Part 1 of the six-part series From Course Factory to Capability Function.
Most L&D functions were built as course factories. A request comes in, a course goes out. Volume is the measure of success: courses launched, seats filled, completions logged. For years that was enough, because nobody asked what the courses changed.
AI ends that arrangement. When anyone can generate a polished course in an afternoon, production stops being the scarce resource. The scarce resource becomes judgment: knowing what the organization needs people to be capable of, and whether a course is even the answer.
That is the shift from course factory to capability function. It is not a rebrand. It changes what L&D is accountable for.
A course factory is measured on output. A capability function is measured on outcomes. The factory asks, “What training do you need?” The capability function asks, “What performance problem are you trying to solve, and what would tell us it is solved?” Those are different jobs, and most L&D teams are still staffed and funded for the first one.
The change shows up in six places. Courses give way to capability: the unit of work becomes what people can do, not what they have completed. Jobs give way to tasks and workflows, because AI changes work at the task level before it changes it at the job level. Knowledge transfer gives way to performance enablement: information is cheap now, applied performance is not. Broad programs give way to high-value use cases, the few places where capability actually moves the business. Completion metrics give way to performance evidence. And content production gives way to capability design, which starts from the work and works backward.
None of this means the end of training. It means a clearer understanding of when training is the right answer. Some problems are skill problems, and a well-designed course is still the right tool. Many are not. They are process problems, incentive problems, tool problems, or clarity problems, and no amount of training fixes those. A capability function knows the difference. A course factory treats them all the same.
Here is a practical test. Take the last five requests your team received. For each one, write down the performance problem in one sentence, then write down what evidence would show the problem is solved. If you cannot write either sentence, you are running a factory: taking orders without diagnosing. That is not a criticism of your team. It is a description of the operating model most of us inherited.
This series is about changing that model. Each installment takes one part of the transformation and makes it usable: how to lead L&D when AI changes the economics of production, how to stop treating every performance problem as a training problem, how to decide who should do the work, how to measure what matters, and what the future L&D function actually does day to day.
The umbrella over all of it is simple: solve the right problem. Before you build the course, before you buy the tool, before you automate the workflow, make sure you are working on the thing that actually needs fixing. Everything in this series follows from that.
