- Canada’s present strength as a global AI leader is in research and technology. However, we need to deliver economic and social benefits from our Canadian investment in AI by moving towards AI adoption in government and business. We need a framework for adoption, including regulations that create certainty in the sector.
- Canadian companies need to focus less on who is using AI and what it looks like in a particular industry, and focus more on how AI changes the way we work and its added value to organizations and their output.
- The primary challenges facing SMEs in adopting AI are financial and human resources. SMEs would benefit by focusing on three things: identifying the value to be gained from AI, developing partnerships within their ecosystem, and financial funding, particularly from government.
Canadian government must do three things to support AI adoption: create certainty through specific AI regulations, split the funding mix to focus on AI implementations rather than solely on fundamental research, and lead by example by employing AI in more governmental operations.
What do you identify as the main strengths and weaknesses of Canada’s AI ecosystem?
Canada’s advantage comes from the power of fundamental research, primarily from three academic institutions in Montreal, Toronto and Edmonton. Vancouver also has an interesting and growing academic environment. In 2017, the federal government invested $125 million towards fundamental research through the Pan-Canadian Artificial Intelligence Strategy. The funding went to these three institutions. Canada is now recognized as having world-class AI research teams, giving us a research advantage based on the levels of investment by the federal government. That is our strength and we need to build on it.
“89% of Canadian executives acknowledge they know how to pilot, but struggle to scale AI across their businesses.”
Our challenge is determining how to deliver economic and social benefits from Canada’s AI investment. Canadian private sector companies, whether large or small, have not adopted AI with any kind of urgency or speed. In fact, a recent global Accenture report titled AI: Built to Scale found that 89% of Canadian executives acknowledge they know how to pilot, but struggle to scale AI across their businesses. Our government has also been tentative about adopting AI solutions and requires a framework to support and encourage the integration of such solutions.
What can Canadian governments and key stakeholders do to incentivize AI adoption that ultimately translates into positive impacts for our SMEs and our overall economy?
This is a really difficult question because the answer is multi-dimensional and involves multiple stakeholders. There is not a one-size-fits-all problem or solution. For instance, the problems large Canadian organizations face with respect to AI adoption are very different from those our SMEs face.
Any framework for adoption, whether it be government led or better yet a public-private collaboration, needs to address methods for creating change in the way organizations accomplish tasks or sets of tasks. For example, I think of AI adoption very differently than I do of AI experimentation, AI proofs of concept, AI pilots or AI minimum viable products (MVPs).
“The goal is to use AI to change the way we work. We need to work smarter, better and more effectively within Canadian organizations to produce economic and social benefits that could not have been achieved without AI.”
AI adoption includes the deployment of solutions into a production environment creating the expectation of new ways of working. It delivers added value to the organization in the form of desired outcomes, including cost optimization, yield, improvements, fraud reduction, and enhanced customer experience. The goal is not just to proliferate AI use and solutions, but, first and foremost, the goal is to use AI to change the way we work. We need to work smarter, better and more effectively within Canadian organizations to produce economic and social benefits that could not have been achieved without AI.Although this sounds like a very long introduction to the answer, I believe people end up focusing too much on who is using AI and what the AI use cases are. If we are not changing the way we work, doing things better and adding value to organizations, then it is not sustainable.
A framework for AI adoption involves two distinct sets of initiatives. I call the first set “enabling initiatives.” This asks what the government and others can do. I think of enabling initiatives as fundamental research. This includes university and college training programs, continuous education for employees, up-scaling employees within companies, and the creation of a viable infrastructure to support a thriving start-up ecosystem. If the goal is to adopt AI technology, then fundamental research is actually an enabling capability, but not a solution unto itself.These initiatives enable an environment where we can adopt AI, but they do not create adoption. Training does not create adoption, funding for start-ups does not create adoption, and a responsible AI framework does not create adoption, although these are all important enabling capabilities.
“An innovation mindset says that we want to be first; we want to take the risk and implement new technology ahead of our global peers. This mindset does not presently exist in all of our industries.”
The second set of initiatives are “direct adoption initiatives,” which support an innovation mindset. These include measurement programs tied to outcomes, procurement policies, and incentives and rewards. In Canada we have struggled with new technology. An innovation mindset says that we want to be first; we want to take the risk and implement new technology ahead of our global peers. This mindset does not presently exist in all of our industries.
What are the main challenges and obstacles SMEs face when seeking to integrate AI into their operations, and what should they prioritize?
The primary issue is resources. These are usually financial resources or the availability of skilled employees who have the capacity to focus on something new in an environment where most entrepreneurs are already working 22 out of their 24 hours. That being said, I think SMEs have an advantage over large organizations since they are not generally dealing with as much legacy infrastructure, legacy processes and old ways of thinking.
