When Christine Karel joined a company-wide rollout of Microsoft Copilot at Ameriprise Financial, the learning and development team wasn’t in the room. The AI tool launched across the organization with technology, risk, and compliance leaders at the table, but without anyone to think through how employees would actually understand and use it. The predictable result was confusion, hesitation, and a workforce that treated a sophisticated AI assistant the way it might treat a search engine.

“If you launch a massive investment like Copilot across an organization,” said Karel, head of enterprise learning at Ameriprise Financial, “how are we, as talent and learning and HR, supporting that initiative?”

That question, and the hard lessons behind it, set the tone for a lively panel discussion at a From Day One’s Minneapolis conference. A panel of leaders dug into how companies are scrambling to build AI fluency among employees, why the pace of change keeps outrunning their strategies, and what it will take for L&D to earn a permanent seat at the table. The panel was moderated by Evan Ramstad, business columnist for the Minnesota Star Tribune.

When the Tool Launches Before the Training

The story Karel told about Copilot wasn’t unique. Michelle Anderson, VP of global learning & development at AmTrust Financial Services, described almost the same experience. 

“A couple years ago, we launched Copilot to the entire company without us, they didn’t come to L&D at all, and then they wondered, ‘Well, why did this fall flat? Why isn’t anybody using it?’” The company pulled the tool back and is now building its own internal solution. L&D has a seat at that table now, Anderson says, but only because the first attempt failed.

The pattern reveals something important: AI adoption isn’t primarily a technology problem. It’s a learning problem. People arrive with wildly different baselines: some treat AI like Google, others are genuinely uncertain what it can do. And without a deliberate strategy to meet them where they are, even well-funded rollouts stall.

Anderson’s team at AmTrust has responded with a framework called Grow, designed to weave learning into employees’ daily workflow rather than tacking it on as a separate task. The system uses job descriptions and self-reported skills to recommend relevant development, surfacing nudges through Microsoft Viva Learning directly inside Teams. Managers are expected to reinforce the habit (a minimum of one hour of learning per month is company policy) because, as Anderson put it, “if your manager doesn’t support something, you’re not going to find the time to do it.”

Measuring Comfort, Not Just Completion

One of the sharper debates among panelists was how, and whether, to assess employees’ comfort with AI tools. Anderson said AmTrust deliberately chose not to survey its workforce on the topic. “We were worried that there would be fear associated with it,” she said. “And I don’t think we’re ready to address that fear yet.”

Carita Hibben, VP of HR at C.H. Robinson, took a different approach. About two years into what she described as a comprehensive AI transformation at the logistics company, her team conducted a pulse survey asking employees to rate their comfort with AI in their daily work. Around 73% said they were comfortable—a result Hibben called encouraging, and one that also yielded “actionable insights on areas that we might need to dive deeper into.”

C.H. Robinson has also leaned into AI-powered role-play simulations. Using a platform, employees and managers can practice challenging conversations by either playing themselves or switching roles to see the scenario from the other side. The same kind of simulation is used by the company’s sales and account management teams to prepare for difficult customer conversations. Hibben noted that utilization spikes during performance review periods and other high-stakes moments, even without formal requirements to use the tool.

AmTrust has built out similar functionality through LinkedIn Learning, with the added ability to create custom simulations tailored to specific roles. The platform scores participants and recommends follow-on coursework based on performance. Anderson’s team is now exploring how to embed those simulations into new-hire programs for claims associates.

Closing the Loop at the C-Suite Level

Moses Berkowitz, chief revenue officer at Censia AI, works closely with CHROs, CFOs, and CEOs on workforce strategy. He offered a bird’s-eye view of what’s driving urgency at the top of organizations. “This is a CEO and board-level conversation right now,” he said, “and the conversation in 10 out of 10 rooms is that this is going to have a massive impact on our workforce. Not in a scary way, but jobs are really going to change.”

What gives him optimism, Berkowitz said, is that conversations previously happening in silos are now happening in the same room: What skills does the business actually need? What do our employees have? Where’s the gap? That alignment, he says, is the foundation for meaningful progress.

Karel echoed that framing, describing how Ameriprise’s AI Leadership Council, originally composed of technology, risk, and compliance leaders, has expanded to include talent, communications, and business unit representatives. The lesson she drew was pointed: you can only get so far with a tool. “The people that run the tool, that think through the tool, and actually work around the tool are really what we need to be thinking about.”

Berkowitz added a striking data point from his firm’s work with one of the world’s largest consulting companies: out of 400,000 employees, the single heaviest user of their internal AI tools is the CEO. “That sends a message to everyone that we take this seriously.”

The Half-Life Problem

Skills are expiring faster than ever, and AI is accelerating that trend at a remarkable pace. Berkowitz says that the half-life of a skill has collapsed to under five years, and for anything AI-adjacent, it’s compressing even faster.

The implications for L&D are significant. Berkowitz described working with a university that invested $250,000 in an AI training program. Six months after launch, the program was obsolete—the underlying technology had moved on. “As the half-life of skills compresses, we need to think about how we build programs differently,” he said.

The panelists largely agreed that the answer isn’t to build more programs faster, but to build differently. That formal, comprehensive training curriculum may simply become too time-intensive to justify, says Anderson. “We’re going to have to get to the point where we’re not building big giant formal programs anymore. We’re building more in the flow of work, in the place that they need it.”

Karel put it plainly: AI makes it faster and easier to create content, but that won’t solve the structural problem if organizations are building around the wrong model entirely. What will endure, Karel says, are the foundational capabilities: critical thinking, adaptability, ethical judgment. “Those are old skills. If you could just base the foundation on some of those things, those are going to be the things that take you along the way.”

The Human Element Doesn’t Disappear

As the conversation turned to productivity, Berkowitz gently pushed back on the framing that tends to dominate headlines. “The topic of productivity, it’s all we want to talk about in the media, and we’re replacing workers, but I think it’s a bit of a red herring.” The more useful question, he says, is how organizations can deliver more value to customers per unit of human capital. He cited the example of bank tellers after ATMs arrived: The work changed, but the role didn’t disappear. Tellers shifted from handling cash to greeting customers, and the experience actually improved.

Anderson made a similar case, pointing to research suggesting that some roles could see 40% to 50% productivity gains through automation. But she was quick to add the counterweight: “There has to be a human at the center of it. How do we teach people to be better critical thinkers and thought partners?” She envisions a future where managers become more important, not less, not because AI will replace their authority, but because coaching, psychological safety, and human judgment will matter more as digital tools handle more routine tasks.

What L&D Looks Like in Five Years

Closing the session with a rapid-fire look ahead, each panelist offered a vision for where learning and development is headed.

Karel predicted that the function will shift its focus from teaching specific topics to shaping workforce design, with learning as one lever among many, grounded in deep knowledge of what skills the business actually needs.

Hibben emphasized agility above all else. “What we know is that there’s going to be continuation of skills needed, and those talent practitioners need to be flexible and agile.”

Anderson was less focused on where L&D sits in the org chart than on whether it maintains strong partnerships with the business and stays aligned on outcomes.

Berkowitz offered a challenging take: the United States spends roughly $150 billion annually on upskilling and reskilling, he says, and he believes that figure still dramatically underestimates what’s needed. “If I were to wave the magic wand, we’re going to invest a lot more in L&D.” The catch is that it will be managed like a business. “We’re going to run it like a P&L.”

Anderson’s response was immediate: “I would be okay with that.”

Grace Turney is a St. Louis-based writer, artist, and former librarian. See more of her work at graceturney17.wixsite.com/mysite.

(Photos by Josh Larson for From Day One)