It’s early December, and the Palais des Congrès de Montreal is buzzing with thousands of smart people. Officially, the event is known as the 32nd Neural Information Processing Systems conference. Unofficially, it is the world’s largest gathering of artificial intelligence experts.
For Jean-François Gagné, the co-founder and chief executive officer of Element AI Inc., the conference is a coming-out party of sorts. In fact, he is determined to make it hard to miss the name and brand of his Montreal-based artificial intelligence software company. At Pierre Elliott Trudeau International Airport, the company has persuaded the airport authority to superimpose “AI” in Element’s corporate purple over the terminal’s “Montreal” sign. When the conference’s 8,000 guests arrive at their hotels, many are welcomed by greeters in Element AI T-shirts. The company is throwing a giant party in Old Montreal, and Mr. Gagné is hosting Prime Minister Justin Trudeau for a G7 event at its headquarters.
The CEO speaks of all this as though he’s defending a castle: “This conference is on our turf and there’s no way we’ll let all the foreign organizations that are already invading from all fronts across Canada just own this place,” says the perennially unshaven Mr. Gagné, wearing a suit jacket, Element T-shirt, jeans and sneakers.
Think big, spend big, and be loud about it: that might as well be Mr. Gagné’s motto. The Canadian technology industry is in the midst of a boom and is teeming with AI startups, but few have launched with a bigger splash than Element, a software company that is not yet three years old. The company launched in October, 2016, with a big vision and a global superstar on its founding team: Yoshua Bengio, the University of Montreal professor known as one of the godfathers of “deep learning,” the foundational science behind today’s AI revolution.
Element set out to build a Canadian AI company to rival the world’s biggest tech giants. “The dream was, ‘The next Google is going to be Canadian,’ ” Dr. Bengio said in an interview.
Big names brought their chequebooks: Microsoft Corp., Intel Corp., Nvidia Corp., Tencent Holdings Ltd., mutual fund giant Fidelity Investments, plus a Korean conglomerate, a Singaporean sovereign wealth fund and several Canadian investors, provided US$102-million in financing in spring, 2017. It was an unheard-of amount for a brand-new Canadian company and enabled Element to hire 500 employees including 100 PhDs. “Clearly I’ve been a factor in attracting capital, attracting talent," Dr. Bengio said. “But it’s not only that.”
Element has positioned itself as a global business star and thought leader on emerging issues around AI. The startup called itself “an unprecedented Quebec success story” in a submission to Quebec Finance Minister Eric Girard this year and attracted generally flattering media coverage. A 2018 headline from Fortune magazine asked, “Can This Startup Break Big Tech’s Hold on AI?” and Fast Company in May named Mr. Gagné one of the world’s 100 most creative people in business.
Element has also become the self-appointed representative of Canada’s AI sector, lobbying federal politicians and government officials 85 times since the start of 2018 – startups rarely do direct lobbying – securing $5-million in federal aid and landing two photo-ops with the Prime Minister.
And there’s more money on the way. Sources familiar with the company’s plans say it is close to raising another US$100-million to US$250-million from a group of investors that includes pension fund manager Caisse de dépôt et placement du Québec, the Quebec government and a global consulting firm. Early backers already provided tens of millions of dollars in capital this spring.
But some are not buying into the hype. One financier who looked at an investment in Element last year, but passed, said: “I just didn’t believe the story … the math is not clear, the product road map is not clear, the only thing that’s clear is [its strategy to] ‘Hire as many people as we can.’”
The skepticism is widely shared across the fledgling Canadian AI sector. Though no one will speak on the record, several rival AI entrepreneurs complain Element’s rampant hiring has squeezed the market for scarce talent, while its progress is at odds with its claims to be Canada’s premier AI company. Element doesn’t yet have a proven track record of delivering products, said one AI entrepreneur in eastern Canada who questioned whether the company even has a viable business model, and yet it’s draining the AI ecosystem.
In response to its critics, Element has unveiled ambitious product rollout plans: It is developing seven products during its fiscal year ending next Jan. 31, Mr. Gagné says, with three set for general availability by summer’s end.
But Element has experienced setbacks on two of its flagship products in development and shed senior staff charged with taking them to market. Other products promised by the company are still in early stages and unlikely to be released for months. According to Public Services and Procurement Canada, Element as of early this year did not yet qualify among the top level of AI suppliers entitled to bid for government business because it hadn’t delivered at least five successful AI projects. Startups consume cash, but with its big roster of expensive engineering talent, Element is burning through it at an unusually high rate.
Element can attract talent. But can it build a sustainable business?
