Human Resources Committee on Nov. 22nd, 2023

[The Chair (Mr. Robert Morrissey (Egmont, Lib.)), Liberal, Bobby Morrissey]

Committee members, the clerk has advised me that we have a quorum and that those appearing virtually have been sound-tested. All are okay except for one witness, but there is another witness from the same group who is okay.

With that, I will call the meeting to order. Welcome to meeting number 90 of the House of Commons Standing Committee on Human Resources, Skills and Social Development and the Status of Persons with Disabilities. Pursuant to Standing Order 108(2), the committee is resuming its study on the implications of artificial intelligence technologies for the Canadian labour force.

Today's meeting is taking place in a hybrid format, meaning there are members and witnesses appearing virtually and in the room. You have the option of choosing to participate in the official language of your choice by using interpretation services. There are headsets in the room. As well, if you are virtual, please use the "world" icon at the bottom of your Surface tablet and choose the official language of your choice.

If there is an interruption in translation, please get my attention. We'll suspend while it is being corrected.

I would ask those participating to speak slowly, if possible, for the benefit of the interpreters. To those in the room, keep your earpiece away from the mike to protect the hearing of the translators.

Again, if you're appearing virtually, to get my attention use the “raise hand” icon at the bottom of your Surface.

Today we will be meeting from 4:30 to 6:00 with witnesses.

As an individual, we have David Kiron, editorial director, Massachusetts Institute of Technology Sloan Management Review, by video conference. From the Canadian Union of Public Employees—Quebec, we have Danick Soucy, president, political official, committee on new technologies, by video conference; and Nathalie Blais, research representative. Nathalie does have issues with sound. If they those cannot be resolved, she will not participate. From SAP Canada Incorporated we have Yana Lukasheh, vice-president, government affairs and business development.

With that, we will start with five minutes for each group, beginning with David Kiron. You have five minutes or less, please.

[Dr. David Kiron, Editorial Director, Massachusetts Institute of Technology Sloan Management Review, As an Individual]

Distinguished members of the committee, thank you for organizing this important study and inviting me to participate in it.

I'll discuss how AI is influencing four categories of work: designing work, supplying workers, conducting work and measuring work and workers. AI-related shifts in each category have policy implications. In aggregate, these shifts raise questions about how to optimize producer flexibility, worker equity and security. More broadly, these trends create policy opportunities for increasing productivity at the national level and strengthening social safety nets.

Although we need policies to address worker displacement from AI-related automation, policy also needs to address AI's influence on a wide range of business activities, including human-machine interactions, surveillance and the use of external or contingent workers. Policy addressing AI in workforce ecosystems should balance workers' interests in sustainable and decent jobs with employers' interests in productivity and economic growth. The goal should be to allow businesses to meet competitive challenges while avoiding dehumanizing workers, discrimination and inequality.

I refer to "workforce ecosystems" rather than "workforces". Our ongoing research on workforce ecosystems demonstrates that more and more organizations rely on workers other than employees to accomplish work. These include contractors, subcontractors, gig workers, business partners and crowds. Over 90% of managers in our global surveys view non-employees as part of their workforce. Many organizations are looking for best practices to ethically orchestrate all workers in an integrated way.

I'll start with designing work. The growing use of AI has a profound effect on work design and workforce ecosystems, including greater use of crowd-based work designs and disaggregating jobs into component tasks or projects. Consider modern food delivery platforms like Grubhub and DoorDash that use AI for sophisticated scheduling, matching, rating and routing, which has essentially redesigned work within the food delivery industry. Without AI, such crowd-based work designs wouldn't be possible.

AI is also driving recent trends to create work without jobs. On the one hand, this modularization of work can facilitate mobility within the firm and improve employee satisfaction by efficiently matching workers with tasks. On the other hand, designing work around tasks and projects can increase reliance on contingent workers for whom fewer benefits are required. Greater numbers of Canadian contingent workers can increase burdens on government-sponsored safety nets.

