Workforce Decision Intelligence

Count the Learning Opportunities AI Leaves Behind

What Siemens, Ingka and current research reveal about maintaining the experiences that build expertise as AI changes work. Connect skills investment to supervised assignments, useful feedback and access to responsibility.

5 min read

PsychFlo Editorial

A young worker can finish a sophisticated task with AI assistance and still be unsure how to recognise when the result is wrong. That creates a difficult question for employers: how will tomorrow's experts acquire the judgement needed to supervise systems that increasingly perform today's learning tasks?

The answer will shape more than training policy. It will affect access to careers, the supply of experienced workers and who can participate in economic growth.

The World Economic Forum's June 2026 report on entry-level work, developed with PwC, finds that more than one in three young workers are in occupations with medium to high exposure to AI-driven task change. Exposure does not mean those jobs will disappear. It does mean that the work through which people learn may change alongside the work through which organisations produce.

As employers and public institutions plan for 2027, they should examine a measure that training totals miss: access to meaningful learning opportunities inside the job.

Look at the experiences behind a capability

An entry-level task often produces two things. It produces an immediate output and gives the person doing it a chance to build understanding.

Preparing a first analysis can teach someone to notice inconsistent definitions. Drafting a response can reveal the difference between an approved policy and an attractive promise. Working through a difficult exception can teach an employee when to seek help.

Automation may improve the immediate output while changing or removing that experience. The sensible response is to identify the learning function and deliberately replace it where necessary. Preserving every repetitive task would waste the opportunities AI offers. Assuming that learning will continue automatically would leave an important investment unplanned.

A useful starting question is: after this task is automated, where will someone practise the judgement it used to develop?

Separate performance from understanding

Evidence from a bounded experiment illustrates why the question matters. In research released by Anthropic in January 2026, 52 mostly junior software engineers learned an unfamiliar Python library. The AI-assisted group performed worse on a subsequent knowledge assessment, while the average speed improvement was not statistically significant. Some ways of using the assistant supported stronger understanding than others.

That study does not establish a universal effect of AI on learning or employment. It examines one programming setting and a small sample. Its practical warning is narrower: completing the immediate task and learning the underlying material are different outcomes.

Training programmes should therefore include an opportunity to explain, challenge or adapt an output, rather than relying entirely on the quality of AI-assisted submissions.

For a proposed assessment, give a learner a new case with an ambiguous source or changed constraint. Ask them to explain what remains valid and what must change. Such exercises need local testing and fair assessment criteria; they should not be presented as universal measures of professional competence.

Map opportunities before redesigning roles

Employers can create a simple inventory alongside their AI adoption plan. For each role being redesigned, identify the capability people must develop, the experience that currently develops it and the replacement opportunity available after automation.

Siemens provides a current example of keeping development inside the job. In its September 2026 training announcement, the company described AI integration across vocational training and dual-study programmes, with around 1,500 new entrants starting that year. It also reported more than 300 training positions for partner companies. These are programme commitments and activity figures, not proof of later career outcomes.

The inventory should also name the person who provides feedback and the time allocated to doing so. A learning opportunity without supervision can become an unsupported exposure to risk. A mentor without protected time can become another overburdened specialist.

Employers should evaluate whether redesigned opportunities improve understanding and allow people to progress.

Make access part of the design

Access also depends on time. In an August 2026 report on its learning pilots, Ingka Group described giving retail co-workers dedicated learning time during working hours in the United States and Switzerland. Its reported business comparisons are not causal evidence; the practical example is the decision to allocate learning time inside the working day.

Review who receives challenging assignments, feedback and supervised responsibility. Consider location, working pattern, disability access and the availability of training in the language people use at work. Use appropriate aggregate information and safeguards rather than creating intrusive profiles of individual potential.

Smaller employers may struggle to provide a complete pathway alone. Industry associations, colleges and groups of employers could share practice material, mentoring capacity and placements that offer progressively more responsibility. Public support could help fund these arrangements while evaluating whether participants gain useful capabilities and access to better work.

The important unit is an accessible development opportunity with feedback, not merely a seat on an online course.

Ask for evidence of progression

The Forum's entry-level framework connects job access, job design, talent pipelines and education. An inventory of learning opportunities could help employers and institutions translate those priorities into decisions about actual work.

Alongside participation totals, ask whether learners can handle a fresh case, find the relevant evidence and recognise when to escalate. Track whether supervised practice leads to additional responsibility. Interpret results in context; a short exercise cannot establish the future performance of an entire workforce.

AI can widen access to useful guidance and make previously difficult work more approachable. Realising that opportunity requires a deliberate route from assistance to understanding, and from understanding to responsibility.

A resilient workforce needs a continuing supply of people who can judge unfamiliar situations. As AI changes the tasks through which expertise develops, employers should count—and create—the opportunities that make that judgement possible.

AI assistance was used in research discovery, structure and drafting. Research findings and organisational examples are attributed to the linked sources.

← All research and insights

PsychFlo guide

Guided chatbot · prepared answers, no language model

Find the right course, software product or consulting review. Messages stay in this page’s memory; don’t enter personal or confidential information.

Full decision copilot