Skip to main content

Has this man discovered the formula for creating high-tech superstars? Sean Silcoff goes inside a bold experiment at the University of Toronto to find out

Cian O'Sullivan is the picture of confidence as he speaks to a packed lecture hall at the University of Toronto's Rotman School of Management.

A tall, lantern-jawed man, Mr. O'Sullivan begins his presentation with a slick video explaining the genius idea developed by Beagle Inc., his Kitchener, Ont.-based startup. Using artificial intelligence (AI), Beagle's software reviews dense contracts and other legal documents and roots out critical information in minutes. Most of the 30 million small- and medium-sized businesses in North America don't use lawyers for commercial deals; Beagle's product saves them hours of aggravation.

The founder throws out some impressive numbers and names. Already, it has 750 customers and prospective investors have shown "an immense amount of interest," he tells the Rotman crowd. So have Volkswagen AG and Thomson Reuters Corp.; Mr. O'Sullivan is spending 20 per cent of his time working on these new giant customer leads. Everything seems to be going Mr. O'Sullivan's way on this mid-April, 2016, day – that is, until the audience starts barking questions.

John Harris, a retired steel magnate, asks why Beagle only charges $83 per month. If it works, that's too low, he says.

Barney Pell, a Silicon Valley entrepreneur and investor who once led U.S. National Aeronautics and Space Administration's AI branch, says Mr. O'Sullivan should raise much more than the $400,000 (U.S.) he's seeking or investors won't take him seriously. "Right now you're in no-man's land," Mr. Pell says.

Mike Serbinis, one of Canada's most successful tech entrepreneurs, says Beagle must decide if it's going after small or large customers, but not both. Others agree: Beagle has a focus problem. "I just want to know what kind of business this is," says Shivon Zilis, a partner with San Francisco venture-capital firm Bloomberg Beta.

Watching the scene unfold is the man who designed this interrogation chamber for entrepreneurs: Ajay Agrawal, one of Canada's foremost business academics and founder of Creative Destruction Lab (CDL), a ground-breaking program for startups housed at Rotman.

CDL has accomplished what few incubators of technology startups in Canada have managed to do, already making a major impact on a teeming startup ecosystem. Now in its fifth year, it has developed a promising method for helping early stage companies- many of them using AI and other leading-edge technology – to grow, flourish and attract private capital.

Mr. Agrawal's aspirations go far beyond that, however. He also hopes that what the CDL team has created can point the way to a solution for a problem that has vexed our economy for decades: our sputtering ecosystem for innovation. Canada is blessed with excellent research universities and a strong pool of engineering and technical talent. Where we often stumble is in the process of turning bright ideas into cash, and spinning out technology businesses that can grow to become global players. Since Nortel Networks Corp. went bankrupt and BlackBerry Ltd. flamed out, there has been a shortage of truly large anchor companies in the Canadian tech sector. That will have to change if Canada is going to have an economy built on the jobs of the future.

With the federal government set to unveil what has been described as an "innovation budget" next week, perhaps CDL's most impressive accomplishment is that it isn't part of a grand public-sector innovation strategy, nor based on a recommendation from one of the many past reports on improving Canada's spotty record for commercializing technology. Rather, CDL has acted like a startup itself, driven by ingenuity and purpose and managed by a skeleton staff on a shoestring budget mostly financed by the private sector and individuals .

At the time of Mr. O'Sullivan's presentation last spring, Beagle was one of 13 AI firms left in CDL's machine-learning program, down from 25 that started the previous fall. All were trying to build new businesses around algorithms that can learn on their own, rather than just perform tasks programmed by humans.

To stay in the program, Mr. O'Sullivan and the other fledgling entrepreneurs had to convince at least one of the "mentors" – accomplished businesspeople who act as advisers, objective setters and judges – that it's worth devoting four hours out of their busy schedules to Beagle over the next eight weeks. If none are willing, participants are booted out of the program.

"When they come here most [CDL companies] are science projects," Mr. Agrawal says following the session. "By the time we get to this meeting" – the fourth day-long session out of five over the course of nine months – "they are ready to raise a seed round of capital." That explains why the 60-odd observers in the room include several San Francisco-area venture capitalists. Their presence "is an important indicator" of CDL's success, said founding partner Dan Debow. "It tells you they're not coming to do you a favour, they're coming to do themselves a favour."

