'A lot of employers are a bit behind on taking action, creating policies or taking a stance on AI use in their organization'
Organizations rushing to adopt AI without a clear use case or training strategy are setting themselves up for failure.
Why? According to Aniklet Zefi, instructional designer at Concordia University and former HR professional, the real problem isn’t the technology, it’s the absence of planning.
“There are different profiles of employees who have different needs. Leaders are one of those, because leaders aren't just responsible for using the tools, they're responsible for creating policies around these tools and creating a strategy. If they don't know how AI tools work, you're going to see adoption lacking,” he says.
In fact, lack of structured rollout has consequences. A recent report from Boston Consulting Group found that only 36% of employees feel adequately trained for AI use. Thirty-seven percent also said their company isn’t supplying the right tools, and only a quarter said they have sufficient support from leadership on how and when to use AI at work.
When employees lack access to the AI tools they need, over half say they'll turn to unofficial alternatives. That opens the door to confusion, security vulnerabilities, and a fractured approach to implementation, the report said.
This is what Zefi calls “shadow AI” – when employees adopt AI tools without formal guidance or oversight.
“Employees are using generative AI tools, but employers haven’t given any guidance on best use, responsible use policies or how to use it in the workplace, which is concerning,” he explains.
Leadership hesitation is stalling AI adoption
For Zefi, one of the core problems is that many leaders aren’t equipped to guide their teams through AI integration. This creates a bottleneck, because “you really need to start at the top, you need buy-in from the leaders,” he says. It’s not enough to simply roll out a platform or provide a license — training is essential.
“You have a lot of leaders who don't have the training themselves, don't understand the technology themselves. So oftentimes, when we fear something, we choose to ignore it,” he says. “A lot of employers expect that them paying for something means that employees will use it. So how do you bridge the gap between you investing in a tool and employees using it?”
This question highlights a core issue: many organizations are confusing deployment with adoption. Solving this starts with a needs-based approach, with training tailored to different departments and employee skillsets, he says.
“I really believe in conducting a training needs analysis or assessment for your organization,” Zefi says. “It's about identifying the different profiles you have within your organization and then seeing what kind of generative AI skills they need.”
Organizations that take the time to define goals and constraints will see more value in their programs, which is especially important when developing policies. Too many companies, he says, haven’t even decided on their stance.
“What are your values as an organization? Based on those values, what do you want to use Gen AI for?” Zefi says. “Sometimes even one sentence is important: ‘We allow employees to use generative AI in these scenarios.'”
But Zefi also warns against paralysis, urging leaders to avoid inaction just because the technology is evolving so quickly.
Recommendations for employers
When it comes to implementing AI within an organization, the responsibility isn’t limited to offering learning materials. According to Zefi, communication, safety and employee well-being need to be at the center of implementation strategies. That includes emotional safety, as the fears around AI displacing jobs are real and prevalent.
“It’s not about controlling employees. It's about giving them a sandbox that's safe,” he says.
Adoption doesn’t have to start with high-stakes automation; Zefi recommends identifying projects or departments where GenAI can deliver quick wins without major regulatory or ethical complexity. This approach helps build momentum and internal success stories, reduce resistance among hesitant teams and encourage feedback loops to improve future rollouts.
“There’s adoption that you might not want; you might be having adoption that is not sanctioned or that breaches your confidentiality or your security. You want the right type of adoption,” he says.