Now I'll move to supplying workers. On the one hand, AI is transforming business access to labour pools. On the other hand, workers have more opportunities to work across geographic boundaries, creating opportunities for more workers. Using AI to find suitable workers can have both negative and positive consequences. For example, AI can perpetuate or reduce bias in hiring. Similarly, AI systems can help ensure pay equity or contribute to inequity through the workforce ecosystem, by, for example, amplifying the value of existing skills while reducing the value of other skills. It remains an open question on whether AI-driven work redesigns in the global economy will increase or decrease the supply of workers for Canadian businesses.

I'll go to conducting work. In workforce ecosystems, humans and AI work together to create value, with varying levels of interdependency and control over one another. As MIT Professor Thomas Malone suggests, people have the most control when machines act only as tools. Machines have successively more control as their roles expand to assistants, peers and finally, managers. Emergent uses of generative AI in each category raise a variety of policy questions regarding worker liability, privacy and performance management, among other considerations.

The last category where AI is influencing work is measurement. Firms increasingly use AI to measure behaviours and performance that were once impossible to track. From biometric sensors to corporate email analysis to sentiment analysis, advanced measurement techniques have the potential to generate efficiency gains and improve conditions for workers, but they also risk dehumanizing workers and increasing discrimination in the workplace.

That's about five minutes. I'm happy to continue. I have a conclusion, but I'm also happy to stop there.

[The Chair, Liberal, Bobby Morrissey]

You can continue during the question period and capture anything that you missed in your opening statement.

We will now go to Mr. Soucy for five minutes or less, please.

[Danick Soucy, President, Political Official, Committee on New Technologies, Canadian Union of Public Employees - Quebec]

Mr. Chair and committee members, thank you for inviting us.

My name is Danick Soucy, and I am the political representative of the Committee on New Technology, Quebec division of the Canadian Union of Public Employees, CUPE for short. CUPE Quebec's Committee on New Technology is attempting to gain a better understanding of emerging technologies that could impact the work of our members, including artificial intelligence, or AI. The committee's objective has never been to oppose technological breakthroughs, but, instead, to find ways of adapting to them.

One year ago, rapid advances by ChatGPT surprised the world and even AI specialists. We now know that generative AI systems are able to perform a variety of tasks. Not only can they allow for the automation of manual labour, but they can also perform numerous professional creative tasks or those normally undertaken by office staff. Generative AI has immense possibilities and could cause serious upheavals in the working world and in Canadian society if no guardrails are put in place.

We believe that it is imperative that action be taken immediately to regulate AI before companies undertake large-scale implementation, so that everything possible is done to avoid bringing in systems that cause problems for workers or for society at large. The old saying an ounce of prevention is worth a pound of cure certainly pertains to AI, which, in spite of its usefulness, can cause dangers on many different levels.

One of the dangers is that many AI systems were trained using the Internet. As a result, they have incorporated biases and inaccurate data that can lead to discrimination or disinformation. However, commercial AI systems are more non-transparent than ever, and their suppliers do not always reveal what data sets they were trained on. In addition, the autonomy of AI systems makes it more complex to determine who or what is responsible when harm is done. The public and employers must be educated on this issue.

In the workplace, this can mean rejections of either applications or promotions, or non-compliance with workers' fundamental rights in terms of privacy or the protection of personal information. AI systems used to assign duties to workers can also impact their health and safety by intensifying their work or by limiting, for example, their decision-making leeway, which is recognized as a work-related psychosocial risk.

AI should not lead to discrimination, result in increased occupational health and safety problems or jeopardize an employee's privacy or personal information.

Available data on the possible impacts of AI systems on labour vary greatly. However, a shocking study published by Goldman Sachs, a U.S. investment bank, estimated last March that AI could result in the automation of 300 million full-time jobs worldwide. This estimate includes the disappearance of a quarter of the work currently done in the U.S. and Europe. This is, by far, the most alarming assessment of which we are aware.

