For the past three years, most organisations have approached AI adoption as a technology rollout. They bought licenses, launched copilots, trained employees on prompts and built entire strategies around implementation.
The assumption was that if people had access to the right tools, transformation would follow. It did not.
Across the AI learning and enterprise transformation space, one statistic keeps resurfacing. According to McKinsey’s State of AI report, around 88% of companies are now using AI in some form, yet only 6% are capturing meaningful business value from it. While the figures vary across studies, the broader pattern is impossible to ignore. AI adoption is widespread, but AI impact is not.
That gap is where the real issue sits. Many organisations have invested in AI, moved quickly and encouraged experimentation. Yet the return has often been limited because the harder challenge was misdiagnosed from the start.
AI adoption is not simply a question of access to technology. It is a question of whether people can change the way they work quickly enough to make the technology useful. That makes AI adoption a workforce adaptability challenge.
The organisations that capture value from AI will be those that can help people build judgement around new systems, redesign workflows and develop new capabilities at the speed the technology now demands.
Why most AI rollouts stall
Software can be deployed quickly, but behaviour changes more slowly.
I see this divide clearly in conversations with learning businesses and enterprise teams. The organisations making progress ask harder questions than which tool to buy or which prompt framework to teach.
They are asking how capability moves through the company.
How do people learn new ways of working while the work itself is changing? How do teams build judgement around AI outputs? Exactly when should employees trust automation, challenge it or apply their own expertise?
These questions determine whether AI remains a collection of experiments or becomes part of how the organisation actually creates value.
This is why every fast-moving organisation is becoming a learning business, whether it describes itself that way or not. A company that cannot learn quickly will struggle to adapt quickly. In an AI economy, those two things are becoming inseparable.
The two paths organisations are taking
As learning strategist Lavinia Mehedintu recently observed, organisations increasingly appear to be splitting into two broad paths. Some are treating AI as a systemic change process. Others are treating it as a technology training project.
The training project approach is familiar. Employees are shown how to use new platforms, generate outputs, automate workflows and experiment with AI tools. That work has value, especially in the early stages. It builds confidence and reduces some of the fear around new systems.
It also has limits.
Tool training can create useful pockets of efficiency, but it rarely changes how the organisation operates. People may become more comfortable with AI without becoming much better equipped to rethink the work around it.
The systemic-change approach goes deeper. In these organisations, AI now prompts questions about roles, processes, decision-making and value creation. Leaders are asking where human judgement matters most, which workflows need to be redesigned and what new capabilities people need to develop.
This distinction matters because the adoption vs impact gap is not caused by a lack of AI activity. Many organisations are extremely active in their use of AI. The problem is that this activity can be mistaken for progress.
Tools are being implemented across departments while the underlying operating model remains largely unchanged. That is where the gap between adoption and impact begins to open.
Why traditional learning models are breaking
The traditional approach to workplace learning assumed relative stability. Skills remained useful long enough to justify structured programmes, fixed content and periodic retraining.
AI breaks that assumption. Workflows now evolve within months. Roles shift before organisations can define them. Knowledge becomes outdated almost as quickly as it is documented.
As Mehedintu has noted, AI is now “coming at L&D from so many different angles” that many teams feel paralysed by the speed and complexity of change. They are being asked to support AI adoption across the business, rethink their own use of AI and respond to the shrinking shelf life of skills.
That is a lot to ask of a course-led model. A quarterly training programme was never designed to keep pace with technology that can change how people work from one week to the next.
For years, organisations treated learning as something layered on top of work. Employees would leave their workflow, consume content, complete modules and then try to apply that knowledge later in real situations.
That model was already under pressure. AI is exposing its limits.
Learning must move closer to work
AI is beginning to make something else possible.
Learning strategist Enrique Rubio recently described “learning in the flow of work” as “the dream of dreams” for workplace learning. The idea has existed for decades, but what is different now is that AI makes parts of it operationally possible in a way that was previously unrealistic.
An employee drafting a proposal can receive guidance on structure, tone or compliance requirements in real time while writing. A junior analyst can receive contextual explanations while working through unfamiliar data. A support representative can receive coaching during a difficult customer interaction rather than weeks later in a generic training session.
This is where AI can change learning most profoundly. It can make capability-building more immediate, contextual and continuous.
Adaptability is not built through content consumption alone. People become adaptable by practising new behaviours, receiving feedback, applying judgement and improving while doing the work.
When the work itself is changing, learning cannot remain separate from the work.
What AI is actually changing
The most useful AI learning applications help organisations create, adapt, distribute and apply knowledge at scale.
Until recently, that was difficult. Building multilingual learning programmes, personalised support, knowledge tools and role-specific guidance required significant time, budget and operational capacity.
Now smaller teams can do far more.
One example comes from the World Bank Group, where a relatively small learning team used AI-supported systems to reach more than 60,000 users across 195 countries through multilingual learning programmes and knowledge tools.
Similarly, the team behind global surgery training academy SURGhub used AI to build a custom reporting tool that replaced a full week of manual data work per quarter, with no dedicated engineering resource.
What is striking about these examples is not simply efficiency. It is that smaller teams can now build learning systems with a reach and flexibility that would previously have required far larger resources .
A small team can reach global audiences. Complex knowledge can be translated into accessible formats.
Courses, resources and support tools can be updated more quickly. Learning experiences can be localised, reused and improved without rebuilding everything from scratch.
This is what learning as infrastructure means in practical terms. It gives an organisation a repeatable system for moving knowledge and capability to the people who need it, when they need it, in a form they can actually use.
The real opportunity is to build learning systems that help people adapt continuously, rather than simply automating old training models.
The rise of always-on learning systems
We are already seeing signs of this across the market. Over the past two years, the market has seen a predictable explosion in AI courses, certifications and tool-specific training.
Some of that work is useful. People need basic literacy and confidence with new tools. The risk is assuming that tool familiarity solves the deeper capability problem.
Knowing how to use an AI tool does not automatically mean someone knows how to work, decide, communicate or lead in an AI-enabled organisation.
At the same time, a second market is emerging. Companies are investing in B2B learning systems designed not only for employees, but also for customers, partners and end users who need to understand new AI-driven workflows.
As AI becomes embedded into products, services and operations, companies will need to educate entire ecosystems. Customers will need guidance. Partners will need enablement. End users will need support as workflows change.
AI therefore increases both the number of people who need to learn and the frequency with which learning has to happen.
Across the market, demand for customer education systems, partner enablement platforms and scalable B2B learning operations is increasing as organisations adapt to AI-enabled workflows. AI adoption is creating growing demand for learning approaches that help people understand, use and adapt to changing ways of working.
Learning technologists will become increasingly important in this environment. Their role will be to help design the systems through which organisations keep employees, customers and partners moving with the technology.
Why adaptability becomes the advantage
I increasingly believe that as AI systems become more capable, human value shifts upward. That AI will ultimately elevate the importance of learning rather than diminish it.
Competitive advantage increasingly comes from judgement, synthesis, context and decision-making under ambiguity. These capabilities do not appear automatically once a company buys software.
They have to be developed.
This is why adaptability will become one of the defining advantages of the AI era. The organisations that thrive will probably not be those with access to the most AI tools. Those tools will quickly become widely available.
The advantage will belong to organisations that can absorb change faster than competitors. Organisations that can continuously reskill, redeploy and support their workforce while technology keeps moving underneath them.
Ultimately, AI transformation is less about teaching people how to use new tools and more about building organisations capable of continuous reinvention.
That’s not a software problem.
It’s an adaptability problem.
