Artificial intelligence has become big business — and the pace of innovation is only picking up. According to Deutsche Bank, 175,072 AI patents were filed between 2012 and 2022, with more than half of them coming in those final three years. The bank anticipates a dramatic spike this year and next in companies adopting AI applications, especially in such fields as product development, sales, marketing and human resources. Legal firms now use AI to generate contracts; travel companies rely on chatbots to provide help during the booking process. Already, the global AI market is worth roughly U.S.$136.6 billion, and it’s on track to reach U.S.$1.3 trillion by 2032. Patents for AI innovations, as seen in the figure below, are being filed in many different sectors. From 2022 to 2030, AI use by organizations around the world is expected to expand at a compound annual growth rate of more than 38 percent.
Machine learning co-occurs most frequently with the life and medical sciences and telecommunications fields; computer vision with telecommunications and transportation
Source: WIPO Technology Trends
It’s clear that AI adoption is climbing at a breakneck rate. Experts predict that as computational power grows exponentially, the capabilities of these AI applications — in reasoning, in accuracy, in specialization and in personalization — will skyrocket. At the same time, regulations and policy can take much longer to develop. The European Union spent three years drafting its 125-page law to regulate artificial intelligence, introduced in April 2021. But none of those 125 pages mentioned generative AI, the breakthrough that powers applications like ChatGPT and that blindsided lawmakers. While regulators work to catch up, business leaders need to take their own steps to ensure that the technology being developed and used today doesn’t have harmful consequences. Policy-makers are having to play catch up. For instance, a bipartisan group of U.S. House representatives proposed new legislation in January to regulate the use of AI to create clones or likenesses of artists. As the technology develops, it’s important for business leaders and policy-makers to ensure AI is used in the service of society.
For more than 30 years, Canadian scientists have played a vital role in developing the foundation for AI technologies. In 2017, Canada became the first nation to launch a national AI strategy. When it comes to average year-over-year growth in AI talent concentration, as well as publications per capita from AI researchers, Canada outpaces all other G7 nations. Canada has one of the highest rates of AI patents filed and one of the highest per-capita levels of VC investment in AI enablers, developers and users. McKinsey recently named Toronto one of 16 global hubs focused on advancing AI and its technologies.
But when it comes to businesses actually using AI, the story looks quite different — and Canada is in danger of falling behind. A recent KPMG survey reported that only 35 percent of large Canadian businesses currently use AI in their operations, compared to 72 percent of large U.S. businesses. Small businesses fare even worse: by 2021 only 3 percent were using AI, Toronto Metropolitan University found in a recent report, and rates of adoption are lower still among several equity-seeking groups. Currently, companies owned by women, Indigenous people and people with disabilities are far less likely than other companies to use AI.
By failing to take advantage of AI, businesses miss out on many opportunities to remain competitive with their global peers and, especially, to boost their productivity, which is directly tied to higher wages and higher standards of living in this country.
Source: Deloitte
Source: Deloitte
Artificial intelligence (AI) is the science and engineering of making computer programs that can simulate human intelligence.
Generative AI is a type of artificial intelligence technology that can produce text, images, audio and synthetic data.
Machine learning is a branch of artificial intelligence, in which computer systems are able to learn and adapt by using algorithms and statistical models to analyze data.
Responsible AI is the practice of designing, developing and using artificial intelligence with good intention to support employees and businesses and fairly benefit customers and society.
Tech stewardship is a professional identity, orientation and practice that enables people to identify work-related opportunities that help make positive use of technology.
“Weak productivity growth has been a serial problem in Canada,” says Bank of Canada governor Tiff Maclem. AI offers the potential to address that head-on, unlocking new capabilities and increasing efficiencies across a myriad of sectors. Google estimates generative AI could save the average worker in Canada about 100 hours a year. And research from Goldman Sachs suggests generative AI could spur productivity growth by 1.5 percentage points over 10 years. All told, according to estimates by McKinsey, generative AI’s impact on productivity could add the equivalent of up to $4.4 trillion in value to the global economy every year.
That said, implementing new technologies demands resources as well as considerable time: it can take up to five years to fully integrate AI into an enterprise. A significant majority of Canadian businesses have yet to implement this technology. According to the 2021 Survey of Digital Technology and Internet Use conducted by Statistics Canada, the biggest barrier to adoption is a lack of motivation: 69 percent of businesses that don’t use AI say they have not identified a business need for it. The next-biggest barrier is a lack of awareness: 28 percent of those businesses acknowledge that they don’t know what tools are available. The cost of AI technologies and the skills gap in companies’ workforce were also cited as obstacles to adoption, as were legal, privacy and security concerns.