In such a scenario, what would happen to laid-off workers? Would employment insurance be all they could count on?

Would companies be responsible for their retraining?

Would they be required to train their staff whose work was transformed by AI?

Would they compensate governments for income tax revenues lost because of the use of AI to protect our public services?

The government cannot consider the use of AI solely from the angle of innovation, productivity and economic growth. It must also take into account the adverse impacts that AI systems would have on citizens and their ability to contribute to Canadian society more generally.

To this end, CUPE Quebec recommends that governments maintain a dialogue with all groups in civil society, including unions, on the subject of AI and that the government entrust Statistics Canada with the mandatory collection of information on the progression of AI and its impacts on work and on labour.

Furthermore, the regulations to be implemented quickly should at least address the following four elements.

First, employers should be obligated to declare any use of AI in the workplace and involve workers or their union representatives prior to the design and implementation of AI systems.

Second, employers should be required to train or requalify personnel affected by the adoption of AI.

Third, implementation of a legal framework is necessary to protect the fundamental rights of workers and to identify those responsible for AI systems.

Fourth and finally, requirements should be imposed relating to the responsible development of AI for the granting of any public funding.

Thank you for your attention.

[The Chair, Liberal, Bobby Morrissey]

Thank you, Mr. Soucy.

Madame Lukasheh, you have five minutes.

[Yana Lukasheh, Vice-President, Government Affairs and Business Development, SAP Canada Inc.]

Thank you, Mr. Chair and members of the committee. We appreciate the opportunity to appear before you today to contribute to the study regarding the implications of artificial intelligence technologies for the Canadian labour force.

SAP is a software technology application enterprise with long-standing operations in Canada spanning over 30 years. We work with organizations of all sizes across the public and private sectors to enable them to become part of a network of intelligent and sustainable enterprises.

Our secure and trusted technologies run integrated AI-powered business processes in the cloud. More specifically, our applications cover enterprise resource planning, human resources and procurement and finance management, including travel and expense claims.

SAP is a global enterprise present in 140 countries, with Canadian operations of strategic importance. We contribute $1.5 billion annually to the Canadian GDP and have a total of 7,000 jobs in our ecosystem from coast to coast to coast. Our innovation labs where our R and D is conducted are located Montreal, Waterloo and Vancouver.

We understand that Canada's labour force today is confronted by the fast-paced evolution of AI technology, and workers are increasingly faced with a series of complex decisions related to implementation and training as organizations are evolving within this new digital era. As AI is increasingly used to automate decisions that have a significant impact on people's lives, health and safety, we recognize that governments have an important role to play in promoting innovation while safeguarding public interest.

Concerns, which we hope to discuss as a part of our testimony today, are common and are often overlooked practices associated with a general lack of AI integration, which we have seen impact many industries, including Canada's public sector. For example, I'm referring to the use of disconnected or complex legacy systems across organizations, outdated manual processes, limited interoperability and few end-to-end processes across human capital management functionalities. When not addressed, these have implications on recruitment, retention and skills training, not to mention the cost associated with the operation such legacy systems.

The boundless potential of generative AI could bolster Canada's economy by $210 billion, greatly boosting Canadian workers' productivity. It's important that organizations seek experienced industry partners that are equipped to guide operations and organizations through their digital transformations, leveraging technologies like AI to level up the workforce. At SAP, we see that potential and opportunity to unlock productivity and value across our economic sectors. For example, AI can address some of the top workforce challenges of our times from recruiting and training to increasing employee engagement and retention.

I'll run through a few examples. Recruiting AI software can remove unconscious biases in job descriptions. Recruiting automation can lighten the administrative burden by automating the delivery and receipt of necessary documents. Specialized AI-enabled training is interactive; it's continuously learning and adapting to each worker's learning style, whether it's visual, auditory or written. AI analytics, specifically sentiment analytics, can identify how workers are feeling. AI performance analytics allow managers to extract bias-free insights from continuous real-time assessments via multiple sources.

