The market for AI upskilling has never been noisier. Every week brings a new program, a new platform, a new promise of transformation. And to their credit, most organizations aren’t standing still: They’re investing, experimenting, and pushing their people to build new skills— even while the full picture of how work will evolve is still coming into focus.
That instinct to act now is right, because the urgency is real. AI investment is accelerating rapidly across industries. Yet more than 60% of organizations report little to no ROI, and nearly 80% of AI transformations fail to deliver expected impact. At the same time, leading companies are shifting their approach: They invest up to 2x more than laggards in upskilling their workforce and allocate as much as 60% of their AI budgets to capability building and retraining.
But the results are still uneven. Despite significant investment in capability building, only a minority of AI leaders are translating that investment into meaningful impact at scale. The issue isn’t effort, it’s effectiveness. In many cases, exposure and learning are happening. But performance is not following.
That gap comes at a cost. Not just in wasted spend, but in missed opportunity. As the pace of change accelerates, organizations that fail to translate capability into performance—in other words, turning know-how into “now do”—don’t just fall behind, they compound disadvantage.
From our work on the frontlines of large-scale workforce transformation, a clear pattern emerges. Capability only creates value when new skills are activated in the flow of work, not when they are taught in isolation. Organizations that get this right build what we call capability velocity: the speed at which skills are not just learned but applied and embedded to drive performance.
The question is no longer whether to invest in AI. It is how to distinguish the investments that will compound advantage from the ones that will evaporate.
Before greenlighting your next AI capability building initiative, consider the following five questions.
Are you building skills people will actually use, or just checking a training box?
Most capability programs still optimize for completion. People finish courses, pass assessments, and demonstrate understanding. But under real delivery pressure, behavior quietly reverts to old patterns or, increasingly, people experience cognitive overload (“brain fry”) as they try to integrate new tools into already demanding workflows. The risk in taking this approach is subtle but pervasive: Organizations mistake exposure for impact.
This matters more than ever because in AI-driven environments, value is created only when new tools, copilots, and AI agents are used effectively inside real day-to-day work. Foundational learning builds awareness, but it does not change how work gets done. Without applied practice—repeated, reinforced, and tied to actual outputs—capability never translates into performance.
The organizations seeing real results design skills training for use, not completion. They embed learning directly into the work itself, so that capability building happens in the moment of need and results in a behavioral shift. At a global hyperscaler, for example, software engineering teams did not learn AI tools in isolation. Instead, learning was integrated into live development workflows, with coaching and feedback tied to real code production. As a result, code productivity increased by approximately 30%, shipped features rose by 27%, and a higher share of code moved to production faster. People learned more about AI through the process. However, more importantly, they embedded new ways of working as a team.
Are you addressing an employee identity crisis in addition to a skills gap?
AI adoption rarely stalls because people don’t understand the tools. It stalls because people are uncertain what those tools mean for their role, their expertise, and their value. For many employees and managers, the age of AI feels like navigating an endless whitewater river—where roles, responsibilities, and ways of working are in constant flux.
When individuals have built careers on being the one who “does the work,” making the pivot to directing or augmenting that work with AI can feel like a loss, not a gain. For many employees, this is a real shift in professional identity, not just a change in tools.
This makes the capability-building challenge behavioral (and personal) as much as technical. AI is not simply adding new tasks. It is reshaping how value is created. Without addressing that shift, even well-designed programs produce a familiar outcome: people who know what to do and still don’t do it.
The most effective organizations treat this transition deliberately. They combine learning design with behavioral science—redefining roles, creating safe environments to experiment, and reinforcing new behaviors through managers and peer groups, with leaders acting as role models.
In a global hospitality company, frontline revenue managers had long made pricing trade-offs based on experience and judgment. A new AI-driven revenue management system was introduced to optimize and automate some of those decisions. Initial training created awareness but adoption lagged. Managers continued to override recommendations rather than trust the new system.
To accelerate employees’ full embrace of the system, the company had to address the identity issue directly. Behavioral interventions reframed the role: from decision maker to decision orchestrator. Managers were shown where their judgment still mattered and where the system added value and their continued interventions had the effect of destroying that value. They practiced using the system in real decisions, building confidence through experience rather than instruction. As trust grew, override behavior dropped and system-driven decisions increased, unlocking the value of the technology. The shift was not about learning a tool. It was about redefining what it meant to perform well and supporting teams with the shift in their workplace identities.
Are you actively enabling a change in habits, or just hoping for one?
Capability building fails when habits don’t change under real delivery pressure.
In AI-enabled environments, adopting new ways of working often comes with a temporary performance dip. Tasks take longer, workflows feel unfamiliar, and output may decline before it improves. If incentives and expectations are not adjusted, people revert quickly to what is efficient and what is measured—the way that worked for them before.