Some liken CDL to a cross between Survivor and Dragon's Den because most firms that start get cut. That comparison irks the cerebral and serious Mr. Agrawal. Dragon's Den "rewards the ability to pitch," Mr. Agrawal said. "CDL rewards an ability to perform … our ratings are the success of these companies."

If so, CDL is a huge hit. Mr. Agrawal's initial goal was to see CDL alumni create a combined $50-million in equity value after five years. That's how he sold the program to early benefactors, including Mr. Debow and Mr. Harris, tech entrepreneurs Michael and Richard Hyatt and ex-Research In Motion Ltd. chief financial officer Dennis Kavelman, who each donated $300,000 to get CDL started.

CDL has already surpassed that goal – 20 times over. Leading the way is Thalmic Labs Inc. of Waterloo, Ont., a maker of wearable technology that was one of the first companies to go through the lab in 2012-13. Thalmic raised $120-million (U.S.) last year from Intel Capital, Amazon's Alexa Fund and others. Much of Thalmic's early guidance and funding came from CDL mentors.

Even without Thalmic, CDL would be a success: Other CDL graduate firms are valued at about $500-million combined and sell such eclectic innovations as anti-frost coatings, low-energy display lights and speech-analysis software that detects the onset of dementia. It's still too early to know if any will break out and emerge into significant enterprises; startups typically need seven to 10 years to reach that point.

Now, like a startup entrepreneur, Mr. Agrawal wants CDL to build on its early success and expand into something significant. This year, the program welcomed 100 companies – half of them focused on machine-learning. It expanded to the University of British Columbia, where "Creative Destruction Lab West" boasts early Yahoo president Jeff Mallett as a mentor. It has fielded interest from Dalhousie University in Halifax, the University of Calgary's Haskayne School of Business and New York University to host similar offshoots. Companies have flown in from France, Israel, Boston, New York and even Silicon Valley to participate.

But there's no time to waste: Canada's early lead in the now hot area of machine learning, a type of AI, has eroded somewhat as U.S. giants swooped to snap up leading Canadian scientists and startups. Google, Microsoft and their ilk have set themselves up to be big beneficiaries of Canada's AI talent pool.

Mr. Agrawal says now is the time to act on another field where Canada has built an early lead before the industry giants clean up: quantum computing. It's an emerging field based on a still developing technology that harnesses the power of subatomic particles. Quantum computers are expected to one day be significantly faster than the most sophisticated machines on the planet, enabling researchers to solve complex problems beyond current capabilities.

If things break as Mr. Agrawal hopes, Canada could hold that lead. To help, he's set this spring to announce a CDL stream exclusively centred on companies developing quantum computing-based AI companies.

"Our mission is that by 2020 the quantum-machine-learning initiative will have produced more, well-capitalized, revenue-generating quantum machine learning software companies than the rest of the world combined, and majority will be based in Canada," Mr. Agrawal says. "We have a model that works for building companies, but we are still in catchup mode."

Ajay Agrawal at a Creative Destruction Lab session.

Doubling down on global leaders

The idea for Creative Destruction Lab came to Mr. Agrawal during a dinner in Toronto about five years ago. The topic that evening was what Canada could do to improve its woeful record for harnessing Canadian ingenuity to stimulate economic growth.

This was familiar ground for Mr. Agrawal, now 47, an expert on innovation economics and one of Rotman's most popular professors; he'd attended an "endless number" of government-sponsored panels, round tables and meetings on the subject during his career, and many of the usual suspects were in attendance, including CEOs of some of Canada's largest companies.

But something bothered him that night. Everyone at the table "was very deferential" to the CEOs "who were talking about how important innovation was," he said. "But they hadn't really done anything that's innovative."

Rather, it was the scientists in attendance who were the true innovators, including regenerative medicine expert Molly Shoichet, quantum computing and robotics visionary Geordie Rose, and Steve Mann, a pioneer in wearable computers. "I thought to myself, 'In this room … everyone is focused on the bank CEOs, but as soon as we go global, they're not really that relevant – and the other people here are.' A light went off and I realized, 'OK, we need to double down on the people who are global leaders and they're not the ones everyone else in the country is focused on.'"

