‘You can't just take tasks in and out without changing the connected systems’: researcher on why Canadian employers risk costly rehiring cycles after AI cuts
When it comes to automation, new research suggests that the bluntest decisions may be the costliest.
Gartner recently predicted that half the companies that have cut staff due to AI will be hiring them back by 2027, “to perform similar functions, but under different job titles.”
This comes after realizing that the gaps in skills – such as “expertise, empathy, and judgment,” according to Emily Potosky of Gartner – left by laid-off humans weren’t necessarily filled by their AI replacements.
As Lisa Cohen, associate professor of organizational behaviour at McGill University’s Desautels Faculty of Management, explains, most AI systems only replace slices of work rather than whole jobs. What this means for employers is that before laying people off because of "AI efficiencies,” they need a solid plan for what happens to the tasks, the people and the pay bands that remain.
“AI is really important. We may or may not overestimate how much it's going to affect the world of work, but we finally figured it out that AI does not destroy jobs, it destroys tasks,” Cohen says.
“It also can create some new tasks. It's going to be relatively rare that AI can do everything in a job, and as a result, it is going to end up being humans who have to make decisions about what they are going to do with the tasks and jobs that remain.”
Job redesign: focus on tasks, not titles
Cohen’s research focuses on how tasks are bundled into jobs and how jobs are bundled into organizations. In practice, she says, job design tends to be messy and incremental, rather than the deliberate, top-down exercise that many textbooks imply.
In one study, Cohen and co-author Sara Mahabadi of the University of Alberta discovered that jobs will evolve even during the hiring process as candidates are interviewed and tasks are added or eliminated according to their skills.
“I think that the reality of how job design happens is different than how job design should happen,” Cohen says.
“The reality is, you just pick up a lot of tasks that are laying around, and you group them together because you think they're similar, or because you think that people with similar skills can do them.”
She gives the example of “data entry and data collection” ending up in the same role simply because “one person can do both of them.”
In a more deliberate process, she says, managers would map all of the work the organization does, understand which tasks sit where, and then decide – in light of AI and other changes – how that work should be grouped.
Let employees craft AI-era jobs
In classic job design theory, Cohen explains, managers assemble tasks into roles to achieve goals like efficiency, control and employee satisfaction. But she points out that the people doing the work also reshape their jobs over time, faster than ever now because of the rate at which AI is changing their jobs and duties.
Given how quickly AI tools evolve, Cohen is skeptical that any central HR or leadership team can keep up with job design and descriptions.
To solve this problem, she says it’s time for employees to be given “permission” to craft their own roles, and for managers to have the tools to make that happen. That also means paying attention to what tasks employees have quietly added or dropped, where there are hidden bottlenecks, and where initiative is already reshaping roles.
“What you can do is help the people on the floor, the line managers and the employees, give them tools to think about how to design their jobs,” says Cohen, and asking them directly what work they think would be valuable to take on.
In a paralegal example, she says, that might mean asking directly: “You have this job as a paralegal, you are no longer going to be doing a contract. What is it that you think you are going to do instead? Is there other work that you've been wanting to do?”
From redesigning jobs to redesigning systems
Going beyond job redesign, Cohen explains how decisions at the task level require organization-wide systems planning. Again using a legal firm as an example, she describes the knock-on effects of reassigning paralegals to different work: “If AI in a law firm takes away the basic tasks from paralegals of writing contracts, what are they going to do instead? Where is that work going to come from?”
This also has direct implications for headcount, pay and hiring, bringing up serious questions about what tasks are valued over others, and how job redesign alters compensation frameworks.
For example, she poses the problem of what to pay someone who is doing less or relatively easier work due to AI: “Should they have a lower salary?” The same question goes for an employee whose lower-level tasks have been replaced by AI.
“Maybe they should get more money, and maybe you should rethink your hiring practices for those jobs, who you should hire, because it's no longer a medium-level job, it's now a high-level job,” Cohen says.
“So you need to think [it] through. You can't just take tasks in and out without changing the connected systems.”