The stakes are high. AI is infiltrating many industries and will change how people work, travel and get information, and how businesses operate as self-teaching algorithms replace human efforts. It’s still early days in this transformation, but AI is already confronting governments and regulators with issues such as data sovereignty, privacy and built-in machine biases.
For its proponents, Element AI is Canada’s bet to build a giant company and capture some of the value in this country, where many of the big AI breakthroughs happened. “I am tired of Canada acting like the world’s greatest poker player walking into casino after casino with two aces in its hands and folding halfway through,” said Matt Ocko, an Element AI board member and managing partner at Data Collective, a Silicon Valley firm that was an early investor in the company. “I want to see Canada have a massive … win.”
Canada already risks falling behind: Chinese rivals Megvii Technology Ltd., and SenseTime Group Ltd. have already raised more than $1-billion each. Mr. Gagné knows that, if he wants Element to be the flag-carrier for Canada’s AI industry, he still has a lot to prove. “It’s really the year of execution for us,” he said.
Element’s beginnings date to 2015, sparked in part by a conversation Mr. Gagné had with a troubled Dr. Bengio. Mr. Gagné had just taken a job as entrepreneur-in-residence with a Montreal venture capital firm, Real Ventures, which had backed his previous tech startup, optimization software developer Planora Inc. Mr. Gagne was responsible for developing Real’s strategy to invest in the fledgling AI sector.
Meanwhile, his friend Dr. Bengio was getting worried about the state of the AI field he had helped create. US tech giants such as Google, Apple and Facebook were realizing the commercial potential of AI breakthroughs made by Canadian academics and their students, and hiring as much academic talent as possible, including fellow deep learning pioneer Geoff Hinton, a professor at University of Toronto.
Dr. Bengio suggested they create a company that would anchor a thriving domestic AI sector and help Canadian universities retain their academics. Their first plan, seeded by Real, was to build a core AI platform and fund an incubator that could spin off startups to deliver machine-learning tools to businesses that could perform functions such as reading and processing insurance applications or detecting and responding to anomalies on production lines. Key to that plan was quickly amassing top-tier AI talent. “Think about how much money is being invested in China and in Silicon Valley," said Dr. Bengio. “There’s no way we’re going to succeed if we don’t push on the gas as much as is reasonable.”
To Mr. Gagné, that meant aiming larger than the typical“lean startup” approach by young technology companies of stretching scant resources to get their first products to market. “We knew it was not going to happen by scraping and bootstrapping your way and getting one solution and hiring a few folks” at a time.
He and his co-founders, including his spouse Anne Martel, research scientist Nicolas Chapados, and Real partner Jean Sébastien Cournoyer, felt their company had to get big, fast.
Element’s founders aimed to establish their so-called “supercredibility,” a notion championed by visionary Silicon Valley entrepreneur Peter Diamandis in his 2015 book, Bold: How to Go Big, Create Wealth and Impact the World. Mr. Diamandis implored entrepreneurs to be brash and unapologetic, acting like they had already succeeded. Some of the basic principles of supercredibility include: “Start at the top, and build your way up"; “When forced to compromise, ask for more"; “If you can’t win, change the rules"; and, “If you can’t change the rules, then ignore them.”
“We really wanted to make a dent,” said Mr. Gagné. “The notion of impact, from the start, was one of the criteria that we wanted to make sure we had.”
When Element launched, Dr. Bengio was front and centre, though his time spent on the company would be limited to an advisory role and board membership due to his other commitments. “I am a very busy person,” he said.
But the timing was perfect. The market was growing aware of the vast potential for AI to help companies derive new insights and make predictions by mining vast troves of their data. Money poured into the sector. Against that backdrop, Dr. Bengio’s involvement anointed Element with supercredibility. Element was flooded with CVs from AI scientists and seasoned executives and more than 20 requests a day from potential customers “from literally every corner of the globe” asking how they could use AI in their business, said Naomi Goldapple, one of Element’s earliest hires. (Ms. Goldapple spoke to the Globe last December when she was a director of industry solutions. She recently left the company and declined to comment further.)
Weeks later, Microsoft announced it was investing an undisclosed amount in Element, one of its first venture deals in AI. Calls followed from other prospective investors, including Data Collective’s Mr. Ocko. His company was “looking for the strongest independent concentration of self-sustaining AI talent that could actually drive product outcomes for global companies. Element AI met that set of criteria,” he said. It wasn’t a Chinese company subject to state influence, nor within the gravitational pull of AI-savvy, data-rich Silicon Valley giants that other corporations might be loath to work with. That was “reinforced for us by Yoshua Bengio’s position as co-founder and his unambiguous moral stature as the only truly prominent AI researcher who had not sold out,” Mr. Ocko said.
