Organizations are entering a new phase of their AI journey. The initial wave of exploration is largely behind them, and employees are increasingly incorporating AI into their daily work to boost productivity and streamline routine tasks. By many measures, that is a success. But for business leaders, it is also just the beginning.
The question is no longer whether employees can use AI. It is whether organizations are using it in the ways that matter most. Beyond individual productivity gains lies a larger opportunity: rethinking jobs, processes, decision-making, and the way value is created across the organization.
That challenge was the focus of Impact Live: AI at Work: Don’t just use it – Lead it, moderated by Fredrik Odegaard, Associate Professor of Management Science at the Ivey Business School. He was joined by Ivey AI Fellows, Michael Pelosi, Country Manager for Canada at Cohere, and Melissa Hartwick, Principal at Boston Consulting Group (BCG). Their conversation spanned everything from productivity and job redesign to governance, scaling, and skills, but returned repeatedly to a central idea: successful AI adoption is as much a leadership challenge as a technological one. Three themes, in particular, stood out.
Start with business value, not the technology
One of the clearest messages from the session was that organizations need to move past treating AI as a set of interesting tools and instead tie it to a specific business outcome.
Pelosi described AI adoption as a progression from augmentation to automation and then innovation. Many organizations, he said, have already become comfortable using AI to support routine tasks such as summarizing documents or producing drafts. But he argued that the larger opportunity lies beyond that.
“The real opportunity we see is not just augmenting employees and or automating certain tasks or processes, but how are you going to use the technology to innovate and build something net new.”
He illustrated the point with the example of the now defunct video rental store, Blockbuster. AI might have helped the company recommend videos more effectively or automate the rental experience, he said, but that would not have changed the core business. “The fact is, they’d still be a video store,” he said.
Hartwick made a similar point using BCG’s framework of deploy, reshape, and invent. Off-the-shelf tools can bring quick gains, especially at the individual level, but she said those gains do not automatically translate into enterprise value. With basic deployment, “you might see kind of a 10 to 20 per cent productivity improvement,” she said, which can be useful, “but you’re not fully translating it to those enterprise value creation type opportunities yet.”
For both speakers, the practical implication was the same: leaders need to define what success looks like before launching pilots or adopting new features. Hartwick, in particular, urged organizations to be explicit about their objectives, whether that is driving revenue growth, reducing costs, mitigating risk, or improving customer experience.
Most work will be reshaped before it is replaced
The panel also challenged the idea that AI adoption is best understood through job loss.
Drawing on recent BCG research, Hartwick said the broader trend is job redesign rather than wholesale replacement. “A job is fundamentally a set of multiple tasks,” she said. If some of those tasks are automated, leaders still need to decide what happens next: whether employees do more of the same work, shift toward a different mix of responsibilities, or develop new capabilities.
That distinction matters because automation at the task level does not necessarily eliminate a full role. Hartwick said the impact will depend partly on whether the work is being augmented or substituted, and partly on whether demand expands as costs fall. In some areas, especially where demand is relatively fixed, substitution may be more likely.
"We are starting to see … the true substitution, where the majority of tasks can be substituted and there’s a fixed demand for that particular job," she said.
Pelosi agreed that the more useful question for leaders is how work changes. He said many executives have been surprised by how quickly younger employees are adapting to AI tools and using them naturally. Rather than seeing recent graduates as especially vulnerable, he said employers are often impressed by how readily they work with AI.
He also pointed to sectors where automation is being driven by labour shortages rather than headcount reduction. In manufacturing, logistics, and related fields, some employers are struggling to fill roles at all. That makes AI part of a broader workforce response, not simply a cost-cutting tool.
Most importantly, Pelosi said the leaders he sees making progress are not starting with the idea of removing people. “All of them are talking about how can they enhance what the employee is doing and get more output, and make them focus on more valuable tasks.”
That perspective carries important implications for how organizations think about work. If roles themselves are being reshaped, then reskilling cannot be treated as a separate initiative; it has to be built into how roles are redesigned and performance redefined.
The hard part is scaling responsibly
A third takeaway from the session was that the biggest challenges often begin after the pilot.
Hartwick said many companies are still in an experimentation phase, running multiple use cases without translating them into sustained business value. She argued that governance needs to be built in from the beginning, not treated as a side issue. Without clear guardrails, organizations can end up with fragmented accountability and inconsistent standards.
She also emphasized that most of the work of adoption sits outside the model itself.
“10 per cent should be on the algorithms, 20 per cent should be on the data and tech, and 70 per cent should be on the people and processes,” she said. That means leaders need to think carefully about operating models, ownership, training, and how employees will actually use AI in day-to-day work.
Pelosi made a similar point from the perspective of deployment. A proof of concept may be promising, but rolling out AI at scale raises a different set of questions around compliance, security, legal risk, and reliability. He noted that there is a major difference between generating something impressive in a demo and putting a tool in front of customers or employees in a way that an organization is prepared to stand behind.
For that reason, both speakers cautioned against mistaking enthusiasm for readiness. At the same time, neither argued for waiting until everything is just right.
“Do not let the pursuit of perfection prevent progress," said Pelosi.
Hartwick closed with a similar point, saying that some organizations are beginning to see stronger results because they have become more disciplined. Instead of launching many disconnected efforts, they decide that “we’re going to pick one to two functions for the next year” and focus there.
The AI adoption opportunity is real, but so is the organizational work required to make it useful. For leaders, the challenge is not only to introduce AI into the workplace, but to decide where it matters, how work should change around it, and what it takes to scale that change responsibly.
Watch the full Impact Live webinar, AI at Work: Don't just use it - Lead it, on Ivey Impact or Ivey's YouTube channel
