Overview | Building Canada’s Future AI Workforce in the Brave New (Post-Pandemic) World
Study Scope
This report explores the supports needed for Canada’s digital workforce to acquire more robust skills in artificial intelligence (AI). A key focus is addressing the knowledge gap between subgroups on AI product development teams in the finance and healthcare sectors.
Grounded in interviews with AI industry leaders, the report includes calls to action for stakeholders in Canada’s AI ecosystem.
Study Context
Recent advances in AI in finance and healthcare hold great potential for innovation leading to better customer and patient engagement, employee empowerment, operational efficiencies, and industry transformation.
The COVID-19 pandemic has exerted a paradigmatic shift in the workplace, and AI will also be critical to improving the efficiencies of the increasingly remote and work-from-home workforce across sectors.
In 2020 during the pandemic, Canadian employment in digital roles grew by 9.7%. LinkedIn data suggests global employment in digital roles will increase another by another 363%.
Statistics Canada identifies web designers and developers, information systems testing technicians, and software engineers and designers among ICT roles that have seen increased demand in Canada.
In November 2020, the Government of Canada announced an additional investment of $1.5 billion in provincial and territorial Workforce Development Agreements, highlighting the importance of skills training initiatives in Canada.
Study Findings
AI in Finance
Financial Services (encompass banking and finance, asset and wealth management, credit, insurance, and financial technology industries):
- Contribute approximately 7% to Canada’s GDP
- Employ more than 800,000 Canadians
- Large clusters of activity in Toronto, Montreal, Calgary, and Vancouver
The AI Advantage
Financial services companies that implement AI solutions outperform their competition and report revenue growth of up to 19% directly attributable to AI.
New AI initiatives have resulted in:
- Cost reductions
- Productivity gains
- Revenue enhancement
- Better customer engagement, including customer acquisition, satisfaction, and retention
How AI is Used in Finance
Financial services mostly use machine learning (ML), often referred to as deep learning, based on neural networks. Typical financial applications include:
- Algorithmic trading
- Prediction (of defaults, credit worthiness, possible fraud, cash flows, etc.)
- Time series analysis
- Natural language processing (NLP)
Challenges
Ethical issues are a significant challenge, i.e., data selection for training the model (historical data contain biases or may not be applicable).
Another concern is the ability to explain how ML models arrive at outcomes (some algorithm decision-making is more explainable than others).
Talent, Training, and Education
Core Finance AI teams typically consist of two layers:
- Technical leads with PhD training, responsible for guiding the research and architecture design
- Software developers with strong ML, math, and coding skills
Given the technical nature of the industry, domain knowledge is crucial.
A key role is Data Engineer: analyzing variables in the datasets and combining them to create new features that improve model accuracy.
AI in Healthcare
AI in healthcare helps in detection and diagnosis, patient treatment, and service delivery.
The continuing integration of intelligent innovations and technology in healthcare are expected to transform healthcare, impacting the nature of healthcare occupations and the economy.
AI in healthcare follows two paths:
- Deep learning for image processing and analysis
- Natural language processing (NLP)
Image processing and analysis is used primarily in assistive tools used by radiologists, while NLP is used in patient case analysis and diagnostics.
Cloud platforms were identified as a core component of clinic and hospital infrastructure, delivering AI solutions through Software-as-a-Service (SaaS).
Benefits
AI will enable information gathering, processing, and analyzing at faster speeds and greater accuracy and precision, potentially allowing for improvement in the health of the population.
AI could result in better healthcare affordability, accuracy, and accessibility.
AI technology is expected to complement healthcare by enhancing clinician work (this separate to “automation” or “robotics”).
AI pandemic advantages:
- Remote counselling
- Providing at-risk patients access with peace of mind
- Reducing strain on healthcare system
Future of AI in Healthcare
Study interviewees expect AI and ML to further expanding in diagnostics (such as medical imaging and analysis) over the next five years.
Expansion of AI-guided treatments such as radiation therapy.
Addressing Concerns
Key concerns (especially in training medical ML applications) are data collection, data privacy, governance, security, governmental regulation (Health Canada is developing data rules and regulations).
Wider public acceptance will require addressing concerns about personal data usage and suspicions of data monetization.
Study interviewees worried AI system development and deployment is outpacing the ethical frameworks that are necessary but not yet fully formed.
More collaboration and cross-disciplinary understanding between medical practitioners and AI experts are required.
Talent, Training, Education
Conventional roles such as computer scientists, systems engineers, software developers, and AI or ML specialists are key to AI teams. However, ethicists, human factors specialists, and government representatives will require consultation on policy and regulations.