The founding team set out to raise US$40-million in venture capital, a sizable sum for a fledgling Canadian company. Mr. Ocko encouraged them to think bigger and raise enough “to be a global champion for Canada and a magnet for capital and talent to Canada.” The founders decided to “be as ambitious as we can [and] take the money,” Mr. Cournoyer said in 2017. The ensuing financing announced in June, 2017. was one of the largest early-stage rounds for a nascent company in Canada.
But Element’s path forward was fuzzy. It abandoned plans to be a startup factory that spun off AI businesses after determining large corporations weren’t yet ready to deal with a slew of AI startups and still needed to wrap their heads around what AI could do for them. Instead Element decided to build those businesses in-house. It dabbled with different ways of helping corporate customers devise AI implementation strategies; most of its work for the first two years involved consulting to Canadian and multinational companies to address specific problems using AI.
The strategic shifts and consulting focus left observers wondering why investors would shell out so much for a company that did custom piecework, since consultants typically generate lower margins and command lower valuations than software sellers. “We created a lot of confusion in the market,” Mr. Gagné acknowledges.
But Mr. Ocko insists the plan was always that Element would evolve into a product company. “That was going to take some time" for Element to experiment and build out AI-based products that met customer needs, Mr. Ocko said.
In early 2018 Element decided to focus on two sectors: financial services and supply chain/logistics. Those target industries were rife with data, repetitive and paper-based processes and deep-pocketed companies willing to embrace AI, making them ideal targets for standardized AI products. Last July, Mr. Gagné told his staff he didn’t want any proposals going out for work “that didn’t have eventual licence-recurring revenue attached to it for a product.”
Starting in late 2018, Element began to reveal its product strategy and some early customers. It was working with investor GIC Private Ltd., the Singapore sovereign wealth fund, to develop a tool that automatically and frequently rebalanced investment portfolios. Another investor, National Bank of Canada, agreed to work with Element to develop a program that would help cybersecurity operators do their jobs by taking on some of the routine tasks of threat detection. The company signed up Gore Mutual, a property and casualty insurer in Cambridge, Ont.,and financial services giant HSBC to develop products for their industries. But it won’t be until 2020 that the first products are ready to go to market.
Element has been quiet about its financial performance but a confidential document prepared for prospective investors last year and obtained by the Globe offers a rare glimpse.
As of June, 2018, the company had generated $4.7-million in revenue (primarily from consulting projects) for 19 customers including Maple Leaf, Barrick, L’Oreal and Hyundai. Sources briefed on Element’s financial performance say revenue in the fiscal year ended Jan. 31, 2019, totalled less than $10-million. Mr. Gagné declined to comment on financial performance except to say 90 per cent of revenues as of last December were for non-recurring business. He added results are on track with plans; Mr. Ocko said Element’s revenues reflected in signed contracts covering future years is “dramatically … larger “ than $10-million.
While that might be an impressive output for many startups, context is important. Software startups typically stretch resources and hire prudently until they build a “minimum viable product” that they take to market. Once they generate sales and market acceptance, they hire engineers and product designers to build out the offering and sales and marketing people to generate revenue.
Element took the reverse approach, snapping up machine-learning research scientists and marketers and amassing a big internal infrastructure before it had products in market. Element “has lots of applied research scientists that think about and develop AI algorithms, but those aren’t the same people that write software and intimately understand a use case and a particular problem you’re trying to solve,” said one recent insider.
As a result, Element is an expensive company to run, even in Montreal where costs are lower than in the Bay Area. And that’s before it has commercially ready products to sell.
Aside from the normal startup accoutrements – staff are treated to a free kombucha and latté bar as well as complimentary lunches, snacks and carbonated water on tap at its Park Ex neighbourhood headquarters – the company also pays for things other startups can’t afford.
That includes a 71-person fundamental research team led by 13 applied research scientists doing basic research whose main purpose is producing work “to have academic impact,” said Dr. Chapados, Element’s chief science officer. “Normally, investing in fundamental research is the stuff of very large corporations,” he acknowledges. (Element has filed more than 50 patent applications.)
The company also compensates 24 outside AI academic “fellows” to provide occasional advice or feedback when needed. It maintains offices in Toronto, Seoul, Singapore and has a 20-person London, U.K., operation that does pro bono work to deliver “AI for good”-- a keen interest of Dr. Bengio’s. Element also advises a Korean AI venture fund and partners with Singapore institutions to help develop their local startup sector.
Element does extensive lobbying (it hired a head of public policy and government relations early last year long before it had a chief financial officer) and even has a “brand guru,” also rarities for early-stage companies.
