Canada’s academic research expertise helped the country become a leader in the AI startup boom. As the ecosystem matures, those early startups — powered by big ideas — are facing the challenge of operationalizing products in the market. This challenge requires industry expertise, but that expertise will still need to be derived from academia’s AI labs.
According to our latest map of Canadian AI players, the ecosystem is becoming leaner as winners and losers emerge. One of our main metrics for ecosystem growth we have looked at has been the number of AI companies — visualizing their logos in a “map” was a way to show to the world (and ourselves) that Canada is a force in the global competition to apply this technology. Now, the growth rate of those logos is flattening as more AI companies leave the market amidst the competition to find product-market fit.
Though, the growing attrition rate does not mean the gross size of the ecosystem is shrinking. Other metrics show that the discovery process of product-market fit is indeed making winners. From 2017 to 2018, Canadian AI companies saw investment more than double to $660 Million, with more of it going to later-stage funding rounds. In the same period, revenue for AI enterprise solutions companies in Canada rose 65% from the year before, and have grown to represent nearly 5% of the enterprise solutions market revenue overall.
Keeping count of the gross total of players in the ecosystem is losing its relevance. The attrition rate of AI companies will continue to rise as winners emerge. To better understand the health and growth of the Canadian AI ecosystem, we will need to take a deeper look at the discovery of product-market fit and how companies succeed or fail at finding it. Below I share what I think will be important to track going forward — I also welcome the community’s ideas for what signals we should be monitoring.
Tracking a More Mature Ecosystem
There are multiple reasons why an AI company may leave the market: some close because they could not find their product-market fit before running out of cash; others pivot out because they could not deliver on the AI promise but had viable products regardless; and still others are acquired and add to the growth of another AI venture.
One connecting theme that needs attention in all of these stories is AI talent. It is a big challenge of the AI ecosystem to prove the potential value of AI’s application, and talent is at the core of that challenge. The particular problems are data, algorithms, and organizational change. Each of these problems require talented people who deeply understand the technology as well as the area of application.
The technical expertise will come from academia, as it always has. Assessing the robustness of this research and training system, and ensuring it is receiving sufficient support, can be started on a relatively trivial basis: number of professors, PhDs, and new programs such as applied masters.
Building applied expertise and passing on critical new developments in the fundamental research will best be achieved through academic labs’ connections to industry through the ecosystem of AI companies and enterprise labs. To be sure, good training experiences can be found in enterprise labs, but the large majority of good talent and scientific progress will still come from academic programs. The opportunities for collaboration are plentiful with at least 70 enterprise AI labs to choose from, though the ecosystem ought to be careful to balance keeping professors on a part-time basis so they can stay teaching. Therefore, directing support and tracking efficacy should be derived from the connections between professors (who remain teaching) and industries.
Making Sure We’re Proving AI’s Potential
As AI applications make their way into everyday use, how can we differentiate companies delivering the full potential of academic progress? It is those organizations that have actually integrated active learning systems into their operations that are maintaining the necessary pipeline between academia and industry. To know Canada is keeping its competitive AI edge, future iterations of the map will need to focus on organizations that are running, at scale, systems that continuously learn and improve from their interactions.
This pool should not include organizations who have downloaded pre-trained models, nor even those that have a lab but still have yet to operationalize systems that learn continuously. These are important supporters of the ecosystem, but represent more of the demand-side of the market. Separately tracking readiness of these organizations (data, infrastructure, willingness, etc.) will be an important metric to understand the health of the ecosystem and the pulling forces that keep it thriving.
Beyond these proposed angles, I invite others to suggest what ought to be tracked to understand the health of Canada’s AI ecosystem to be sure we are keeping our edge and making the best use of our world class research.
Continuing to Provide the Right Support
One other reason for these reports has been to encourage the right kind of support and behavior by showing over time the positive impacts of our collaboration as an ecosystem. One of the biggest issues for us when the world started to recognize the importance of AI was the outpouring of talent from Canada into the US, including professors. Today we’ve nearly eliminated the net-negative growth of AI experts due in part to a huge turnaround in immigration policies (thank you Global Skills Strategy). As well, the surge in funding has done a great deal to create viable work for fresh graduates coming out of our universities. Getting to strong net-positive talent growth will mean persistence in these policies, as well as increasing the amount of people pursuing AI training.
In the last several years, academia and business have come to an important understanding that even as professors take on private work to transfer research into industry, professors should keep one foot in universities to teach the next generation. Again, critical funding has paired well with this shift in mindset: Canada and provincial governments have doubled down on educational and upskilling programs that focus on this field thanks to tax credits and other incentives.
It will take more time to see the results of these changes — PhDs take 5-7 years to complete and aligning policy has its own particular timelines. Persisting in our collaboration and efforts for the long-term health of the ecosystem will mean encouraging students earlier in their education to pursue science and maths, not letting up on temporary immigration policies, and standing by the liberal Canadian values that make this country such an attractive place to work and have a positive impact on the world.
To find this year’s full analysis, go here.