The job description was never a good description of work — it was a container defined by organisational logic. As AI agents become capable of performing specific, well-defined tasks and platform architectures make it easier to route work dynamically, the job as the primary unit of workforce design is becoming a liability.
Jobs measure headcount; work units measure what actually needs doing
- When organisations think in jobs, they think in headcount: how many people do we need in this role? A work-unit decision asks a structurally different question: what work needs to get done, what capabilities does it require, and which combination of human and AI capacity is the right answer for this specific set of work units? As AI capability expands, the difference between these two approaches will determine which organisations can deploy capacity flexibly.
- Workers who understand their work in terms of the work units they execute can evaluate their own AI-adjacency with precision. Workers who understand their work only at the level of job title cannot — they will be the last to see the change coming and the slowest to adapt.
Map your team in work units to see what AI can already do
Map one team's work in work units rather than job titles: for each work unit, what is the input, the output, the capabilities required, and what proportion of that capability could AI provide today? You do not need to be precise — an order-of-magnitude estimate will reveal which work units are most exposed to AI substitution and which require human judgment AI cannot yet replicate. D5 places work-unit architecture at the centre of how organisations design human-AI capability combinations for Economy 4.0.
