL will code information and every single aspect of an enterprise. At some point there will be a new programming language, a new way to code, and new software thanks to AI.

What are the biggest limitations to your growth and the AI related market in general?

I would say the lack of talent in the AI field that’s necessary for growth. We are growing as fast as we can and are now around 200 people. We have 7–8 full-time researchers, about 20 university professors that work for us part-time, and students who work on our problems in labs; but access to talent is still a major limitation.

How can the talent problem be solved?

There’s no silver bullet, you know? There are toolsets, libraries, open source, and trainings to take into account, but it will take time. This is nothing new. It’s been around since 2012, which leads to tech giants strategizing to invest in every startup that does AI (investing in anyone that has something to do with AI or Machine Learning). They are spending millions of dollars: In China the valuation of a company is $5 million US dollars per legitimate AI researcher. This is insane! The situation in North America is not so crazy.

An incubator is also operating within your company. What is the goal of that venture?

An incubator is complementing the product. We’ve developed a great business network, so we can help startups build expertise and a quality product and help them get access to Fortune 2000 companies. There is no specific application process. Currently, companies are reaching out to us, but we have slowed down the incubation process as it’s not our priority right now. We are working with a few companies and want to make sure this approach is successful and that it works. We are trying to be a good partner for those we are already working with, but we’re not looking for new startups to work with. Overall, as a company we are balancing between growth, execution, and restructuring, which needs more attention. That’s why the incubator part is in testing mode.

What is it like being an entrepreneur?

Being an entrepreneur is hell. But you get a lot of freedom, and it can be a fulfilling adventure. You learn a lot about yourself and about people, but it’s very hard. Every single one of your weaknesses is highlighted and exposed, and you hear about it.