That lightbulb moment had been 20 years in the making. Over the course of his academic research Mr. Agrawal, who studied engineering before earning his PhD in strategy and economics at UBC had focused on why most universities struggle to help their research scientists achieve commercial success from their inventions.

That was a particularly acute problem in Canada. This country's 36 leading research universities earned a combined $62-million (Cdn) in licensing income from campus inventions in 2015, according to the Association of University Technology Managers, a paltry fraction of the billions of dollars in on-campus research expenditures annually. By comparison, U.S. commercialization powerhouses Stanford University in Palo Alto, CA. and the Massachusetts Institute of Technology alone last year brought in $94.2-million (U.S.) and $62-million in licensing income, respectively.

It wasn't that Canadian scientists were less inventive than their U.S. counterparts; in a 2008 paper, Mr. Agrawal found U of T scientists disclosed as many inventions per capita as those at MIT. However, the "technology transfer" system whereby Canada's top research university helped campus inventors commercialize their research breakthroughs was less efficient than MIT's as there were fewer patent applications, issued patents and licensing income, he wrote. When Mr. Agrawal broadened the study to 160 Canadian and U.S. universities, he found a similar gap between the two countries.

"In Canada we have a lot of important technical breakthroughs," Mr. Agrawal said. "But we have comparatively very little capital flowing in to commercialize those. That begs the question, 'Why?'"

He rejected conventional wisdom that Silicon Valley had better ideas, smarter people and more money. The main reason for the gap, he believed from his years of research, was what he described as a lack of good judgment in Canada – judgment that was essential to help scientists turn into successful entrepreneurs. That key ingredient was in abundance in Silicon Valley's established technology commercialization ecosystem.

Mr. Agrawal cited a compelling example to make his point.

A 1997 e-mail in Stanford's archives shows a campus graduate student working through a potential deal to license technology he has developed to early Internet company Excite. The student writes to Stanford's technology transfer office that he is willing to sign over rights for Excite to use his technology in exchange for a salary. He figures his time is worth $100,000 a year.But his e-mail also reveals someone named "Vinod" has advised him to instead buy a company that would own the technology. The "punchline," Mr. Agrawal said, is that the technology in question was Google's search algorithm. The student was Larry Page, who co-founded Google a year later. "Vinod" was venture capitalist Vinod Khosla, whose early backing helped to turn Google (now Alphabet Inc.) into the world's second most valuable public company.

"The most remarkable thing about [Mr. Page's] e-mail is how unremarkable it is," Mr. Agrawal said. "There's not an ounce of genius there. That could have been from a computer science student [anywhere in Canada]. In my view, the only reason this turned into a company the way it did is because of the judgment from people like Vinod."

Canadian venture capitalist Haig Farris, centre, provides feedback at a Creative Destruction Lab session.

The market for judgment

Mr. Agrawal started to develop an idea he called the "market for judgment." He wanted to bring together some of the sharpest and shrewdest investors to advise cutting-edge Canadian research scientists and help them build companies around their technology breakthroughs, providing them with the kind of business guidance Mr. Khosla had shared with Google's co-founder.

"The problem is if you are a person who needs judgment – like a PhD in computer science who's come up with a new way to organize the world's information online. You can't go down to Bay Street and buy five units of judgment. It's not for sale. And therein lies the problem. [Creating that market] became the essence of the Creative Destruction Lab."

The academic began developing the idea with a small group that included his former UBC entrepreneurship professor and pioneering Canadian venture capitalist Haig Farris, the late Osler, Harcourt & Hoskin LLP technology lawyer Geoff Taber, and Dan Debow, a successful Toronto tech entrepreneur and Rotman alumnus that Mr. Agrawal had met through Next 36, a program the professor had helped create to expose bright Canadian undergraduate students to entrepreneurs. Mr. Debow initially told him the world didn't need another accelerator, to which Mr. Agrawal replied, "'Well, I don't want to build that, I want to build something different,'" Mr. Debow recalled.