Additional needed roles:
- Graduate-level employees such as health informaticians and health services researchers
- Chief Privacy Officers to address privacy concerns and guide engineers on privacy policies
- Ethnographers versed in qualitative research methods for detailed studies of cultural groups
The healthcare sector also requires cross-disciplinary teams, consisting of domain and technical experts.
Upskilling could train experts on complementary fields within the development team. Upskilling can provide insight and basic competency but does not create experts (who typically require university education).
AI specialists in healthcare are mostly post-graduate-level educated in ML. Employers tend to regard these high performers as “easy to upskill when necessary.”
Typical ML profile: employee who has studied in the domain for multiple years, either in a university setting or as industry employees with experience executing ML implementations.
Building an AI Team
Key Roles and Skillsets
Ideal teams in healthcare and finance consist of graduate-level trained domain experts, experienced developers, or developers with some technical undergraduate degree:
Collaboration, teamwork, and functional understanding of the other knowledge domains is essential:
- This intersection of knowledge is made possible through collaborative teamwork and upskilling
- Domain experts must have sufficient understanding of the market and business interests to not stray into pure research (applies equally for business and technical)
Education and Experience
Fostering an AI Talent Pipeline
Cross Training vs. Upskilling
Fostering an AI talent pipeline for the financial services and healthcare sectors in Canada essentially means building more “ideal type” AI teams.
This most effectively accomplished through cross training and mentorship programs that target the three development subgroups (domain, business, and AI experts).
The goal would be to cross train each expert in the relevant subject matter of the other two.
Domain Cross Training
The upskilling landscape offers many online courses and university certificates for business and computer science but very few that teach specific domain knowledge, i.e., specific to primary care or credit risk assessments.
Domain upskilling needs to be tailored to product-specific knowledge and include research trends and relevant regulatory and legal frameworks.
Business Cross Training
Unlike traditional business courses, which cover a wide range of topics, upskilling should focus primarily on product and client-specific work that identifies and defines business problems, product management, quality assurance, and client relations in the context of developing AI capabilities and products.
This cross-training should include current market research on domain applications and product trends.
AI/ML Cross Training
There is currently a severe gap in upskilling opportunities for AI/ML:
- Bootcamps are too watered down
- Graduate-level certificate courses are too costly and time intensive
- Massive open online courses (MOOCS) have little to no consistency or quality control
Effective cross training in AI/ML to fill this gap would focus only on the core components of AI/ML that are important for specific sectors, i.e., healthcare or financial services.
Cross-training example:
- How to train ML models specifically on ambulatory, inpatient, and adjudicated claims data using electronic medical record systems with the objectives of discerning which data points correlate to the rates of health events in a population, the history of disease, and finding the most successful treatments
Regulatory, Ethics, and Legal Cross Training
Teams will require training in legal, ethical, and regulatory matters specific to their domain, including privacy and security.
Specific course content will depend heavily on the sector and subsector, and the associated jurisdiction
Mentorship
Mentorship programs could be an important complement to cross-training program courses to facilitate on-the-job knowledge and experience:
Mentorship programs could foster cross-trained talent internally (by designating AI/ML, business, or domain lead as an assigned mentor to a recent hire or new team member).
Cross-Training Challenges and Market Opportunity
The most pressing challenge is devising courses that are sufficiently comprehensive while ensuring they can be delivered quickly.
The following considerations are important:
- Determining which delivery method is best suited to this type of skillset acquisition
- What is the fastest way to facilitate cross-training courses
- How to minimize the overall learning curve without oversimplifying the material
A market opportunity exists for the development of quality, industry-certifiable training courses that are delivered by companies with domain expertise.
Call to Action
- Businesses and organizations that develop AI products for use in the healthcare and financial services sectors should prioritize cross-training for development teams
- Canadian academic institutions should assess and seek to fill the availability and accessibility of AI-related courses in non-technical programs such as business, finance, or medical programs and vice versa
- Stakeholders in heavily regulated sectors like financial services and healthcare need to establish industry-wide data governance standards and secure data-sharing mechanisms to enable greater and more secure data access
- Canada’s strategy for AI skills development should include cross-training as a fundamental pillar in workforce development efforts. For example, government could include AI cross training in its programs under the Workforce Development Agreements
- In light of the resilience of the tech sector during the pandemic, the federal government should continue to prioritize and support remote upskilling programs to address broad information and communications technology (ICT) skill needs