When it comes to generative AI specifically, a recent report from the AI Accelerator Institute found that a lack of knowledge is the main reason companies don’t use these models. Some businesses have misconceptions about generative AI’s limitations, while some overestimate what the technology can do, leading to disappointment. Others are concerned around data privacy and intellectual property, or are distrustful of the outputs.
Improved awareness of the uses and benefits of generative AI can help to alleviate decision-makers’ concerns. Testing out AI applications can establish how companies might use this technology to solve specific problems in their business or industry. Many Canadian startups are gaining the trust of corporations by implementing good practices to ensure the technology will be used responsibly, as well as demonstrating real-world use cases of their AI solutions and offering free trials.
New technologies also bring new risks, and when it comes to AI, there can be a lack of transparency in how algorithms make decisions, and how those decisions can replicate and reinforce biases. There are also concerns about how companies address data privacy when using AI tools. These issues can negatively impact customers’ trust in businesses that use AI.
Responsible AI involves designing, developing and using AI systems that are safe, transparent, traceable and non-discriminatory — and governments around the world are making it a priority. The European Union Artificial Intelligence Act (EU AI Act) categorizes AI systems according to the level of risk they pose to fundamental rights and user safety. The Government of Canada has established guiding principles on the effective and ethical use of AI, and in December called out special consideration for vulnerable groups such as children and communities that have historically experienced discrimination and bias. Investors and corporations are also increasingly looking for companies to adhere to certain principles to help ensure the responsible application of AI. Radical Ventures recently released an assessment tool to help VCs assess early-stage AI businesses, and some organizations are incorporating responsible AI as part of the ESG framework.
The widespread use of responsible AI, most importantly, benefits society. But responsible AI can also benefit businesses by minimizing bias in their data, ensuring transparency in their processes, supporting internal governance of AI systems, and protecting the privacy and security of client data.
To ensure that new technologies are beneficial for everyone, businesses and customers must actively engage in tech stewardship. This practice involves the development and use of technology that is purposeful, responsible, inclusive and regenerative. It’s achieved by designing and implementing technology with the intention of creating a positive impact; by anticipating, monitoring and managing the complex impacts of technology; by broadening who and what is involved in decision-making; and by choosing to care for the environment, the economy, communities and individuals.
Canadian ventures can differentiate themselves as tech stewards by considering any technology’s social and ethical implications. They can also empower their teams to use technology that drives business success while creating a positive impact. This can promote tech stewardship across networks and industries.
“There is a tendency to view technology too narrowly. By better understanding the full impacts of technology on people and society, and considering the underlying values that are involved, tech stewards uncover new opportunities,” says Mark Abbott, director of Tech Stewardship at MaRS Discovery District. “For companies, this can yield opportunities to grow profit and positive impact. Beyond specific opportunities, building tech stewardship into a company’s culture can lead to sustainable competitive advantage as it enables them to more quickly and effectively realize the potential of the technologies of today and tomorrow.”
*Startups are defined as companies that were active in the last five years, have received at least one investment deal in the last five years that is more than U.S.$1 million. AI companies included those that are either AI developers, offered AI services, used AI to drive their services, or had plans to implement AI in their businesses within two years.
Source: Deloitte
The following six ventures in the MaRS portfolio are leaders in AI and machine learning. These companies are working on solutions that can help companies use AI responsibly to enhance their productivity, identify key insights, improve customer service and transform their business.
Combined, these six ventures:
Mike Murchison, CEO and Co-founder
Founded: 2016
What it does: Trained on company documents and policies, Ada’s chatbot responds as a human agent would and is able to resolve inquiries without human involvement. It can be deployed across multiple channels and modalities, including telephone, web, mobile and social. Companies can measure the AI chatbot’s performance by tracking the quality of every interaction, evaluating how well it provides resolutions and training it to improve its responses.
What it offers: Ada’s AI chatbot reduces the amount of time and human resources a company spends in customer service. On average, Ada’s clients resolve 40 percent of customer inquiries without needing human intervention.
“Ada is an all-in-one generative AI,” says Yochai Konig, Ada’s vice president of machine learning. “We’re offering outcome-based pricing, where clients pay only for actual customer resolutions. The tech is good enough today to bring material value to customers and clients.”