Unlike most accelerators and incubators, CDL wouldn't offer companies free office space nor take an equity stake. This was meant to be a purely altruistic venture "that emphasized something very specific: the judgment mechanism," Mr. Debow said. Mr. Agrawal also had a distinct idea how he program's success would be measured: not by number of jobs or startups generated – a typical yardstick – but equity value created by CDL companies as they raised money. "That's how investors measure the success or failure of their investments," said Mr. Debow.

To build one side of the judgment market, Mr. Agrawal felt it was crucial to take a different approach to mentorship. While mentors are a common feature of incubators and accelerators, "they generally fall into two categories," said Mr. Agrawal: people who were "well-meaning but haven't really built a significant business," or "the occasional superstar mentor" who was typically too busy to provide much help. "We wanted judgment that was neither mediocre-and-highly committed, nor superstar-but-not-committed. We wanted superstars that were committed."

Mr. Agrawal set out to assemble a stellar group of seven entrepreneurs – the G7, as he called them – who had built successful businesses from scratch through to a successful sale or public offering, and who would be willing to commit five full days to CDL, spread out over the program's nine months. "Time is the most valuable thing they have," he said.

He was also wanted them to put skin in the game by investing in CDL companies they liked, though he made a handful of exceptions for entrepreneurs whose companies hadn't yet paid out from a sale or IPO, such as Geordie Rose, a former UBC classmate who had gone on to co-found quantum-computer maker D-Wave Systems.

The G7 would be matched with an initial annual cohort of 23 early stage companies, sourced from mentors, benefactors, the Next 36, other accelerators and, most of all, university science labs. When they met, the G7 would collectively determine three short-term goals judged to be the most likely milestones to advance each CDL startup. The firms would then have eight weeks to deliver. If they couldn't they would be out.

If they did, they'd get three more objectives to meet eight weeks later. At any meeting, the G7 could invest – or vote them out. Every meeting, at least one company would be automatically cut. "We knew half [of the CDL startups] probably wouldn't make it through," said Mr. Agrawal, noting yet another difference from other startup assistance programs. "It turned out to be less than that."

Inside the workspace at Thalmic Labs in 2013.

Thalmic for the win

Mr. Debow, Wind Mobile founder Tony Lacavera and the Hyatt brothers were among those who signed up as mentors and financial supporters. Mr. Agrawal had further success wooing other mentors through Canada's tight-knit tech sector, including Mr. Serbinis and Calgary entrepreneur Chen Fong, who were sold on his pitch that they could make a meaningful contribution to Canada. Many were impressed by the company they'd be keeping. "I say no to almost everything," said Ted Livingston, co-founder of popular Waterloo-based mobile instant messaging service Kik and an early G7 member. "I just looked at the entrepreneurs [Mr. Agrawal assembled] and was like, 'Wow, yes. I would be honoured.' I thought this was one of the highest densities of great people I've ever seen."

But when he asked U of T and Rotman for help, Mr. Agrawal said "we didn't get a nickel." The extent of Rotman's support was allowing him to book rooms in the building. "This was really happening despite the university. We weren't really doing anything that fit within the university mould," said Mr. Agrawal. (To date, CDL has received $2.5-million from individuals, $900,000 from companies, and $400,000 from the publicly funded Ontario Centres of Excellence)

The plan came together within months and the first CDL cohort started in the fall of 2012. Almost immediately, several of the G7, including Mr. Debow, got excited about Thalmic, which just months earlier had been a mechatronics class project at University of Waterloo to make a gesture-control armband guided by electrical signals of a user's muscles.

That December Thalmic closed a $1.1-million seed financing round. "A big chunk of that came from the G7 and others they introduced us to," Thalmic CEO Stephen Lake said. "One of the first value-adds was that not only were we getting advice from people in the room but they actually were writing cheques and becoming investors." The following June, Thalmic raised another $14.5-million; Boston's Spark Capital, an early backer of Slack and Twitter, and Intel Capital led the deal. "I feel comfortable saying we can put [Thalmic] in the win column," said Mr. Debow, who joined the board.

The Creative Destruction Lab in March, 2017.