Element has so many employees that by early 2019 it was no longer defined as a “small business,” meaning it qualified for significantly lower research and development tax credits than companies with fewer than 500 people. (By comparison Shopify Inc. only reached the 500-employee mark the year it surpassed US$100 million in revenues.) That prompted Element to ask Quebec’s Finance Minister to reconsider the rules, saying it was still a startup “in hypergrowth mode.”
Outsiders estimate Element’s “burn rate” exceeds $5-million a month. The burn “is pretty high,” said Ms. Martel, Element’s senior vice-president of operations, without confirming the amount. “It’s higher than a typical startup, definitely.” Said Mr. Gagné: “If burn equals risk, then yes, we’re taking more risk … It’s an ambitious play, we never shy away from saying that.”
Anchored by the belief that it needed to offer a full range of AI tools, and not just a single product to corporate customers, Element set out to launch seven products this year. “Yes, we could sell one product at a time but we wanted to sell very quickly a suite of products to be able to penetrate larger organizations … that can help [solve] many of their problems,” Ms. Martel said, calling it a “contrarian, bold [strategy that] makes sense. Our expectation is that the revenue will also follow once we’re in place in all these organizations."
With the market for AI solutions still nascent, Mr. Gagné acknowledged that the revenue potential of its offerings isn’t clear. Selling to large corporations typically takes a long time, and it’s hard enough for most startup to build a business around a single product, let alone seven.
Element has already had some product setbacks. In January, Element staff told The Globe and Mail during a product demonstration at its headquarters that its cybersecurity project with National Bank was in the advanced prototype stage and close to being sellable once it had been “trained” on National Bank data. But within a few months, The Globe has learned, the company quietly shelved the flagship project this past spring and parted ways with many of the business leaders hired to take the product to market. Element declined to comment on the status of the project and referred questions to the bank, which also declined to comment.
Element has also had some challenges developing a program to help the clogged Port of Montreal to develop software that would ease congestion by predicting wait times for truck drivers delivering or picking up loads. The predictions, available to truckers through the port’s smartphone app, would come from mining data sources, including vessel arrival times, train manoeuvres and weather.
Element has worked with the port authority since early 2017 and progress on the client side has been slower than expected, Ms. Goldapple said in January. When the port expanded operating hours last fall, that threw things off for months because Element’s algorithms were relying on historical data; testing is still ongoing and will only conclude later this year.
The port authority’s director of information technology, Serge Montpetit, said the product had the potential to make his operation more effective. “We want to be a smart port,” he said. “We’re creating history here, we’re at the beginning of something.”
But is this a smart product for Element AI? The company hasn’t worked out pricing yet but the cost would have to come out of the port’s information technology budget, which is less than $10 million and already covers everything from business software to cellphones and computers. And it’s not clear whether the product can be sold in a standardized form to other customers – ports are typically distinctive, depending on geography, local dynamics and differing operating and ownership models.
Asked how much he thought Element’s product would have to be altered for each customer, Mr. Montpetit replied “my guess is 30 per cent” – a high amount that would hurt margins because of the work required with each sale. Element insiders also expressed doubt about the size of the potential market for the port product, which Mr. Gagné acknowledged would be highly customized and sold to operators that “don’t have a super large IT budget.”
The company’s foray into financial services may be more promising. Gore is Element’s lead customer on a product that automates the intake and uploading of data from insurance applications for home, automobile and commercial insurance policies. Gore’s chief information officer Sean Christie said the software will save the insurer money, increase its efficiency, reduce errors and provide the opportunity to capture and use more information from application forms. “It’s a slam-dunk business case for us. If they can price the product in the right space, they will be very successful.”
But with a pilot project under way, Element will only be ready to market the product near year’s end. The project with GIC is in the early stages and Element hasn’t yet started to deploy the product with the Singapore fund. And while HSBC’s head of transformation for global banking and markets, Chuck Texeira, said the UK-based bank picked Element AI after “scour[ing] the world to find the best AI firms," it won’t be until 2020 that the parties determine what products they will pursue together.
Nearly three years in, Element is still a work in progress. Then again, the entire AI ecosystem is. With deep-pocketed backers, Element still has the means to improve on its outcomes to date, and some observers in the Montreal investment community expect it will sharpen its focus and make staffing changes necessary to commercialize products in the months ahead with a greater sense of urgency.
“Element AI is not a perfect company,” said Mr. Ocko. “They are still for all intents and purposes a startup. They have a long way to go to be a global export champion for Canada … They will make mistakes. They have made mistakes.” But Mr. Ocko says he’s “very happy” with the company’s progress toward commercialization. "Anyone who had a bad attitude about the company is going to be gravely disappointed.”