In one industry-leading intelligent data services company, teams were expected to adopt new AI-enabled ways of working. While the tools were available and training was completed, productivity initially declined as employees adapted. Because performance metrics remained unchanged, many avoided the new approach in practice. Instead they prioritized short-term output over long-term capability building.
The organization turned this around by realigning incentives and reinforcing new habits in the flow of work. KPIs were adjusted to reflect the transition, and progress became visible through peer comparisons and performance tracking. Managers played a central role by actively modeling the new ways of working and reinforcing expectations through ongoing performance conversations. As a result, adoption accelerated. Teams integrated new tools into daily workflows, performance recovered, and productivity improved as the new way of working took hold.
These signals matter. People follow what is measured, rewarded, and demonstrated. If you do not redesign incentives and actively reinforce new habits, resistance is inevitable. When you do, behavior shifts because the new way of working becomes the path to performance.
Are you using AI to teach AI?
Most organizations are still treating AI as something to be taught, rather than something that can transform how learning itself happens. The vast majority of AI skill building programs still rely on static modules, standardized pathways, and periodic training cycles. These approaches were already under pressure before AI accelerated the pace of change in how humans can learn, not just what they need to learn.
AI is the first technology capable of personalizing capability building at scale, adapting to individual readiness, providing real-time guidance, and reinforcing behavior directly in the flow of work. Ignoring these rapidly advancing features leaves learning in a slower, less effective mode that cannot keep up with ever-changing demands placed on the workforce.
Leading organizations are already embedding AI into capability building itself. Coaching agents provide contextual support during real tasks, answer questions in the moment, and reinforce new behaviors through continuous feedback loops. In a leading global consumer goods company, AI-enabled coaching is used to guide employees as they apply new tools in live environments. Similarly, in a large-scale health care transformation in North America, a GenAI-powered platform shifted learning from passive consumption to real-world mastery, accelerating impact delivery by up to 9x.
Increasingly, organizations are also using AI to rapidly and continuously diagnose emerging capability gaps, simulate future workflows, and pressure-test redesign choices before implementation — accelerating both organizational learning and transformation execution.
At scale, this requires organizations to not only learn faster but to teach faster, leveraging AI to free up capacity and elevating leaders to act as active developers of their teams in the flow of work. When AI becomes part of the learning system, capability building shifts from episodic training to continuous activation.
Are you building enduring human skills alongside AI skills?
There is a growing tendency to focus capability building on tool proficiency: how to prompt, how to automate, how to integrate AI into workflows. While necessary, this is only half the equation. As AI takes on more execution, the capabilities that differentiate human performance, judgment, communication, systems thinking, and adaptability become more critical, not less.
At the same time, organizations must guard against over-reliance on AI, as the “augmentation trap” can erode critical thinking and independent judgment if these are not actively strengthened.
Organizations must strive for balance. Programs that focus only on AI usage risk commoditizing the workforce. Programs that ignore it risk irrelevance. Neither creates sustained advantage.
What distinguishes leading organizations is that they treat these as two distinct but interdependent capabilities. They invest in helping employees direct AI effectively by framing problems, interpreting outputs, and knowing when to override, while also deliberately building the human capabilities that AI cannot replicate wholesale.
This is becoming the next frontier of capability building. As technical skills become more accessible and, over time, commoditized, human capabilities—such as problem solving, judgment, collaboration, and relationship building—become increasingly valuable, yet remain harder to teach and measure. Traditional signals like credentials or training completion are often weak predictors of performance in AI-augmented environments.
Leading organizations are therefore embedding these capabilities into real and simulated environments, using AI-enabled coaching, role play, and interactive learning experiences to develop and test judgment in context rather than knowledge in isolation.
At a global technology company, for example, internal programs emphasize decision making, collaboration, and creative problem solving in AI-augmented environments alongside technical fluency. The goal was not just higher productivity, but better decisions and more resilient performance.
As AI capabilities continue to scale, the advantage will not come from access to technology alone, but from how effectively organizations develop the human capabilities that shape how that technology is used and thereby push the human performance frontier to previously unattainable levels.
From Capability Building to Performance Infrastructure
The organizations pulling ahead right now are not the ones spending the most on AI training. They are the ones that have recognized a more fundamental shift: Capability building is no longer a support function; it is performance infrastructure.
They move beyond fostering employee awareness of new tools to ensuring that capability is applied in real work and embedded through systems, incentives, and reinforcement. They measure success not by participation, but by changes in output and outcomes. And they treat the speed at which skills are built, applied, and embedded—known as capability velocity—as a strategic asset.
In a world where change is continuous, these new ways of thinking and acting matter. Organizations no longer fall behind because they fail to invest in building those capabilities. They fall behind because they fail to translate their capability investments into performance.
The goal is not to teach people about AI. It is to enable them to perform with it consistently, at scale, and under real conditions. That is a harder problem. It is also the only one worth solving.