Just-in-Time MBA

Far from celebrating CDL's early successes, however, Mr. Agrawal took the approach of an entrepreneur, tinkering to make it better. He recruited world-renowned U of T professors including Ms. Shoichet and Mr. Mann to act as CDL "chief scientists," testing technology brought in by the startups. He asked professional services firms including Osler and Ernst & Young Global Ltd. to advise the startups. Additional observers, dubbed "G7 Associates," were invited to participate. These carefully curated guests – "we look for community builders, not lone wolves," Mr. Agrawal said – included venture capitalists such as Mr. Farris and successful entrepreneurs who were expected to provide feedback, introductions and even funding to the startups.

He also changed how the G7 weeded out CDL companies. At first the selection was done by a show of hands after each meeting. Mr. Agrawal didn't like that, so the G7 was instead asked if they'd be willing to devote four hours of their time to each startup over the following eight weeks. If just one put up a hand, the startup would stay. "The best voting mechanism is individual conviction," Mr. Agrawal said. Committing four hours "makes it expensive for them to raise their hands."

That proved to be the saving grace for some. Mr. Livingston took to two founders in CDL's second year who were developing a robot tea dispenser called teaBOT. Their rough prototype looked like a high school science project but Mr. Livingston was intrigued by the team: one was a U of T aerospace robotics engineering PhD candidate who loved to build machines, and the Kik founder wanted to see what they could accomplish. "I just believed in these people," Mr. Livingston said. He alone voted to keep them in initially. At each successive meeting the prototype and pitch improved, and several G7s ultimately invested. There are now 17 teaBOTs in service, including machines in three "365 by Whole Foods" stores.

The biggest change came at the behest of steel magnate John Harris, a founding donor and regular observer of CDL meetings. He was learning from the CDL sessions himself and said "It's crazy this is happening right here in a business school and there is not a single MBA student in the room," Mr. Agrawal recalled.

Mr. Agrawal and a PhD student designed a self-directed course for CDL's second year whereby students did standard MBA analysis, observing the startup-G7 interactions and producing slides with prim recommendations. Mr. Debow called it "a lot of wasted effort, a classic MBA make-work project. Their insights were not that helpful."

But there was such a surge in demand from students and inquiries to admissions about CDL the following year that Rotman administrators took notice.

That coincided with the arrival of new dean, former Bank of Canada deputy governor Tiff Macklem in July, 2014. After giving many high-level speeches over the years about Canada's chronically stagnant productivity, Mr. Macklem quickly embraced CDL for the impact it was having. The dean has since thrown Rotman's support behind the program and the launch in 2015 of an annual CDL machine learning conference, promoting the lab publicly and meetings with government and business. "CDL is very results-oriented but it's also an experiment in how you teach entrepreneurship and how you actually accelerate companies," Mr. Macklem said. "It's working, and there are not enough examples of accelerators that are truly working."

To improve the MBA class in the third year, Mr. Agrawal decided to get students directly involved by working with the startups. Now young companies led by scientists with limited business know-how had access to "this team of super motivated, super engaged, highly talented MBAs working for them," Mr. Debow said.

Forget reading stale Harvard Business School case studies: students were now living real-time case studies. "They're not doing an academic exercise, they're taking their academic tools and applying it to a real problem – and it's a problem the founders and the G7 care about," Mr. Agrawal said. Meanwhile their professors were expected to write class notes based on the latest challenges CDL startups faced, not just wheeling out well-worn lectures. One visiting academic from Harvard called it "just-in-time education." The CDL course became "the most competitive at Rotman to get into," said Mr. Agrawal.

"It was one of the most beneficial classes I had," said Sasha Kucharczyk, who took the CDL course as an MBA student in 2013-14. "It taught you to think and act differently [to] help you attack and execute on abstract problems [to get] meaningful results." Two years later he entered his startup Preteckt, which uses AI software to predict when heavy vehicles will need maintenance, into CDL.

