A junior employee prepares a first draft. A supervisor marks it up, explains what’s missing and asks for another version. The process takes longer than anyone would like, but something valuable happens between the first attempt and the final approval. The employee learns how the business thinks.

Now introduce an AI tool that produces a serviceable draft in seconds. The immediate benefit is easy to see. The developmental consequence is harder to measure, especially if the employer decides it no longer needs as many junior employees.

For Canadian HR leaders, that’s a workforce planning risk worth bringing into the next budget discussion. Employers can improve productivity through automation while gradually removing the experiences that develop their next generation of capable employees. When an experienced person leaves, the business may discover that it has streamlined its work and weakened its succession pipeline at the same time.

The answer isn’t to preserve every repetitive task. It’s to make the development of judgment an explicit part of work redesign, with an owner, a budget and evidence that people are becoming capable. Otherwise, today’s efficiency savings can become tomorrow’s recruitment premium.

The adoption decision is already a workforce decision

Statistics Canada reported that 25.2% of businesses planned to use AI to produce goods or deliver services over the next 12 months in the third quarter of 2026, compared with 14.5% a year earlier. Those figures measure intentions, not completed deployments or job losses. They nevertheless show why HR needs to participate before technology decisions become approved staffing assumptions. [1]

A vendor proposal may describe hours saved. Finance may translate those hours into avoided hires. Operations may assume the remaining employees can absorb the exceptions. None of those assumptions answers a crucial question: how will the organization develop people who can handle the work the system can’t?

Consider an accounts payable employee who learns to recognize suspicious invoices by reviewing ordinary ones, or a customer service representative who develops judgment through increasingly difficult conversations. An analyst’s first assignments may seem basic to a senior colleague, yet they establish the context needed to challenge an implausible result later.

Remove the routine work and the learning process changes. That can be beneficial if the employer deliberately creates a better one. It becomes dangerous when the organization assumes development will continue automatically because the job title still exists.

AI can accelerate learning when employers design for it

There’s credible evidence that AI can help less experienced employees perform better. In the 2023 working paper Generative AI at Work, researchers studied the introduction of an AI assistant among 5,179 customer support agents. They reported a 14% average increase in issues resolved per hour, with a 34% improvement among novice and lower-skilled workers. Results varied substantially by experience and skill. [2]

That finding supports a more useful conversation than a blanket prediction that AI will eliminate junior work. A tool can distribute useful practices and help employees handle tasks sooner. It may become part of a stronger learning environment.

However, performance with assistance and independent understanding are different things. An employee may select a good suggested response without recognizing when the next suggestion is inappropriate. The research doesn’t establish that every AI deployment builds durable expertise, and its results shouldn’t become a universal productivity assumption for Canadian workplaces.

HR should therefore ask two questions during a pilot. Does the tool improve the employee’s work? And can the employee explain, verify and appropriately challenge that work? The first question measures operational assistance. The second tests whether the organization is also developing capability.

Find the learning hidden inside ordinary tasks

Traditional job descriptions usually explain responsibilities, reporting relationships and qualifications. They rarely explain how someone learns to become good at the job. That missing information matters when an employer starts removing tasks.

Ask experienced employees to identify the situations that changed their understanding of the work. What did they repeatedly encounter before they could recognize a problem quickly? Which mistakes required a supervisor’s explanation? Where did they learn that the written procedure didn’t cover every circumstance?

An insurance administrator might point to inconsistent client information. A maintenance employee might describe a sound that indicated an equipment fault. An HR coordinator might recall a leave request that required escalation rather than a standard reply. These examples reveal learning opportunities that a task inventory alone can miss.

For each activity proposed for automation, record its operational purpose and its developmental purpose. Producing a report may also teach someone how the numbers connect. Handling a complaint may build knowledge of customer commitments. Reviewing documentation may teach employees what an incomplete record looks like.

Some tasks will have little developmental value and can disappear without much consequence. Others should change rather than vanish. Employees might review selected automated outputs, reconstruct the reasoning behind a recommendation or investigate why the system handled two apparently similar cases differently.

The savings can look larger than they are

Imagine a Canadian service business with 180 employees. It plans to introduce automation across an administrative team and leave three junior vacancies unfilled. Each position has an assumed annual employment cost of $65,000, including salary, employer contributions and benefits. The apparent annual saving is $195,000.

