Skills are becoming a decisive factor in AI adoption
The OECD's June 2026 policy brief AI and skills: What we know so far places an important issue at the centre of the artificial-intelligence debate. Much public discussion focuses on model capability, computing infrastructure and investment, yet organisations cannot realise those benefits automatically. The skills of the people using AI are increasingly one of the factors that determine whether adoption produces productivity, better decisions and sustainable change.
The OECD highlights that a lack of relevant skills is already holding back adoption. Its evidence indicates that businesses, including many SMEs, identify skills shortages as a significant barrier to effective use of AI. This should change the way employers think about AI strategy. Technology expenditure is only part of the investment; the other part is human capital.
The capabilities required for effective use
The first requirement is AI literacy. Employees who use AI systems do not all need technical knowledge of machine learning, but they do need an informed understanding of how the systems behave. They should recognise that outputs may be incomplete, biased or inaccurate, understand basic confidentiality risks and know how to verify important claims. This level of literacy allows users to benefit from AI without confusing fluency with reliability.
Problem framing is a second capability. Generative AI is much more useful when the user can define the objective, audience, constraints and source material clearly. Inexperienced users often ask vague questions and judge the technology by the vague answers that follow. Skilled users understand that effective interaction begins with a well-structured problem. This is less a matter of clever prompting than of disciplined professional thinking.
Evaluation is a third requirement. As AI output becomes more persuasive, the ability to judge relevance, evidence, logic and context becomes increasingly valuable. In many professional roles, producing a first draft will become easier while deciding whether the draft is good enough to use becomes more important. Subject expertise therefore remains valuable. A trade finance professional can detect a subtle error in a documentary-credit analysis that a novice may accept; a marketer can distinguish a superficially attractive campaign from one that fits the customer journey; a trainer can assess whether an AI-generated question measures the intended learning objective.
AI changes tasks as well as occupations
The OECD's wider work on skills and the future of work supports the view that AI changes the task mix inside occupations rather than producing a uniform elimination of entire jobs. This has significant implications for training. Organisations should not focus only on teaching employees how to use a new tool. They should help people understand how their own role is changing and which parts of the work should remain human.
Workflow design becomes an important capability in this context. Businesses gain more value when AI is integrated into repeatable processes rather than used for occasional experimentation. Staff need to understand how information moves through a task, where AI can remove friction, what requires checking and how the output connects to the next stage. These skills are particularly important as systems move from answering questions toward taking actions.
A practical workforce-development model
For employers, a useful training model can be organised into three layers. The first layer is foundational literacy for everyone who uses AI. This should cover basic concepts, confidentiality, verification, common failure modes and responsible use. The second layer is role-specific application. Sales teams should practise sales tasks, finance staff should practise finance tasks, and export teams should practise market research and documentation support relevant to international trade. Generic demonstrations are useful initially, but durable productivity comes from application to real work.
The third layer is advanced workflow capability for managers, process owners and power users. These employees should understand how to standardise useful methods, evaluate tools, design workflows, set permissions and introduce automation responsibly. In a small organisation, one well-trained internal champion can have a disproportionate effect by helping colleagues document good practice and avoid common mistakes.
This model also recognises that AI development is continuous. Tools and capabilities are changing quickly, so a single annual workshop is unlikely to be sufficient. Short practical updates, internal demonstrations, shared workflow libraries and periodic policy reviews can sustain capability more effectively.
Measurement should focus on behaviour and outcomes
AI training should not be judged by attendance alone. Organisations should look for changes in work. Are staff checking sources more consistently? Are repetitive tasks taking less time? Are teams producing stronger first drafts? Are confidentiality problems being prevented? Are useful workflows being documented and shared? These measures provide a better indication of capability than the number of people who have completed a course.
There is also a risk that employers define productivity too narrowly. AI can save time, but it can also improve the quality of work by allowing people to compare more alternatives, identify gaps in an argument or communicate more clearly. A skilled user may not simply perform the same task faster; they may produce a better result. Training should therefore develop critical judgement as well as efficiency.
Conclusion
The eBSI AI Skills Academy is built around this practical view of capability. Its purpose is not merely to familiarise learners with product interfaces, but to develop transferable skills that remain useful as platforms change: problem framing, verification, workflow design, responsible information handling and professional judgement. The OECD's message should be encouraging for businesses because it suggests that the benefits of AI are not reserved for organisations with the largest technology budgets. Companies that invest in practical skills and clear processes can create significant value with tools that are increasingly accessible to everyone.
Source
OECD (2026), AI and skills: What we know so far, OECD Publishing, Paris. View the OECD policy brief.
What employers should do with this evidence
The OECD findings suggest that employers should treat AI capability as part of workforce planning rather than as a one-off technology initiative. A practical starting point is to identify which roles are already changing, which tasks are suitable for augmentation and which employees require deeper capability because they supervise automated processes or make high-impact decisions. This allows training expenditure to be connected to actual business need rather than distributed uniformly.
Employers should also recognise that the value of domain expertise may increase as AI becomes more capable. When first drafts, summaries and routine analysis become inexpensive, the scarce capability shifts toward evaluating whether an answer is appropriate, interpreting exceptions and understanding the wider consequences of a decision. Professional development should therefore combine AI literacy with the underlying subject knowledge required to challenge AI output. In this sense, the future workforce is not simply more technical. It is more interdisciplinary, combining digital confidence with judgement, communication and specialist expertise.