Already, Ada handles about 70 percent of Canadian fintech firm Wealthsimple’s client inquiries. The budgeting app Qapital is now able to handle more than 25,000 customer service inquiries per month. AirAsia was able to reduce customer wait time by 98 percent over a period of four weeks.
How it fosters tech stewardship: Ada adheres to global security and privacy regulations. Any data collected from customers can only be used to resolve their enquiries. Data cannot be stored or used for other purposes or to train other generative AI models. Sensitive personal information is automatically redacted from automated conversations, and the chatbot only pulls information from first-party support documentation to provide assistance. If a customer inquiry falls outside the company’s guidelines for information it can legally provide, the chatbot moves that inquiry to human review. Ada also employs analytics to identify potential blind spots in its algorithms and systems. The AI can improve internal policies to resolve customers’ issues and eliminate potential customer service problems.
Clients can use the technology for a trial period to see how it would work for them. They provide the guidance and policies for the AI chatbot, so they direct what answers the technology will offer customers. Ada encourages clients to try to “break” the chatbot to see what could go wrong.
Traction to date: Ada handles billions of interactions every year for more than 350 businesses across North America and around the world, making it one of the larger players in the chatbot industry. Ada currently supports more than 50 languages, with more languages on the way, and has expanded the chatbot’s ability to take action on customers’ requests.
Humera Malik, CEO and Co-founder
Founded: 2016
What it does: Canvass AI develops AI software for process manufacturing industries, such as oil and gas, and chemical. Using companies’ sensor, text and visual data, the software builds and selects AI models to provide predictive and prescriptive solutions that fast-track their operational efficiency, profitability and sustainability goals. The AI software can quickly identify the causes of potential disruptions to prevent delays and minimize production quality resulting in losses. Decisions and actions can be made faster, dramatically improving response times to operational problems and reducing maintenance costs.
What it offers: “Large companies are supposed to be super advanced, but many are still running like they did 60 years ago,” says Josef Zankowicz, vice president of corporate development at Canvass AI. “They follow the ‘If it’s not broken, don’t fix it model,’ so things run how they have always run.”
Canvass AI helps companies modernize and improve their operations by making sense of their data. The Canvass AI software can determine the state of a process based on data from indirect sources. One manufacturer used the solution to predict product quality and yield flow in real time, which led to improved product quality while reducing costs by 5 percent. Canvass AI also acts as a repository for institutional data so changes in human resources — whether from retirements, shift changes or new hires — do not affect existing or newly created efficiencies.
Canvass AI has improved companies’ productivity, reliability and quality control, while reducing their costs, waste, water consumption and carbon and GHG emissions. Depending on the sector, companies have seen efficiency improvements ranging from 20 to 30 percent. Engineers are able to do their work up to 40 percent faster, often by using AI to reduce the time spent analyzing trends on spreadsheets.
How it fosters tech stewardship: Canvass AI operates as a software as a service from the Microsoft Azure Marketplace. Customers operate and control their data and the models trained on that data, which are protected with Microsoft’s cloud-level security. The company provides “explainable AI” — users can query the software to understand what data and inputs were used to achieve specific results.
Canvass AI believes in humanizing the AI experience. People are at the centre of its software to ensure the technology and its solutions are accessible and explainable. Canvass AI encourages clients to verify and validate both the inputs and outputs, and to rely on context, experience and knowledge to evaluate the results.
Traction to date: Canvass AI has more than 100 pilot projects underway across several industries, including pulp and paper, oil and gas, and food processing. It recently signed a multi-year deal with one of North America’s largest oil refineries, creating 30 use cases across its network. These use cases can be applied to 700 other refineries around the world, as well as 5,000 chemical plants.
The company has partnered with Microsoft to deliver its predictive analytics software solutions on the Microsoft Azure cloud platform. It is also available as an on-premises solution on request. The goal is to help industrial companies accelerate their Industry 4.0 strategies and benefit from using the Internet of Things.
Faisal Ahmed, Co-founder & CTO
Founded: 2016
What it does: Knockri uses AI-powered behavioural assessments to find companies the best-talent for the job. It conducts and scores structured interviews that focus on behavioural skills such as problem-solving, teamwork, communication, leadership, adaptability and work ethic. Companies can use Knockri’s skills framework, which has been validated by an industrial organizational psychologist for the best performance prediction, to guide their assessments. Knockri can also analyze and audit its clients’ frameworks for use in assessment.