This is an illustrative scenario, not a reported company case. All figures are Canadian dollars and are planning assumptions, not market benchmarks.

The business expects annual software costs of $30,000. During implementation, it also uses 400 hours of experienced employee time at an assumed loaded cost of $70 an hour, or $28,000. If a supervisor spends six additional hours a week reviewing exceptions across 48 weeks at $80 an hour, that adds $23,040 of allocated capacity cost.

After those items, the first-year benefit is $113,960 before any other implementation, quality or transition costs. That can still be an attractive result. But it’s a different result from $195,000, and it depends on whether the organization can genuinely remove the positions while maintaining service.

Two years later, a senior employee leaves. Suppose the absence of an internal successor results in $18,000 of recruitment expense, 300 hours of temporary specialist support at $140 an hour and 200 hours of internal onboarding support at $70 an hour. Those assumed items total $74,000, before any service disruption.

HR shouldn’t attribute that entire amount to automation without evidence. The departure might have created costs under either staffing model. The relevant question is the incremental exposure created by removing development capacity: would the earlier structure have produced someone ready to step up, and what would maintaining that pathway have cost?

That’s the comparison executives need. Present cash expenditure separately from the value of employee time, avoid counting the same hours twice and show the assumptions that would change the conclusion. A credible business case can support automation while rejecting an unrealistic staffing saving.

Protect development through better work design

Preserving learning doesn’t require assigning people obsolete work simply because their managers once did it. A stronger approach identifies the experience employees need and builds a deliberate route to acquiring it.

For a junior analyst, that might mean examining the source data before reading the AI summary, checking selected calculations and explaining which conclusions the evidence supports. For customer service, it might involve supervised exposure to exceptions and a review of why a suggested response was rejected. In maintenance, it could mean diagnosing a controlled fault with a qualified colleague before comparing the diagnosis with the system’s recommendation.

Historical cases are useful if they’re appropriately de-identified and approved for training. Employees can examine incomplete information, make a decision and then compare their reasoning with the eventual outcome. Include ambiguous cases where escalation was the correct response. Otherwise, the exercise may reward confidence rather than judgment.

Managers should also create a progression in responsibility. Observation leads to practice, practice leads to supervised decisions, and demonstrated competence leads to a defined level of authority. The progression should be explicit enough that employees understand how to advance and supervisors know what evidence they need before signing off.

Where work is regulated or safety critical, simulation and AI assistance don’t replace required qualifications, competent supervision or prescribed training. Development must fit the actual work and applicable requirements.

Budget for the people who teach

One common weakness in a reskilling plan is the assumption that experienced employees will teach while maintaining an unchanged workload. A manager is assigned responsibility for the transition, a senior specialist is asked to coach and both retain their existing targets.

The result may be rushed review, delayed feedback and growing resentment. The employer then sees weak development outcomes and concludes that junior employees lack initiative. In reality, it may have created a learning programme without providing the capacity to deliver it.

Cost the teaching work. Identify who will design practice cases, review decisions and assess readiness. Estimate the hours required, adjust workloads and decide which responsibilities can move elsewhere. Include backup coverage so the programme doesn’t depend on one generous employee staying late.

Recognition matters too. If promotion and performance measures reward only individual output, coaching will compete with the behaviour the employer pays for. Managers need clear expectations that developing capable colleagues is part of successful performance, supported by time to do it properly.

Canadian employers need more than one pathway into the work

The consequences of removing entry roles depend on the employer’s circumstances. A business in a regional market may have few nearby candidates for a specialist vacancy. A smaller employer may rely on one experienced person rather than a deep internal bench. A multi-site organization may have expertise in one location that’s difficult to transfer to another.

These are planning considerations to investigate locally, not reasons to assume every Canadian employer faces the same shortage. HR should test the actual hiring market, the availability of remote work, qualification requirements and the time it takes to develop competence in its own operations.

Career pathways deserve the same attention. Entry-level work can provide a route into an occupation for graduates, career changers and newcomers whose experience employers haven’t yet recognized. If the employer removes that route and continues demanding several years of directly comparable experience, it may narrow its own future applicant pool.

