HR leaders must ground recruitment analytics in explainability and workforce planning – not AI hype – to hire fairly and fast, says talent exec
Skills-based hiring is reshaping Canadian recruitment – but data science, not AI, determines whether it works. According to Robert Half Canada's Demand for Skilled Talent report published in July 2026, 53 per cent of Canadian hiring managers say finding qualified talent is harder than before, and 48 per cent have cancelled business projects outright because they lacked the skills to execute them. More AI isn’t the answer, according to one senior talent acquisition leader – the missing discipline is how data is used, interpreted, and legally defended.
"We're moving from a place where basically we had a job title, and the job title really told you everything you needed to know about that job," says Travis Windling, Senior Director of Talent Acquisition, Contingent, Operations and Strategy at Royal Bank of Canada (RBC) in Toronto. "The problem in today's workforce is that because everything is so cross-functional and intermingled, that job title doesn't really mean anything."
Skills-based hiring – prioritizing demonstrated capabilities over credentials such as degrees or previous job titles over simple job descriptions – has become a prominent discussion in HR circles. Windling unpacks what the term demands in practice for the talent acquisition function trying to overcome the talent challenge revealed by the Robert Half Canada report.
Understanding skills and organizational needs
Moving beyond the job title requires assessing candidates at two levels, according to Windling: technical skills, meaning what a person can functionally do; and what he calls "durable skills" or "human skills" – the deeper qualities that define how a person operates. "You need to get to that next layer of assessment in terms of how do I actually understand what this person functionally does, both from a technical skill-set perspective as well as what that person's superpowers are, effectively," he says.
That dual assessment requires an organization to understand its own demand before it can properly evaluate supply. Windling identifies workforce planning as the most common stalling point for Canadian employers trying to move from reactive hiring to proactive, pipeline-driven recruitment. Organizations exploring data-driven talent acquisition and workforce planning strategies will find this discipline consistently separates high-performing talent functions from those still operating on instinct.
"Unless you understand the demand of what you're looking for, you can't translate that through to the right supply and you're going to have a mismatch," says Windling. "Any kind of operational efficiency that you've gained is out the door because you're pipelining the wrong people and you're still moving in that reactive, just-in-time type model unless you get really clean on the workforce planning side."
The AI pendulum – and why it’s swinging back
Windling's view on AI in recruitment is measured for someone who oversees analytics, technology, and vendor management for one of Canada's largest employers. "I think we went to over-index on the AI predictive analytics fit scoring side of things," he says. "I would almost suggest that we're almost starting to see the pendulum start to swing back."
What concerns him isn’t technology itself, but the absence of understanding around it. For example, Ontario's amendments to the Employment Standards Act, 2000, in force since Jan. 1, 2026, require employers with 25 or more employees to disclose in every publicly advertised job posting whether AI is used to screen or assess applicants. For HR leaders wanting to understand the implications of Ontario's employment standards legislation on recruitment processes Windling says disclosure obligations are a floor, not a ceiling.
"That is effectively an automated decision in some sense of the word," he says about something as routine as sorting candidates by date of application. "Regulators are getting a lot smarter with the fact that we already do some of these things, and we just need to be able to articulate why we make the decisions that we make."
Windling believes this type of requirement will push HR back towards focusing on science over just AI, because AI models aren’t able to provide the transparency that’s increasingly being required legally.
What explainability actually requires
The solution Windling points to is not a retreat from data – it’s a return to the science underneath it. He argues that large language models (LLMs), which power most generative AI tools, are fundamentally unsuited to high-stakes hiring decisions.
"LLMs are notoriously unreliable in terms of giving you the same result over and over again from a scalable process perspective," he says. "We actually need to turn back to things like [industrial-organizational] psychology and the 70-plus years of research that we have in that discipline – this is what the actual role or skill set requires, this is what the person is driven and drained by, and then creating that match."
Industrial-organizational (IO) psychology applies scientific methods to the study of human behaviour at work, with a specific focus on selection, assessment, and performance. For talent acquisition leaders navigating the growing complexity of responsible AI use in Canadian workplaces, a statistically grounded, repeatable assessment model will be far more defensible than an opaque algorithmic score. “LLMs are great for certain use cases like summarization or theming – those types of things,” he says. “But in terms of actual decision-making, it needs to be rooted in statistics, math, repeatability, explainability – If you don't have that tie back, you're asking for trouble."
Bias, liability, and the accountability gap
The compliance dimension is growing. Windling points to litigation involving Workday's Human Capital Management platform – where an algorithmic screening tool was alleged to have produced age-discriminatory outcomes – as a signal of where employer liability is heading.
"It was looking at the top end of years of experience and demoting from a match perspective those people, as opposed to understanding the fact that years of experience correlates with years in the workforce and thereby age,” he says. “I might counsel my people to say there's a certain level of minimum experience that's required but never, ever set a maximum."
His broader warning is for HR leaders who have ceded process ownership to IT or internal configuration teams. "You are the process owner. You are the product owner of how you bring talent into the organization,” he says. “You need to understand how that talent is being assessed, and you need to be comfortable from a fairness and bias perspective, as much from a reputational risk as a regulatory risk perspective."
When guardrails are in place and accountability is set, AI-powered predictive analytics can be a useful tool in modernizing talent acquisition to be more focused on skills and strategy – but as a counterpart to human decisions, according to Windling. "Predictability is your friend, and having that level of explainability, transparency, and predictability is really where you want to over-index on using data, especially for a human-based process like recruiting," he says. “It doesn't have to be a fancy model, but in terms of actual decision-making, it needs to be rooted in statistics, math, repeatability, and explainability.”