Another area where technology can support is accessibility. Software solutions can enable the inclusion of members of the disability community into today's workforce. As a co-founding member of the ministerial advisory board that established the Canadian business disability network, SAP advocates for the acceleration of the adoption of technologies that embed tools like AI to onboard members of the disability community into today's workforce.

Canada's potential in this space is vibrant and remains globally competitive, with a diverse AI ecosystem that attracts more AI talent and brings more women into AI-related roles than all of our G7 peers.

The high concentration of talent in Canada contributes to a rising volume of AI patents filed nationally and the highest number of AI publications per capita in 2022. It is even more important that public policy favour retention of top AI talent in this country to uphold our competitive edge and support sustained innovation.

The impact of AI to Canada's labour force remains undeniable, and public policy must allow for better digital integration with Canada's industrial base to strengthen our local ecosystem that is inclusive of SMEs, minority-owned businesses and indigenous businesses.

Mr. Chair, thank you. I'm happy to take questions.

[The Chair, Liberal, Bobby Morrissey]

Thank you, Ms. Lukasheh.

We'll now begin questioning with Mrs. Falk.

Mrs. Falk, you have six minutes, please.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

Thank you very much, Chair.

Thank you to each of our witnesses for taking the time to come here to share your experiences and thoughts regarding AI.

My first question would be for SAP.

You did mention a few side effects or outcomes of some of your clients using AI. I'm just wondering if you have more examples—either for the better or for the worse—of things they have experienced, anticipate or fear having to encounter, if that makes sense.

[Yana Lukasheh]

Absolutely.

I think you'll notice that within our customer base, they realize the value that AI brings into their business processes, and they see the value it can unlock.

I'll probably use, at a very high level, a few examples. Take banks, for example. They have a lot of financial reports and data that they have to manipulate through different data sources. AI can automate a lot of these tasks and summarize a lot of that data. That would provide a lot of efficiency for the workforce in that particular bank to dedicate the time to a lot more strategic work, instead of a lot of the data analysis.

Another example would be within manufacturing. Some of our customers are leveraging AI technologies to look at sales performances, identifying where the underperforming regions are, looking at their procurement and their supply chain, and looking at their HR and trying to find efficiencies across....

That's probably what I would give as an example.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

In your example of a bank, would the job being completed right now by a human then be displaced, or are some of your clients finding alternative work, still in the bank? Are we anticipating or seeing a job loss for a person?

[Yana Lukasheh]

I won't be able to speak on behalf of the banks, but what I can say is that, at SAP, we always view the necessity of the human in any of the work being done. We see AI as an augmentation tool, and not as a replacement for a particular job. This is the case for the various industries that we cover across the board, whether in Canada or around the world.

That brings an important question about how we support the employees to better skill them for the new tools they're going to be leveraging, to be able to use their time in more strategic ways that are linked to business processes.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

For sure. Thank you.

You also mentioned different industries. Would SAP say that they anticipate that there may be a different quantity of AI that would be used in different industries?

[Yana Lukasheh]

Are you able to explain that a bit better? I'm not sure I understand the question.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

Are there some industries that might be utilizing or may utilize AI more than other industries?

[Yana Lukasheh]

That will depend on the industry itself and how they intend to leverage the technology and AI.

What I can say for the industries that are using our human resource application—the software applications, as an example—is that the AI is already embedded in the tool, so it's being leveraged in the same way within the different industries that are leveraging that particular product or solution. I would say that.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

Thank you.

Are you able to speak at all about the impacts that AI would have on the working conditions of workers? Would the workload increase or decrease?

[Yana Lukasheh]

What we can say is that AI definitely optimizes a lot of the manual workloads that are currently being done by humans, so there's an efficiency gain that is being done there.