A practical response is to assess transferable capability through relevant work samples. Give candidates a realistic problem, provide the information needed to attempt it and assess their reasoning using consistent criteria. Don’t confuse familiarity with a particular software product with the ability to understand the work.

Existing employees should have access to those opportunities as well. Publish development assignments, explain selection criteria and make it possible for shift workers and employees outside head office to participate. Review who gets practice time, who receives coaching and who is approved for progression. Access to development can become an unseen source of workforce inequality if managers distribute opportunities informally.

Keep a human review that can change the outcome

A staffing plan can look safe because a person remains responsible for reviewing the system’s work. That safeguard is meaningful only if the reviewer has the competence, information, time and authority to intervene.

An employee who approves hundreds of recommendations under a demanding throughput target may have little practical opportunity to investigate them. A junior worker may also hesitate to question a system that management has described as more accurate than people.

Define what review involves. Specify which records must be checked, which conditions require escalation and who can stop or reverse a decision. Build time for that work into capacity calculations. Then examine whether employees use the escalation route and whether managers respond constructively when they do.

Privacy is part of competence. Canada’s privacy regulators have made clear that generative AI operates within existing privacy frameworks and that organizations must comply with applicable laws. HR should ensure employees understand what information they may use in approved systems and where to obtain help. Obligations vary with the organization and its activities. [3]

A realistic training case might ask an employee to improve a draft without entering identifiable medical or performance information into an unapproved tool. The test is whether the employee can complete the work appropriately, rather than merely repeat a rule from a course.

Run a talent pipeline stress test before reducing junior hiring

HR can make this practical by choosing one job family affected by a planned technology investment. Work with its manager and experienced employees to trace how someone currently progresses from first assignment to independent performance.

Identify the tasks scheduled to disappear and mark which ones provide important learning. Then identify the capabilities the organization will still require, the employees who currently hold them and the evidence that someone else could take over. A name in a succession spreadsheet isn’t enough. Ask what that person has successfully done and what supervision they still need.

Test a realistic disruption. If the most experienced employee became unavailable for three months, who could manage the exceptions? Which work would stop? What external support would be available, at what cost and with what lead time? The exercise should reveal an operational dependency, rather than attempt to predict individual departures.

For each significant gap, decide whether to retain a development position, redesign assignments, create a rotation, document and practise a process, or establish external support. Assign a manager and a review date. A small employer might complete this exercise in a working session; a larger one may need separate reviews for critical functions.

Before approving the staffing change, take five questions to the executive team:

• Which capability are we removing from the payroll, and which capability are we still depending on?
• What experience will prepare someone else to handle that work?
• Who has time and accountability to teach and assess it?
• What evidence will show that the new pathway works?
• What will we do if development takes longer than the business case assumes?

Measure whether the next person is becoming ready

Training completion provides limited evidence of succession readiness. Track how long employees take to perform defined work at an acceptable standard, how often their decisions require correction and whether they escalate appropriately.

Combine speed with quality. A faster employee who misses material exceptions hasn’t reached the same capability as someone who produces reliable work. Use a range of cases and have supervisors explain their assessments so employees receive actionable feedback.

It can also be useful to test selected tasks without assistance in a controlled setting, where safe and relevant. The purpose is to establish what employees understand and whether the team can operate during an outage. Don’t turn the exercise into an unrealistic ban on tools that are integral to the job.

Watch the experienced employees’ workload as closely as the juniors’ performance. Rising review hours, recurring corrections and delayed approvals may indicate that the employer has moved work rather than removed it. Those signals should trigger adjustments to staffing or training before service deteriorates.

Make capability part of the investment decision

HR’s most valuable contribution is to connect today’s work redesign with tomorrow’s operating capacity. That requires a clear view of how people learn, what expertise the business depends on and what replacing that expertise would involve.

An employer may conclude that fewer junior positions make sense. It should reach that conclusion with a funded development pathway and credible coverage for critical work, rather than assuming that future expertise will be available when required.

The next time an automation proposal promises an avoided hire, ask the team to show how the next capable employee will develop. A convincing answer makes the investment stronger. An unanswered question belongs in the business case before the saving belongs in the budget.