What it offers: “Structured behavioural interviews have been proven best for predicting on-job performance and having the least amount of adverse impact,” says Faisal Ahmed, co-founder and CTO. “It creates a balance between fairness and utility.”
According to Knockri’s research, when it comes to job performance, structured behavioural assessment is 16 times more predictive than looking at previous experience, two times more predictive than personality tests or unstructured interviews and four times better at achieving diversity than cognitive ability tests. However, these assessments are costly and time-consuming to administer, and can be difficult to apply consistently across candidate pools. Knockri uses AI to automate the assessment process and level the playing field. Its technology is able to apply the same assessment with the same criteria to every individual, which advances fairness during the hiring and promotion process.
Knockri has found that recruiters are only able to review 10 percent of all applicants. Its technology enables recruiters to review the entire applicant pool in a short period of time. For example, the Department of Defence accumulated 232 hours of interviews, which would normally take several recruiters at least three weeks to review. Knockri’s AI completed the review in only three days.
How it fosters tech stewardship: “Many industry hiring processes are fraught with bias and inefficiency,” says Ahmed. “AI can help to address these issues and ensure fairness in hiring. Companies should do their due diligence to see what data to validate, what questions to ask and what processes to put in place to use the technology.”
Knockri follows industry best practices to safeguard the security and privacy of candidates’ data. The company deliberately designed its machine learning algorithm not to use historical hiring data, which is known to contain biases. It monitors how its algorithm makes predictions across groups to identify and eliminate behaviours and responses that could result in an unfair advantage.
Knockri reports on its findings internally and to customers through whitepapers and industry reports. It continually seeks to uncover biases found in historical hiring practices, such as choosing candidates based on their university degree or personal familiarity. It also moved away from video interviews (except when necessary), which made some candidates uncomfortable and introduced bias into the hiring process. Knockri primarily uses audio interviews, which provide the required responses and eliminate unnecessary data.
Traction to date: Knockri’s customers include medium to large enterprises in the finance, technology, consulting and government sectors. It’s also recently onboarded one of Ontario’s large electricity providers.
Nauman Jaffar, CEO
Founded: 2012
What it does: MarkiTech develops products, platforms and services geared to the underserved seniors’ healthcare market. It uses a combination of technologies and approaches, including software development, machine learning, computer vision and AI software, to design different applications for enterprises and startups.
MarkiTech has created a number of technologies that address specific issues within the healthcare space. Your Doctors Online is an app that enables patients to consult securely with a board-certified physician, who can provide advice and prescriptions. Veyetals is a vital-signs camera app that lets patients collect and monitor data for telemedicine visits. Designed for organizations and their employees, Safe2Work is an AI-powered health intelligence platform that provides non-invasive real-time risk monitoring and health screening. SenSights is an analytics and insights platform that enables home health agencies to monitor, care and track seniors around the clock. Carefall is an AI application that monitors seniors to predict and prevent falls based on behavioural, biological, demographic and environmental factors. And DiabetesPredict.AI is an AI application that monitors individuals to predict the probability of chronic diseases.
In addition to these applications and SaaS platforms, MarkiTech develops solutions to support mobile learning, provides staffing and team augmentation, and conducts penetration testing to identify companies’ IT vulnerabilities.
What it offers: “We are project focused, and we aim to make a difference in healthcare one project at a time,” says CEO Naumann Jaffar.
MarkiTech’s AI-based tech benefits everyone along the healthcare chain. Using technology that reduces paperwork and notifies physicians of changes in patients’ health means that medical professionals can spend more time with people. AI also removes the need for patients to use expensive technology to track their health and access a physician. Patients can manage their own data.
How it fosters tech stewardship: Because it deals with patient data, MarkiTech is thorough in ensuring data security and privacy. The company has experience in conducting privacy and threat assessments and has completed more than 40 PHIPA- & HIPAA-compliant projects in healthcare technology. SenSights.AI is ISO 27001:2022 certified, providing the highest level of information security. MarkiTech also goes through third-party auditing to ensure it meets industry security requirements.
“Internally, nobody has access to private patient data. All data is anonymized and not shared outside any of the apps,” adds Jaffar.
Traction to date: Your Doctors Online has been spun off into its own company. It employs more than 50 people worldwide, serves patients in more than 60 countries, manages more than 3,000 virtual visits per day and has handled more than one million transactions to date.
Ryan Ferguson, Founder and CEO
Founded: 2019
What it does: Riskfuel develops risk models that enable banks and financial institutions to calculate valuations and risk sensitivities in real time. Rather than estimating profits and losses at the end of closing, companies can use these models to make informed trading decisions based on what is happening in the market at any given moment.
