At HRFest Canada 2026, two executives shared what their AI rollouts cost – and what they’d do differently
The pitch is familiar: AI will save hours, eliminate friction, and transform how human resources gets done. At HRFest Canada 2026, two executives who have lived through real AI rollouts offered a less polished account of what implementation actually looks like.
Kalifa Oliver, Executive Advisor and Senior Director of Technology at Lowe’s Companies in Atlanta, and Diana Valler, Chief Human Resources Officer (CHRO) at HIS Canada in Toronto, joined a panel titled “The AI Implementation Files: Case Studies in Ethics, Risk and What Got Paused.” The session delivered something most AI conversations do not: an unvarnished account of decisions made under pressure and hard stops that rarely appear in vendor pitches.
Hype versus reality in the HR suite
“AI in the way we know it in HR right now is heavily marketed, and we all get to hear about is wins,” said Oliver.
Valler recalled the early appeal of the technology: “When first you heard about AI, you’re like, ‘Oh, this magical tool is going to be coming to simplify everything for us. It’s going to be saving us hours,” said Valler, adding that half an hour spent crafting the right prompt was often the more truthful experience.
The shift from consumer to collaborator changes what practitioners expect from the tool – and how they evaluate its output, according to Valler. “What about if you change the mentality and say, ‘Now I work with AI,’ because truly that’s what it is,” she said.
It is a question HR leaders have explored for years: whether AI can genuinely put the human element back into HR strategy.
Drawing the line between AI and human judgment
The session’s most practical question was where, exactly, AI belongs in HR work and where it does not – the key to determining responsible use of AI.
Oliver offered a framing that cut through the theoretical: “Would you let a three-year-old make a multimillion-dollar decision? Probably not. And that’s what we’re doing with this brand-new technology that can fail.”
Valler drew the same boundary from experience. Administrative tasks – drafting job descriptions, researching benchmarks, summarizing industry data – were appropriate use cases. Proprietary decisions were not. “When it’s about something proprietary, something where the critical thinking and the real human is required, that’s a definite no,” she said.
Valler cited recruitment screening as an example of the technology’s limits. AI platforms can match a candidate’s skill set against a job description with high accuracy, “but I challenge you: do they really have the cultural fit?” Cultural fit requires human perception – reading adaptability, appetite for learning, and interpersonal quality that no automated system can reliably assess, said Valler. Canadian employers navigating AI-driven recruitment analytics are being cautioned that skills-based hiring demands data science, not algorithmic guesswork.
What got paused – and why
Oliver described circumstances from previous roles where AI use was reversed. In one, she said she stopped leaders from accessing verbatim employee survey responses, as comments could expose individual identities, particularly in cases involving employee relations allegations. The solution was to redirect leadership attention toward themes, using AI to surface patterns without naming individual voices, said Oliver.
The second example was more costly. After broadly encouraging a team to use ChatGPT with minimal guardrails, costs mounted rapidly, according to Oliver. Employees were querying basic headcount data – information available on existing dashboards – at significant token expense. “We were about to kill our budget,” she said.
The result was a full halt and having to retrain employees on how to use AI, said Oliver. “And that was a massive pause because you’re at a company that’s like, ‘Let’s use it, let’s embrace it, let’s do it,’” she said. “You have leaders who are saying, ‘Do it, do it, do it,’ and then you have to make this hard decision to say, ‘Stop.’”
Valler said her governance response was to restrict staff to approved platforms only, paired with joint HR and IT training covering which tools were approved and why. Research on how AI tools are adding to HR team workloads shows a widening gap between adoption speed and governance readiness. “It’s not an easy decision because it goes a bit against the experiential,” said Valler.
Skills, cognitive offloading, and the recruitment reckoning
As AI absorbs transactional HR tasks, Oliver argued that what organizations traditionally classify as soft skills are being reclassified into critical skills. As routine tasks are delegated to AI, HR practitioners must bring more sustained judgment to their remaining work – raising both the cognitive load and the stakes, she said.
The inverse risk – cognitive offloading – emerged as a parallel concern. Oliver said she observed it directly after her team removed download access from HR dashboards. Requests that followed revealed employees had been exporting data specifically to feed it into AI, rather than reading it themselves. “It’s really scary in HR, where people’s lives are literally in your hands,” she said, “Every data point is a human, and I can’t make mistakes like that.”
Valler’s closing advice for any organization beginning the AI journey: “Start with the people and the purpose of the company you lead, and then technology comes.”
Oliver’s final test was simpler: “If I remove this AI tool tomorrow, would you even notice it’s gone?”