Again, it's not to replace that individual person, but really to make their job a little bit easier and have them concentrate on a lot more strategic work, rather than spend hours on automated work where they could be leveraging AI, which they can do in a span of minutes. ChatGPT is an example of that.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

Sure.

Would we find the probability of errors going down? If it's augmenting data, for example, would AI do it better than a human?

[Yana Lukasheh]

Given the fact that the AI technology is able to access data sets from different sources, it will definitely be able to do the job quicker and be less time consuming, because it has the end-to-end vision from the whole process. Whether the data is better or not, that is left to be determined by the user, but the human factor always has to remain there to be able to be that oversight, as well.

[Rosemarie Falk, Conservative, Battlefords—Lloydminster, SK]

Thank you.

[The Chair, Liberal, Bobby Morrissey]

Thanks, Mrs. Falk.

Joining us now, we have Ms. Nathalie Blais.

Mr. Coteau, you may go ahead for six minutes.

[Michael Coteau, Liberal, Don Valley East, ON]

Thank you so much, and thank you to the witnesses for being here.

This is our last day listening to witnesses. All of the witnesses have really contributed to a very interesting conversation on the subject. Many and various points have been brought forward that have complemented each other.

I want to speak to a point that's been brought up by more than one person, specifically about how machine learning works and how data.... I think you just referenced data coming from many, many different sources. If the data that we're using is building the AI through machine learning, there's no question that bias will be embedded into the technology we're building.

Technology mirrors society as a whole. Here's a good example. If AI were being used in the judicial system, it would look at, let's say, the last 70 years of court cases. If that were the case, and if we acknowledged that the AI would be built from that machine learning and datasets that have lasted the 70 years, we would now be making decisions based on that data, and there would be a bias embedded in it if we acknowledged that the system had systemic barriers in place.

The big question is this. I think Mr. Soucy brought up the fact that we need to be careful that the technology that we're putting forward doesn't set bias against some workers. I guess my question for the union representative is, how do we use the collective agreement process and how do we hold companies accountable when the datasets they're using are often in a black box-based information set that's not shared with the public? These algorithms are private.

How do we ensure that we can find a balance between what's being built and how it serves workers in general?

That question is for Mr. Soucy.

[Danick Soucy]

I'm going to let Ms. Blais answer that.

[Nathalie Blais, Research Representative, Canadian Union of Public Employees - Quebec]

Good afternoon.

Thank you for your question. It's a good one.

I was at a telecommunications symposium recently, and one of the issues discussed was how reliable AI systems were when trained on data that aren't entirely reliable. For example, when the Internet is used to train an AI system, it really captures everything out there, even though some of that information is false and some is true.

How do you make sure an AI system trained on those data is reliable?

When that question was put to business people in the telecommunications sector, they all evaded the question. The reason I'm telling you that story is that, afterwards, I spoke with the person moderating the panel discussion. She, herself, is a technology expert, and she said that the only way to make sure the data are high quality is to require companies to disclose where the data used to train their systems came from. Developers would have to tell companies purchasing AI software whether the systems were trained on data pulled from the Internet, private corporate data, academic data or government data.

[Michael Coteau, Liberal, Don Valley East, ON]

Thank you for that response.

I will move over to SAP, based on the response to the question I asked. If we're going to use technology like AI for recruitment and training, which you mentioned earlier in your testimony.... Part of a company's competitive edge is making sure that the algorithms and software it's using are private, because that's intellectual property. At the same time, we need to make sure that the datasets that are being used are fair and come from reliable places. Many big organizations like the Amazons and the Microsofts may not be unionized, so there's a disconnect with that collective agreement process.

How do we make sure that big companies like SAP are bringing forward AI based on machine learning that is equitable and transparent? How do we go about doing that? At the same time, how do you keep your competitive edge? That's a tough question, but maybe you have some thoughts on that.