The workforce capacity problem AI creates: the operating model decisions it requires

WEF 2025 research finds that half of leaders already report 10 to 20% overcapacity in their workforce due to automation. By 2028, 40% of leaders expect that excess to reach 30 to 39%. These figures describe a structural shift that is already in progress in many mid-market organisations. Most have not named it. The standard framing is productivity gains: AI frees time for higher-value work. The research finding is more specific. AI is creating excess capacity that the organisation has not yet allocated. That unallocated capacity will either be redeployed to higher-value work through deliberate design, or it will be absorbed by existing work patterns until it is no longer visible in any metric.

Excess capacity created by AI is an operating model question before it is an HR question. Three decisions must be made explicitly. First: which roles will be redesigned to absorb the capacity AI frees up in genuinely higher-value work? The answer requires identifying what higher-value work looks like in each function, which is a workflow design exercise rather than a policy statement. Second: which functions will be scaled to do more with the same headcount using the capacity AI releases? Third: which redeployment paths exist for practitioners whose current roles are substantially automated? These decisions require explicit design. They do not emerge from a general commitment to reskilling. An organisation that deploys AI workflows without designing the capacity redeployment path is creating a structural tension that will eventually resolve itself through either attrition or reduction.

The organisations that realise productivity gains from AI workflows are the ones that answer the capacity question before the programme ends. For each workflow in design or production, they quantify the capacity it releases, then answer two critical questions: where does that capacity go, and who has the authority to decide? When these questions have documented answers, the capacity is redeployed to higher-value work and the productivity gain appears in the business metrics that motivated the investment. When neither question has a documented answer, the capacity is absorbed by existing work patterns until it is no longer visible. The productivity gain that the AI workflow was designed to produce never appears in any metric because the freed time was never deliberately reallocated. Why does timing matter? The capacity question must be answered while the governance structures and design authority of the programme are still in place. After the programme ends, the competing priorities of a functioning organisation make it a far harder conversation to have. The capacity redeployment decisions made during the programme window compound. Those deferred to after programme close typically do not happen at all.