Big deals for CDL alumni

Series B funding for Thalmic Labs, makers of the Myo Gesture Control armband (2016)
Series A funding for Bionym, makers of a heartbeat authentication wristband (2014)
Seed funding for Atomwise, developer of a drug-discovery platform (2015)
Seed funding for Kepler Communications, a satellite communications company (2016)
Seed funding for Deep Genomics, a Toronto-based bioinformatics startup (2015)
Seed funding for Taplytics, which specializes in mobile A/B testing (2015)
Seed funding for Bridgit, whose cloud-based platform helps manage construction projects (2016)

All figures in U.S. dollars for single rounds of funding (Source: Company releases)

'World class talent'

It's 8:01 on April 13, 2016, and the fourth meeting of the CDL's "machine learning" stream for the year is under way. Mr. Agrawal, who acts as master of ceremonies, runs a tight ship. He asks everyone in the room to describe themselves in one sentence – and cuts off anyone who lingers.

One by one the 13 startups still in the program get up for their 30-minute slots. Each firm is introduced by a mentor who revisits how they've performed and proposes the next three objectives. Most of the discussion is among the associates and Mr. Agrawal, who pick over the companies as if talking about specimens while their founders stand awkwardly at the front, occasionally responding to pointed questions. "I'm really not sure yet about the company," says Mr. Pell, the former NASA AI head, of Heuritech, a startup from Paris that assesses social media posts using AI to determine fashion trends. "They're smart people," he says of the team of four AI and machine learning PhDs. "My concern is: how big is the business opportunity?"

In 2015-16, for the first time there are two CDL streams, with this one focused exclusively on startups based on "machine learning," a form of AI where algorithms automatically build themselves based on big data troves fed in. (CDL has since doubled down and now runs two machine learning streams, each starting with 25 companies.) The idea was initially met with some skepticism internally that CDL was getting too narrowly focused, but it's worked so far. Two San Francisco-based machine learning experts are mentors, known here as the ML7: Mr Pell, introduced to Mr. Agrawal a year earlier by a CDL alumnus, and Ms. Zilis of VC firm Bloomberg Beta.

The machine-learning focus has also attracted interest from top U.S. venture capital firms, which regularly send partners to observe, including Bessemer Venture Partners, True Ventures, and Google Ventures. "All the companies have been very impressive," says Adam D'Augelli, a True partner. "They have real technology and at this point will have early revenue or early product traction, which for something coming out of university is relatively rare." One Silicon Valley early stage investment firm, FundersClub provides $50,000 to any CDL company that raises money from at least one ML7 and another $100,000. By this point some CDL companies have landed seven-figure venture financings.

"There's world class talent" in Toronto, Mr. Agrawal says. "When we built [a CDL program for machine learning], that was the first time people got on a plane from Silicon Valley to Toronto rather than the other way around. Once they make that investment, they are schlepping to Canada every quarter for a board meeting. Making a second investment is a far lower hurdle. One of the most important things we're doing for the country is getting some of the world's top investors to go from zero to one" in Canada.

As the meetings progress, the ML7 ask who are the startups' customers – the Beagle dilemma – what they should charge, what key positions they should fill and whether their businesses can scale into something big. "If I don't have an understanding for how this could potentially be a billion-dollar company or if I don't believe the founder has that ambition, I categorically can't invest," Ms. Zilis says later.

Mr. Agrawal says there are three main startup risks: technology, market and entrepreneur. Through CDL, he says, entrepreneurs can show their technology works and determine initial market interest – while addressing the third point by showing they can steadily deliver on time-sensitive objectives.

Some companies are doing well. Preteckt seems to be winning over the ML7 with news it has completed its second-generation protoypes, has installed its technology on several vehicles and is working on a letter of intent with Volvo.

"I like everything they are doing," declares ML7 and tech entrepreneur Shahram Tafazoli. The ML7 assigns Preteckt objectives to install 100 revenue-generating units, file a patent application and return with a signed letter from Volvo in the next eight weeks.

