The implementation conversation often starts too late.
An organization selects a tool, configures a pilot and then asks, “How do we get people to use it?” By that point, employees may already have concerns about their roles, unclear expectations or workarounds that make the new process harder to adopt.
Canadian data shows that AI is already changing how organizations manage people.
Among businesses that reported using AI in the previous 12 months, Statistics Canada found that 44.4% had made changes to training or staffing practices because of AI. Nearly one-third trained existing employees and more than one-fifth trained existing executives. Larger organizations were even more likely to invest in training.
That is a reminder: adoption is operational change.
AI can change how information is created, reviewed and approved; which tasks are automated; how employees spend their time; what skills managers need; and who is accountable when an output influences a decision. Those changes need to be designed, not left to emerge accidentally.
New Brunswick’s people-first pilot reinforces the point.
The McKenna Institute and Wallace McCain Institute structured their 2026 SME pilot around leadership, organizational readiness, implementation support and peer accountability. The program’s premise is practical: the technology only creates value when the organization is ready to absorb it.
Start with the workflow, not the job title.
Broad statements such as “AI will replace this role” or “AI will make everyone more productive” are usually too vague to manage. Break the work into tasks. Which tasks involve searching, drafting, summarizing, classifying, analysing, deciding, communicating or executing? Which could be assisted? Which should remain fully human? Which new review or exception-handling tasks will be created?
Design the new workflow with the people doing the work.
Employees often know where the real friction sits: duplicate entry, unnecessary approvals, missing context, awkward handoffs and exceptions that never appear in a process map. Their input can prevent an AI pilot from solving the wrong problem.
Training should include judgment, not just prompts.
People need to understand when the AI is useful, when it is unreliable, what information is restricted and how to verify important outputs. Managers also need enough knowledge to set realistic expectations and assess whether performance is actually improving.
Measure adoption as well as technical performance.
A model can be accurate while the workflow fails. Track whether people use the new process, whether they override or avoid it, whether cycle time changes, whether errors move elsewhere and whether employees feel the tool makes the work easier or more complicated.
Change management does not have to be heavy.
For a small pilot, it can be as simple as a clear sponsor, a few involved users, defined training, a feedback channel, weekly check-ins and a decision log. The important point is that someone actively manages the transition.
The organizations that treat AI as a people-and-workflow change will learn faster.
They will be better positioned to distinguish between a tool problem, a process problem, a data problem and an adoption problem—and to scale only what genuinely improves the work.
Next step
If your AI initiative affects multiple teams or workflows, combine the technology pilot with practical change and delivery support from the beginning.