ADP Research is working with Stanford to unbundle jobs into specific tasks. What have been the most surprising findings so far regarding which tasks are more automatable than expected and which ones aren’t? Given Canada’s heavy reliance on resource, financial services, and public sector work, do those findings hold up the same way for Canadian job categories?
Naomi Lariviere: ADP Research’s “Unbundling” Project with Stanford is underway with the goal of unbundling jobs into discrete tasks and, using ADP payroll data, assign a value to each of these responsibilities. AI is not so much about job creation and destruction, but rather how it is reshaping work at the task level. As AI augments certain job responsibilities, which tasks are becoming the most valuable because they require the creativity, strategy, judgment, and critical thinking that AI can’t replicate? ADP’s massive reach and anonymized workforce dataset makes this possible because we not only have the data on job postings and the skills that comprise that posting, but then we can assign a wage value for each of those tasks because ADP pays so many workers.
We’re uniquely positioned to measure how AI is changing the world of work and help both employers and employees adjust accordingly. For example, employers will not just be able to wholly focus workers on their most high-value tasks, but also give workers a clearer picture of how AI can be working alongside them in more of a complimentary role while they zero in on the work that requires uniquely human attributes like judgment and strategic thinking. It can also help companies with workforce planning because breaking jobs into a bundle of tasks makes it clear where there are opportunities for transferrable skillsets. For example, a marketing professional might have some skills that the company needs in the legal department.
For employees, they’ll have more information to make decisions on education and career choices, in addition to transition options. What are the most high-valued skills as AI automates others and how can I hone my expertise in those areas? Where might my most high-valued skills be useful in a different department?
For workers, their concerns around AI are real and understandable. When people hear conversations that focus only on automation or efficiency, it’s natural to wonder what that means for their role or their future. The Unbundling project can absolutely be a tool for workers as they navigate this next era of work. AI can be incredibly useful in reducing repetitive administrative work, helping people find information faster, or supporting better and quicker decisions. However, that’s very different from replacing human judgment. This is especially true in areas like HR and payroll where the human side still matters a lot.
These are sensitive moments in people’s lives. Pay, benefits, leave, career decisions, workplace issues. People want accuracy, but they also want accountability and trust. I also think employees can tell pretty quickly when AI is being introduced in a way that genuinely helps them versus when it’s treated purely as a cost-cutting exercise.
The organizations getting it right are usually the ones being transparent, involving employees in the conversation, and thinking carefully about how the technology actually improves the experience of work, which is core to the Unbundling project. In Canada especially, where there’s often a strong focus on worker protections and governance, those conversations become even more important. I think that’s healthy. AI adoption works better long term when people feel informed, respected, and included in the change.
You mention that 95% of enterprise AI implementations fail. What’s the single most common mistake you see companies make, and what does a successful AI rollout actually look like in practice? Is that failure rate any different for Canadian businesses where SMEs dominate the economy and AI adoption has lagged compared to the US?
Naomi Lariviere: One of the biggest challenges right now is that many organizations feel intense pressure to “do something with AI” or position themselves as “AI-first,” and that urgency can sometimes push companies to move faster than their operational readiness. The biggest mistake I see is organizations jumping too quickly into the technology without fully understanding the problem they’re trying to solve, the quality of the data behind it, or how the work actually happens today.
In practice, a lot of companies are trying to layer AI onto processes that are already inefficient, inconsistent, or overly complex. And if you automate a bad process, you usually just create problems faster.
The organizations getting the most value from AI tend to be much more disciplined. They spend time understanding where people experience friction, where AI can genuinely improve outcomes, and where human judgment is still essential. They also recognize that good AI outcomes depend heavily on good data and a clear understanding of how work is performed.
That’s especially important in environments with complex regulatory requirements, lean operating models, or limited resources, which is true for many organizations globally, including Canadian small businesses. In those environments, AI adoption tends to be much more grounded in practical business outcomes than innovation for innovation’s sake.
In areas like HR and payroll, that distinction really matters because the margin for error is very small. These systems affect people’s pay, benefits, compliance, and livelihoods, so trust and consistency are non-negotiable. At ADP, our trusted service model pairs AI-driven efficiency with expert human judgement. Nearly 77 years of hands-on experience with HR processes, workflows, and regulatory nuance built into our products, services, and systems. ADP Assist, our intelligent, chat-based HR assistant, is grounded in this institutional knowledge.
With ADP paying 1-in-5 US workers, you have a macro-level view few companies can match. What signals in payroll and HR data are already showing the early impact of AI on the workforce? What does that same data show specifically in Canada and are there sectors here, like financial services or public administration, where the early signals are particularly pronounced?
Naomi Lariviere: Looking at the most recent ADP National Employment Report, 122,00 private-sector jobs were added in May, which is the strongest growth since January 2025 and followed up on more than 100,000 jobs being added in April as well. So strictly looking at job creation via the ADP data, there’s nothing right now that would point to widespread AI job displacement.
In August 2025, ADP Research collaborated on a research project with Stanford using ADP payroll data that found signals of AI job displacement in early-career professions that are most exposed to AI automation, like software developers and customer service reps.