Deep Genomics, meanwhile, doesn't have a product yet, but for Silicon Valley investors, that doesn't always matter. What Deep Genomics does have is CEO Brendan Frey, a renowned deep learning pioneer and U of T professor who is trying to apply his lab work to improve diagnostic yields. Like many AI experts who've started companies, he's attracted Big Silicon Valley money at a very early stage, raising $5-million from Bloomberg Beta and others. It's a reminder that south of the border, investors are willing to take larger, more speculative bets than more conservative Canadian VCs, who as a result miss out on the gigantic returns when some of those flyers turn into global giants. He's the rock star of the room, looking cool and confident in a silver paisley shirt and salt-and-pepper beard. "I got on the phone with Brendan and within five minutes I said 'You can have my money and I will help you no matter what,'" Ms Zilis says in an interview later.

Brothers Alex and Eric Dolan have developed a smartwatch app that can detect when a person with epilepsy is having a seizure and alert others. The technology was inspired by their epileptic mom and 2,600 people are testing it. "We are working with people who are scared … giving them tools so they can manage their lives," Alex Dolan tells the room.

But while some ML7s praise the Dolans, they're not convinced their Neutun Labs Inc. is a billion-dollar opportunity. One concern is that the Neutun sends out "false-positive" alarms close to half the time to emergency contacts about attacks that aren't happening. "Can it deliver something that has value? I think it's been nebulous," says ML7 Moe Kermani, a Vancouver venture capitalist.

Alex Dolan isn't backing down. "I've found a high satisfaction among users," he says in an emotionally charged and defensive tone. "I'm thanked profusely for it….we're focused on the one metric that matters – growing our users and engaging with them." Neutun is cut later that day.

"The ML7 just didn't think they could add a lot of value to the company anymore," Mr. Agrawal explains later. He says CDL "can really only help companies that are willing to be coached."

One such company is Validere Technlogies Inc. It entered the non-machine learning CDL program in 2015-16 with technology to sniff out knockoffs of perfume brands. At the behest of the G7, Validere changed its pitch and is now a hazardous waste detection service. Alberta-based G7 Chen Fong introduced the founders to oil patch executives, securing an entree that might have otherwise taken years, and the team raised $3.3-million (U.S.) in 2016 from investors in Canada and Silicon Valley.

Mr. Agrawal is particularly proud of Rotman MBA student Alyssa Randall, who "went rogue" after a machine learning company she favoured was cut early from CDL. She helped the team anyway and thanks to her efforts they have had some breakthroughs – and been accepted back into CDL for this meeting. MBAs "are viewed as suits that are overpriced and underdeliver," Mr. Agrawal says later. Ms. Randall, on the other hand, has taken big risks and overdelivered on a long shot, rather than taking the safer route of being reassigned to a company that made it through. Her company, called Algocian, graduated from CDL and hired her. "I was really happy," she said months later, though Algocian ultimately shut down. "It was an opportunity for me to solidify a positive reputation for the company."

As for Beagle, the company did end up graduating and focusing on larger enterprise customers, and CEO Cian O'Sullivan praised the program as "a fantastic initiative … an absolute perfect crash course for running a company."

That said, there are ways it could improve, he said months later. Not all advice he got from mentors from helpful, and Mr. O'Sullivan said he would have appreciated advice earlier on how difficult it would be "to train the marketplace" for a new type of product rather than attacking an existing market. He wondered if there were too many companies for mentors to keep track of, sometimes diluting the effectiveness of advice he received, and felt the mentors could have been more open to CDL companies airing their vulnerabilities, rather than "looking for strength after strength after strength. Highlighting weakness doesn't fit well in that environment."

Some wonder whether CDL can replicate its success as it expands. "The real challenge is it's all about Ajay," said Boris Wertz, a Vancouver venture capitalist who is one of CDL West's G7s this year. "As long as Ajay stays motivated he can create a real legacy."

CDL now faces the same challenge as any flourishing startup: not getting too cocky about its early success. "CDL's measure of success will ultimately be 10 years in the making,"said ML7 Lisa Shields, who is now mentoring CDL West companies. Says Mr. Debow: "One of the worst Canadian diseases is to convince yourself you're world class. Have we created Facebook yet? No? Then keep moving."

Indeed, Mr. Agrawal, who actively solicits feedback about improving CDL, hardly seems satisfied; he rejected out of hand recent suggestions CDL throw a party to celebrate its alumni hitting the $1-billion value creation mark in value. "We are so far away from achieving the mission" of improving Canada's competitiveness, he said.