Artificial intelligence is forcing organizations to rethink job roles at a speed few learning functions have experienced before. Tasks are being automated. Workflows are being redesigned. Decision-making is shifting between people and technology. Activities once performed by experienced employees are increasingly supported—or performed—by AI.
For learning and development leaders, the instinctive response is understandable: identify the new skills employees will need and build the programs to develop them.
But that sequence has a problem. By the time L&D is asked what people need to learn, some of the most consequential decisions may already have been made. The workflow has changed. Technology has been selected. Tasks have been redistributed. Roles have been redefined.
Only then does someone ask: “How do we train people for the new way of working?”
In an era of AI transformation, that is, increasingly, too late.
The emerging challenge for L&D leaders is not how to become better at reskilling employees. It is moving upstream to help shape how work changes and define the resulting capability requirements.
A useful example comes from Singapore’s health care system.
A different starting point
The Centre for Healthcare Innovation at Tan Tock Seng Hospital developed an approach known as the CHI Innovation Cycle. The cycle connects three elements: Care and process redesign → automation, IT and robotics → job redesign
For L&D leaders, the lesson here lies in the sequence of the center’s transformation. Using continuous Plan-Do-Study-Act cycles allows changes to be tested, evaluated and refined.
To start, organizations must first reconsider the work. They must examine processes, challenge inefficiencies and design for a desired future state. Then, teams introduce technology where it can enable future workflow.
Only after these steps can job roles be revamped through approaches such as upskilling, shifting appropriate work between occupational groups and expanding employees’ contribution into higher-value activities.
In many organizations, learning happens after the final stage. Once the redesigned job role is established, L&D receives a competency list, develops training and supports implementation.
However, involving a learning function earlier is much more strategic.
Skills requirements are consequences of work design
Organizations frequently begin workforce development by asking: “What skills will we need?”
But skills do not exist independently from work. What employees need to know and be able to do depends on what work they will perform, which decisions they will own, what technology will support them and what accountability will remain human. Change those variables and the required capabilities change with them.
Consider a workflow in which AI takes over the first draft of an analysis that previously required several hours of manual work. It would be easy to conclude that employees simply need “AI skills,” but that tells L&D very little.
The more useful questions are: What will employees do with the time released? Are they expected to validate AI output? Interpret it? Challenge it? Combine it with contextual information? Advise a client? Make the final decision?
Each answer reveals a different capability requirement. This is why L&D may not design reliable future capabilities without understanding future work. The implication is significant: Capability architecture should follow work architecture. And L&D leaders should have a voice in designing both. Do not automate away the learning system.
Another reason L&D needs to be involved upstream: Some work does more than produce an output. It develops expertise.
Junior employees build judgement by researching, preparing first drafts, handling straightforward cases, observing consequences and receiving feedback. Managers develop decision-making ability partly because they have previously performed the operational work beneath those decisions.
AI is capable of taking over many of those activities. This can improve productivity. But if organizations remove developmental work without considering how expertise will be built in its absence, they may inadvertently weaken their future capability pipeline.
The question, therefore, cannot be only: Can AI perform this task?
L&D leaders should ask another: What capability has historically been developed through performing this task, and how will we develop it if the task disappears?
That changes the nature of L&D’s contribution. Instead of designing training after automation, the learning function helps leaders understand the capability consequences of automation before implementation.
This is workforce development by design rather than by reaction.
Redesign should create better work, not simply less work
The CHI cases illustrate another principle that L&D leaders should pay attention to: that streamlining work is only half of a successful redesign.
The more strategic question is what replaces it. In one reported transformation, the “Ward of the Future,” physical redesign, workflow changes and technological support were combined with role changes. Nurses subsequently spent 24.8 percent more time providing direct patient care and walked an average of 3.9 kilometers less per shift. Job satisfaction improved, while staff attrition fell from 8 to 6 percent.
The important lesson is not simply that technology saves time. Human capacity was redirected. That distinction is critical as organizations introduce generative and agentic AI.
Suppose automation releases 20 percent of an employee’s capacity. Discussing productivity may focus on the hours saved. A talent discussion would focus on:
- What higher-value contribution will now occupy that capacity?
- Will employees solve more complex problems?
- Spend more time with customers?
- Exercise greater judgement?
- Coach colleagues?
- Interpret information rather than assemble it?
- Develop new services?
- Improve processes?
If no deliberate solutions exist, automation can make space without creating greater capability or greater value. L&D and talent leaders have an opportunity to connect the two.
From course design to work-capability