Riskfuel’s models can work under all types of market conditions, apply a wide range of risk levels, support “what-if” scenarios and achieve accurate results. The technology decreases calculation costs, as it reduces the runtime of the models, which in turn reduces their computing requirements and environmental footprint.
What it offers: “Trading is an arms race,” says Ryan Ferguson, CEO and founder of Riskfuel. “Better tools for managing risk enables more aggression when going after customers.”
In order to make buying and selling decisions, financial institutions depend heavily on risk evaluation — but it’s a computer-intensive process. With Riskfuel’s technology, clients can evaluate risks more thoroughly before conducting trades. Trading professionals can make more frequent calculations to better understand what could happen in markets under different high-risk scenarios. For example, Riskfuel’s models proved resilient when COVID-19 caused volatility in the market.
Companies do not need to purchase proprietary systems or implement new tools. Models can be deployed one-by-one, so clients can validate the model before deploying it across the portfolio.
How it fosters tech stewardship: “Banking is a heavily regulated industry, and it takes time to build trust in the technology,” says Ferguson. “We mitigate their concerns by building fast replicas of existing models. We then compare the results of both models to demonstrate transparency of risks and pricing, and to see if the results are the same.”
Riskfuel copies clients’ existing models to demonstrate what the technology can do. It validates the quality of the build by inputting thousands of pricing requests to ensure it accurately replicates results generated by the client’s model. Because Riskfuel creates its own model, it can input synthetic data (instead of actual client data or trades) to test the system.
In minutes, a handful of servers can complete calculations that would have taken a data centre all night — using less than 1 percent of a client’s in-house computers. A Microsoft case study determined Riskfuel’s model performed 20 million times faster than the original target model while maintaining accuracy.
Traction to date: “It typically takes six months to build complicated models, and we’re working to reduce that time to six weeks,” says Ferguson. “We know what problems our clients share, and we’re developing tools to automate processes, remove iterations and standardize certain elements.”
Riskfuel currently has six clients in the banking industry, including Scotiabank, BMO and several banks in Europe. It contributed core technology and capabilities to Scotiabank’s risk engine, which can calculate nearly one billion valuation adjustments (XVAs) per second. The tool enables traders and risk managers to better understand exposure across asset classes, and delivers accurate results in real time at much lower costs. Riskfuel’s work with Scotiabank led to the bank winning the 2021 Risk.net award for Technology Innovation of the Year.
Braden Ream, CEO
Founded: 2019
What it does: Voiceflow helps conversational AI teams design, test and manage AI agents. Its platform is easy to use, enabling clients to scale quickly, and the no-code builder allows companies to create a variety of AI products, such as workflows, chatbots, voicebots and other tools. Voiceflow also supports synchronous and asynchronous collaboration, including features like commenting and versioning.
Companies can harness their data to train their own AI chatbots using Voiceflow’s knowledge base. This results in more accurate and contextual responses, automates the provision of FAQs and improves the chatbot’s ability to perform tasks. Voiceflow also helps companies build and test shareable prototypes to get feedback more quickly.
What it offers: “We enable companies to experiment with AI in a safe and controlled environment,” says CEO Braden Ream.
Voiceflow provides a free trial of its AI solution. Clients can watch video tutorials to learn more about the capabilities, and explore community-built tools and plugins to determine how to extend its functionality. Its technology allows customers to get to market 50 percent faster without developers or a technical team, resulting in additional cost savings.
How it fosters tech stewardship: Voiceflow ensures the security of clients’ data and end-user interactions. It maintains all necessary certifications and provides the highest levels of data, network, organizational and application security. To support responsible AI, it scrapes all sensitive data so no identifiable personal information is used to develop models. The company uses clean data in training models and tests all models vigorously to eliminate model bias.
Traction to date: Voiceflow’s clients have seen significant results after implementing the technology. For example, eBay auction sniper eSnipe has been able to implement an AI-powered search agent to automate 70 percent of its help centre tickets, reducing the burden on its support team and increasing customer response times. BMW reduced its time for in-car assistant prototyping time by 50 percent. The Home Depot was able to scale the number of interactive voice response (IVR) user tests from 12 to 300 within the first week of testing.
Voiceflow has received funding through Scientific Research and Experimental Development (SR&ED) tax incentives. It has more than 200,000 global users, as well as more than 1,500 customers. To date, its technology has automated more than 1 billion messages.