Beyond those anecdotes, one of the things ADP Research has been looking at is the gap between a lot of the public conversation around AI and what people are actually experiencing at work. The findings, published in ADP Research’s Today at Work report, are very interesting from an employee experience perspective. 30% of workers who use AI daily in their work were fully engaged, compared to only 14% of those who never use AI. Only 11% of daily AI users reported feeling overloaded, compared to 23% of non-adopters.
Workers who use AI frequently were also more likely to say with confidence their job was safe from elimination. At the same time, daily users of AI at work were four times more likely to be feeling less productive, which is a very interesting finding as AI resigns workflows and automates tasks that might have given a more concrete sense of accomplishment. In Canada specifically, ADP Research’s analysis found 13% of workers use AI nearly every day, while 28% use it multiple times a week and 22% have never used it. In the US, those percentages are 19%, 24%, and 23%, respectively. At Canadian small businesses with 1-9 employees, 15% of workers are using AI every day, while 34% have never used it. At those with 10-49 employees, 10% are using it everyday while 31% have never used it.
Beyond what’s showing up in the ADP data, trust still matters a lot. People want to understand how AI is being used, where human oversight exists, and how decisions are being made, especially in areas connected to work, pay, performance, or career growth. Work is also starting to shift in smaller, practical ways. Organizations are looking more closely at where employees spend time, where work gets bogged down, and where AI can help simplify tasks or reduce administrative friction without removing the human element. People want workplace technology to feel easier and more intuitive.
They want faster access to information, less time spent searching for answers, and fewer repetitive tasks getting in the way of meaningful work. You can also see strong interest in AI across industries with a lot of regulation, documentation, and knowledge-based work, including areas like financial services and public administration. In many of those environments, the real opportunity is helping people navigate complexity and spend less time buried in manual processes.
To me, one of the biggest shifts already underway is that AI is changing what employees expect workplace technology to do for them. There’s a growing expectation that these tools should help people work smarter, move faster, and remove some of the everyday friction from work. A lot of this falls to the employer to be able effectively leverage change management communications so that AI workforce integration is transparent and effective.
The employer also has to upskill their workers so that they can get the most out of AI at work. ADP Research found 17% of Canadian employees strongly agree their employer is investing in their skills development, while 21% of Canadian employers strongly agree they currently have the skills for the future of work, so there is some room for growth to prepare workforces for the future of work.
Ethical AI is easy to talk about but hard to operationalize across 140 countries with different labour laws and cultural norms. How does ADP actually build guardrails into its AI systems? Canada has its own distinct complexities: federal vs. provincial labour law, Quebec’s language requirements, and PIPEDA data privacy obligations. How does ADP navigate that specific layer, and where does Canada sit in terms of regulatory readiness compared to other markets?
Naomi Lariviere: Responsible AI sounds straightforward when people talk about it at a high level, but I think in practice it’s a lot more complicated. At ADP, trust and responsible AI are at the core of everything we do. Our clients have relied on us for decades of change and complexity, and this next era of work is no different. Leveraging human oversight in everything we do helps ensure fairness, transparency, and fairness are never compromised.
In HR and payroll, these systems deal with very personal parts of people’s lives. Their paycheck. Their benefits. Their time off. Their career. So people need to trust that the technology is accurate, secure, and being used responsibly. That’s why responsible AI cannot just be a set of principles on a website. It has to be built into how the technology actually works day to day.
A lot of it comes down to basics done well. Making sure sensitive data is protected and anonymized. Putting guardrails around how information is used. Continuously checking systems for accuracy, bias, or unexpected results. And making sure humans are still involved in important decisions that affect employees.
Context also matters a lot. AI in the workplace cannot give useful answers if it does not understand things like company policies, labor laws, payroll rules, or what a person is actually trying to do in that moment. That becomes even more important in places where the rules are more complex, including many Canadian organizations managing different provincial requirements, privacy expectations, and bilingual environments. At the end of the day, people are not looking for AI that feels impressive. They want technology they can trust when something important is on the line.
A lot of AI conversation centers on business efficiency. But for the average worker, AI can feel more like a threat than an opportunity. What’s your message to workers who are worried about what AI means for their jobs? In Canada, where unionization rates are significantly higher than in the US, how does that message land differently with organized labour and is ADP engaging directly with unions as part of responsible AI deployment?
Naomi Lariviere: A lot of workers are not questioning whether AI is coming into the workplace. They’re questioning whether they’ll have visibility into how it’s being used and whether they’ll have a voice in how the change affects them. That’s an important distinction. Most people can adapt to new technology. What creates anxiety is when change feels unclear, rushed, or done without transparency.
The organizations navigating this well are usually the ones having more open conversations early. They’re explaining where AI is being used, where human oversight still exists, what skills employees may need to build over time, and where they still want people deeply involved in decision-making and customer interactions. In HR and payroll, trust matters a lot because these systems touch highly personal parts of people’s lives. Employees want to know there’s still accountability behind the technology and that important decisions are not being handed over blindly to a model.
In many Canadian workplaces, there’s already a strong culture around employee participation, governance, and worker protections. So naturally, there tends to be a lot of focus on transparency, fairness, and making sure people are included in the conversation as workplace technology evolves. The companies that will get the most value from AI long term are probably not the ones moving the fastest, but the ones building the most trust while they move.
