Can Organisations Learn Fast Enough for AI?
Topic: Technology, Human Resources
Format: Article
Published Date: September 2026
Chief Human Resources Officer Programme
AI is exposing that many organisations are structurally slow at adapting to a world where skills and ways of working change faster than traditional learning systems can keep up.
Until recently, the race around enterprise AI focused on acquiring early access. Companies scrambled to hire machine learning engineers, secure cloud capacity, and build relationships with model providers because experimenting at scale remained expensive and concentrated among a relatively small group of players. Much of the early momentum came from the belief that organisations that are able to secure technical capability early would be better positioned to build products around it later.
That assumption no longer holds much water. AI tools are now embedded into the software employees already use every day, but this ease of access has not automatically translated into usage. McKinsey’s 2025 Superagency in the Workplace report found that while 92 per cent of companies planned to increase AI investments, only one per cent considered themselves mature in deployment, with AI integrated into daily work and producing measurable outcomes. The World Economic Forum’s Future of Jobs Report 2025 found that 63 per cent of employers viewed skills gaps as the biggest barrier to transformation.
Deploying AI tools is easier than changing how people work at a structural level. The harder, more complex task is helping employees learn new ways of working quickly enough, for those investments to matter.
Mistaking familiarity with AI for expertise
AI systems produce fluent, confident-sounding output, whether they are right or wrong. The less someone understands the underlying problem, the less likely they are to recognise when an answer is plausible but wrong.
Vishal Karungulam, Clinical Assistant Professor (Teaching) of Information Systems at ISB, calls this the “illusion of competence.” Confidence produced by repeated AI use can mask real gaps in judgement because nothing in the experience signals that something has gone wrong.
Research by Erik Brynjolfsson and colleagues found that a GPT-based assistant increased productivity among customer-support agents by roughly 15 per cent overall, with the largest gains among less experienced workers. The people seeing the biggest productivity gains were not necessarily the people best able to spot mistakes.
The same concern is emerging in professions built on apprenticeship. For decades, junior lawyers built judgement through the laborious work of drafting documents, reviewing cases, and conducting research. A LexisNexis Mentorship Gap report surveying nearly 900 UK lawyers found that AI tools improved productivity, particularly among junior lawyers. At the same time, 72 per cent identified deep legal reasoning as the biggest emerging skills gap among junior professionals. Performance gains seem to be arriving faster than the experience needed to evaluate them.
Duolingo’s experience in 2025 shows the risks of replacing human-generated content with AI-generated content faster than quality-control systems can adapt. In April, CEO Luis von Ahn published a memo announcing the company was going “AI-first” and would accept small hits on quality as it phased out contractors for work AI could handle. The memo was framed entirely around internal efficiency, with nothing about improving the learning experience. Users treated the quality changes as a breach of trust. After losing more than 400,000 followers across social media, Duolingo acknowledged the controversy had dampened customer growth.
Most corporate learning systems were built for a slower world
AI is exposing weaknesses across organisational systems, from governance and hiring to compliance and workforce planning. Among them is a problem that many companies have spent decades taking for granted: how people learn new skills at work.
PwC’s 2025 Global AI Jobs Barometer found that industries more heavily exposed to AI generated three times higher revenue per employee than less exposed sectors. The difference increasingly reflects how quickly people inside organisations can learn new skills and put them to use.
Amazon’s Upskilling 2025 initiative committed $1.2 billion to retraining more than 300,000 employees across cloud computing, machine learning, and AI, with more than 700,000 employees participating in free training programmes since 2019. Yet the underlying structure remains much the same: formal programmes delivered at intervals.
What AI is teaching organisations about learning
Mastercard offers a different approach. Its Unlocked platform allows employees to find projects, mentorships, and learning opportunities across the organisation. When its fraud-detection division needed AI capability quickly, the platform redeployed employees with adjacent data skills. By 2024, 90 per cent of employees were registered, with 500,000 project hours logged, and a third of participants had seen a role change or promotion.
The organisations adapting most successfully are making three shifts:
Moving learning closer to business decisions: In many organisations, training still follows transformation. Leaders decide what will change, and learning programmes arrive later. As AI reshapes work more quickly, that sequence is becoming harder to sustain.
Ensuring AI adoption moves beyond the tech team: The value of these tools increasingly depends on whether people across the organisation can use them with sound judgement. Organisations where employees use AI superficially are finding it difficult to translate adoption into meaningful results.
Focusing on learning even as the job itself transforms: Formal programmes still matter, but they struggle to keep pace with technology that evolves faster than annual planning cycles. Employees increasingly need to learn while systems, tools, and workflows are still evolving.
The companies that benefit most from AI are unlikely to be those spending the most on tools or training. They will be the ones who help employees build skills and judgement while using the technology. AI can shorten the path to proficiency and allow people to take on new responsibilities faster. But organisations still need ways for employees to develop the foundational knowledge that good judgement depends on. Traditional training models were designed for a world where skills changed more slowly. AI is compressing that timeline. The challenge for leaders is not simply helping people use the technology, but ensuring they continue to learn from it.