Russell Sarder, CEO & Founder of AI CERTs – advancing global AI certification & education.gettyMost companies no longer have an AI access problem. They have a conversion problem. Companies have co-pilots, assistants and pilots in motion. What many still lack is a reliable way to turn those tools into faster decisions, better output, lower costs or stronger revenue performance.McKinsey’s 2025 global survey makes that gap hard to ignore. While AI use is broadening, more than 80% of respondents say their organizations are still not seeing a tangible enterprise-level EBIT impact from generative AI. McKinsey also found that workflow redesign has the biggest effect on whether companies see EBIT impact, yet only 21% of respondents said their organizations had fundamentally redesigned at least some workflows. That is why this conversation needs to move beyond training volume. The real question is not how many employees completed an AI course, but whether the business has built enough role-specific capability to change how work gets done. Adaptive learning is a business system, not a learning feature.This is where adaptive learning becomes strategically important. In many boardrooms, it is still framed as a learning innovation. That is too narrow. Adaptive learning can be better understood as a capability-allocation system. It matches learning to role, readiness, workflow and business priority, then adjusts as those conditions change. In an AI environment where the shelf life of skills is shortening, that matters. PwC found that workers with AI skills command a 62% wage premium, and that productivity growth in AI-exposed industries has accelerated sharply. That is a clear signal that AI capability is not a soft talent issue, but an economic asset. Traditional upskilling models were not built for this moment. Annual curricula, broad enterprise rollouts and completion-heavy dashboards may create awareness, but they rarely create operating leverage. This is because AI changes too quickly. Roles do not need the same depth of capability. Generic learning paths usually sit too far from the workflow to influence real performance. BCG’s 2025 AI at Work research reinforces this point. Only 36% of employees said they felt properly trained in AI use, and regular usage was markedly higher among employees who received at least five hours of training, especially when that training included in-person support and coaching. Blanket upskilling is often a low-return use of capital. Start with workflows rather than course catalogs.A better starting point is the workflow, not the course catalog. Leaders should ask which workflows carry enough economic weight to justify targeted capability investment first. In most organizations, the answer tends to fall into three buckets:• Revenue-critical workflows• Cost-heavy workflows• Risk-sensitive workflowsOnce that priority is clear, adaptive learning can do what static learning cannot. It can tailor the depth of instruction to a specific job, embed practice inside the flow of work and evolve the learning path as the tool set and task change.ROI begins when behavior changes inside the workflow.That is also why adaptive learning should not be judged by participation. Completion rates are not ROI, logins are not ROI and even usage alone is not ROI. ROI begins when learning changes behavior inside a meaningful workflow, and that change shows up in business metrics leaders already trust. I've found that those metrics usually sit in four places:1. Productivity: Cycle time, throughput and time saved on repetitive work2. Quality: Better decision support, fewer errors, stronger customer interactions and more consistent execution3. Labor Leverage: Getting more value from existing teams, reducing unnecessary external hiring and redeploying talent into higher-value work4. Strategic Speed: How quickly the company can move from experimentation to scaled execution McKinsey notes that fewer than one in five organizations track well-defined KPIs for generative AI solutions, which helps explain why so many AI programs still feel promising but hard to prove. Visa offers a useful example of what this looks like when it is done well. In LinkedIn Learning’s 2025 Workplace Learning Report, Visa describes embedding AI-powered training and coaching into its broader product knowledge and solutions learning program. The point was not to create a more modern course, but to give sales teams a safe environment to practice pitches, receive automated feedback and build fluency across a rapidly expanding product set. This approach led to a 78% increase in seller confidence, while 83% of leaders saw value in sellers using the program and the tool to practice their pitches. That is a strong case because it ties adaptive learning to a real commercial moment, in a real role, with an outcome leaders can understand. ​New York Life shows the enterprise version of the same idea. The company unified its talent systems to upskill 12,000 employees in AI. What stands out is not simply the scale of the effort. It is the operating discipline behind it. The company moved away from expensive long-running programs that did not have clear ROI, refocused on companywide priorities such as leadership and AI skills, and used that architecture to support internal mobility and skills agility. New York Life now fills 33% of roles internally, and the company has logged more than 15,000 hours of AI learning. That is what adaptive learning looks like when it becomes part of talent infrastructure rather than a stand-alone initiative. This is an operating model decision.There is a larger leadership point here. In the AI era, learning is no longer just a development function. It is a speed-to-value function. Companies that treat AI upskilling as a benefits program will likely struggle to prove returns. Companies that treat it as execution infrastructure will have a better chance of turning AI investment into measurable performance. The winners will not be the organizations that produce the most content or enroll the most employees. They will be the ones that build the fastest loop between new technology, role-specific capability, redesigned workflows and business metrics that matter.​Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?