The first wave of widespread AI adoption rewarded speed and experimentation. Now, organizations have enough experience to ask harder questions about results. Recent research shows employers reconsidering AI-led workforce cuts, encountering unexpected costs and human oversight requirements, and putting greater emphasis on work design, skills and measurable business outcomes. The conversation around AI is becoming more deliberate, and that may be a sign of progress.
For the past several years, business leaders have heard a consistent message about artificial intelligence: move quickly or risk falling behind.
Organizations experimented. Employees adopted new tools. Leadership teams launched pilots and invested in automation. In some cases, businesses went further, making workforce decisions based on expectations about what AI would soon be able to do.
We are now far enough into widespread adoption to see what happened next.
Recent research from Careerminds offers a useful look at what organizations are learning as AI moves beyond early experimentation and into broader implementation. In its February 2026 study, the company surveyed 600 HR professionals who had made layoffs during the previous 12 months. More than three-quarters said their organizations had eliminated roles because of technologies such as AI.
With the benefit of hindsight, very few said they would make those decisions the same way again. Less than 9% said their AI-driven restructuring delivered what had been promised and that they would repeat the process unchanged. More than 90% would make at least some changes to their approach.
The Careerminds findings are particularly notable because they are not appearing in isolation. We reviewed additional research from Harvard Business Review, Orgvue and others and found similar patterns emerging: organizations are reassessing assumptions about workforce reduction, recognizing the continued need for human oversight and placing greater emphasis on work design, reskilling and measurable business outcomes.
What does this mean for the average Canadian organization?We shared the findings with colleagues across Stratford who work closely with AI strategy, organizational design and intellectual property to get their perspective on what the numbers may be telling us. Across those conversations, one shift stood out: as organizations gain more practical experience with AI, the questions they are asking are becoming more deliberate.
The savings were not always where organizations expected themWhat we can learn from the Careerminds research is that replacing work with AI did not necessarily remove the need for people. More than half of respondents (54.6%) said their organizations found themselves providing more human oversight of the technology than expected. Nearly a third reported losing critical skills and expertise when employees left, and 28% found that the remaining workforce could not fill the resulting knowledge gaps. Many employers then began hiring again. Among organizations that rehired for eliminated roles, 52% did so within six months. Financially, only about 27% of respondents said their organization ultimately came out ahead after accounting for the cost of bringing roles back. Roughly 31% spent more rehiring than they had initially saved, while another 42% effectively broke even. That does not diminish AI's potential. It changes the conversation about how that potential should be evaluated. For leadership teams, cost reduction is one possible outcome of AI adoption. It is a narrow measure, however, if it is considered without looking at customer experience, capacity, quality, organizational knowledge, risk or the ability to execute the company's strategy. |
About the research: These findings come from separate studies with different samples, geographies and methodologies. Percentages should be interpreted within the context of their original studies and not combined into a single dataset. |
|
“We’re seeing leadership teams move from asking, ‘Where can we use AI?’ to asking, ‘Where will AI materially improve an outcome we care about?’ That is a much more useful conversation. A compelling new capability can still be the wrong investment if it doesn’t help the organization execute its strategy or solve a problem that actually needs solving.” - Majd Karam Senior Stratford Consultant & Technology Enthusiast |
The Careerminds findings echo a broader shift in how organizations are thinking about AI and their people.
Research released by Orgvue in 2025 found that 39% of surveyed business leaders had made employees redundant as a result of AI adoption. Among that group, less than half of those leaders (45%) felt that this was ultimately the correct decision. At the same time, 80% planned to reskill employees to use AI and 41% had increased learning and development budgets.
Harvard Business Review made a similar argument in an article published August 28 of this year, warning that some organizations are moving faster on workforce reductions than the evidence supports and encouraging leaders to focus on redesigning work around AI.
An employee's role is rarely a collection of isolated tasks. It also contains judgment, relationships, institutional knowledge, exception handling and responsibilities that may be difficult to see until they disappear.
AI may eliminate the need to perform some activities manually. Leadership still has to decide what happens to the time that creates, how the surrounding role should evolve and what capabilities the organization will need next.
|
“The next phase of AI adoption will not be defined only by the technology organizations put in place, but by how well they redesign work, build new capabilities, support leaders and employees through change, and measure whether AI is actually improving performance.” -Pierre Cote President, Stratford People & Culture |
|
Careerminds found that 55% of organizations had not formally discussed reskilling or redeployment before making their workforce changes. More than half later believed at least some of the eliminated positions could potentially have been transitioned into other roles.
Those findings suggest an opportunity for leaders to broaden the AI conversation before making permanent organizational decisions.
The technology decision is only one part of the decision. The same discipline applies beyond workforce planning.
AI changes how information enters and moves through an organization. Employees may be entering confidential information into third-party tools, using AI to develop new ideas, working with generated material whose origins are unclear or incorporating AI into research and product development.
As adoption expands, those questions become increasingly connected to intellectual property strategy.
|
“Every time an organization changes how work gets done with AI, it should also be asking what happens to its information, know-how and ideas along the way. What are we putting into these systems? What are we getting back? What valuable knowledge are we exposing or creating? AI decisions are increasingly business strategy and IP strategy decisions at the same time.” - Myriam Davidson Vice President, Stratford Intellectual Property |
The goal is not to make every AI initiative cumbersome. It is to understand the implications before widespread use makes them harder to unwind.
There is little evidence that organizations are abandoning AI. But, recent research instead suggests that expectations are becoming more grounded.
Deloitte's 2026 research found that productivity benefits are widespread, yet only 30% of organizations are redesigning key processes around AI. Another 37% remain at a relatively surface level of adoption. Deloitte has described the resulting challenge as “pilot fatigue,” with clearer strategy becoming increasingly important as organizations try to move useful ideas into production.
Here in Canada, BDO reported in June that 46% of Canadian business leaders are experimenting with AI without seeing meaningful ROI, while just 18% are actively embedding it into workflows and operations. Its findings point toward governance, workforce readiness and clear business outcomes as important factors in moving beyond experimentation.
That may be where the next phase of AI adoption begins.
Leadership teams have had several years to ask what the technology can do. They now have more evidence to help them decide where it deserves attention.
That means examining the work before automating it. Understanding the business outcome before selecting the tool. Considering what new capacity could make possible, rather than treating reduced headcount as the default measure of value. And being willing to conclude that some AI opportunities are worth pursuing while others can wait.
There is no advantage in chasing every new capability simply because it exists.
The organizations that get lasting value from AI will need to become increasingly selective about the problems they ask it to solve.
Stratford's AI Impact Assessment helps leadership teams examine where AI could meaningfully support their business objectives, where organizational readiness or risk needs to be addressed, and which opportunities warrant further investment.
Only 47% of Canadian Leaders Feel Confident in Their AI Strategy (Deloitte)
Is Stratford’s AI Impact Assessment Right for You?