“Some colleges have become creative about working with local small businesses by incorporating the businesses’ problems or AI challenges into their curriculum.”
The SME strategy should focus on three fundamental things: One is value discovery. In the context of a particular organization, the main questions to ask are: “What is the value to be gained from AI?” “What does it look like?” and “Where and how can we get it?” Secondly, SMEs need to develop partnerships within their ecosystem. Academic institutions, particularly those running undergraduate and master’s program in AI, are desperate to find challenging projects for their students. Some colleges have become creative about working with local small businesses by incorporating the businesses’ problems or AI challenges into their curriculum, through whichthe students endeavour to solve the problems as part of their course work. The third focus for SMEs should be on developing a funding approach. Once an SME is satisfied that there is value in the AI application and the SME has reasonably priced models for implementation, it needs funding. We need to get creative, look at how we can improve existing government policies, and push the government to support AI adoption initiatives through funding.
What specific steps should Canadian governments take to drive AI adoption among our SMEs?
For the last few years we’ve had a supportive government, not only in terms of promoting entrepreneurship, but also in terms of supporting AI. Providing $125 million to the country’s three major AI institutions was a huge step, and the government should be applauded. However, government can do more on the AI adoption side, including creating more certainty through regulation. I’m not suggesting that onerous legislation and regulations are required. Rather, certainty in regulation has been slow and non-specific in Canada. In Europe, for instance, the General Data Protection Regulation (GDPR) has specified what organizations can and cannot do. I think the absence of regulations in Canada provides an excuse for organizations to delay their full commitment to new technologies and initiations until there is certainty around regulations.
“The absence of regulations in Canada provides an excuse for organizations to delay their full commitment to new technologies and initiations until there is certainty around regulations.”
The second thing we need is more leadership by example. The federal and several provincial governments are very big employers running very large operations across multiple business lines, functions and domains. However, they still haven’t adopted AI in a major way. It is difficult for Canadian governments to say they have made major investments in AI and want Canadian companies to adopt it, when they themselves have not adopted it in a serious way.This is not to say they haven’t done anything. At the federal, provincial and municipal levels we have seen examples of small AI projects and lots of experimentation, but I think we require more in terms of accelerated adoption.
“It is difficult for Canadian governments to say they have made major investments in AI and want Canadian companies to adopt it, when they themselves have not adopted it in a serious way.”
The third thing I suggest moving forward is a shift in some funding from the enabling focus to a focus on direct deployment, whereby funding is measured more by a program’s outcomes and procurement policies, offering the organization incentives and rewards. There is a great example of this in Singapore called 100 Experiments (100E). If any organization comes to them with a problem statement and they can’t find an AI solution that easily addresses the problem, they collaboratively engage a team of university researchers and their own team of engineers to design one. The government matches the amount that the company invests, up to $250,000, helping them to develop an AI solution within their operations and to train their people to continue the solution’s development and implementation.
What are the types and degree of improvement SMEs can gain from successful AI implementations?
The benefit and value a company can achieve through AI will vary depending on the industry and what the organization does. Generally speaking, there are a few categories of value for organizations considering AI adoption.
The one that gets the most attention, because it’s easy to understand, is intelligence automation. This is basically doing with AI what we could not do before. By way of example, someone who is making investment decisions may have a research team whose members read as many articles as they can on a given day in order to decide what to invest in that day. All of Canada’s large banks and many banks around the world now realize that with AI and natural language processing, they can ingest thousands of articles a day, get to the gist of those articles, get machines to summarize key themes and trends, and arrive at an investment decision with greater ease and efficiency. It does not change the nature of the work performed by the investment managers, it just changes the source of the input–from thousands and thousands of unstructured documents to simple summaries of information, making it quicker and easier to get the job done. Is that a cost savings? Maybe – it depends on the organization and its focus. The organization might redeploy the many people who were reading articles to different tasks or new ones on its radar. Maybe it’s not a cost-play but rather the organization’s ability to do more or to look at more industries or participate in more markets. But certainly for some companies, it presents a cost optimization opportunity.
“70% of industry use cases could see incremental value from AI over and above traditional techniques.”
There are also examples that apply to small and medium-sized production and assembly settings. In traditional assembly settings, such as auto parts for example, quality process measures are in place where the part has to conform with certain specifications. The part is tested at the end of the line to determine if it conforms. If it conforms, it gets a pass. And if it doesn’t, it gets a fail and leads to wastage. With AI we are starting to ingest data points along the production process through sensors. We are able to make course corrections in the middle of the production process, so that waste is diminished.
The value an organization stands to benefit from AI adoption will be different depending on the industry, but we believe most industries can derive benefits over and above those they achieve through traditional automation techniques. There are several recent research papers that support this and I would say that 70% of industry use cases could see incremental value from AI over and above traditional techniques.