The May 2026 NBER Digest presents Firm Data on AI, Working Paper 34836. The research draws on a survey of nearly 6,000 executives in four countries and examines AI use, reported effects and expectations. The paper was issued in February and revised in March; the Digest provides the May summary.
Adoption and realised effects are distinct. Expectations about productivity or employment differ from reported changes over a past period. Survey responses are evidence about what respondents report, rather than an automatic causal estimate of an organisation’s technology programme. The distinction matters when workforce teams use such research in planning.
Define the employer’s own questions This editor recommends deciding which outcome is being examined before selecting a measure. Access to a tool, use of a tool, task time and a staffing change are separate questions. A review should state the unit being measured and period covered so a result can be interpreted without relying on the broad label AI adoption. The selected measure should correspond to the decision the organisation intends to make.
An operational change should also be separated from a forecast. A manager’s expectation can inform planning, but should not enter a dashboard as a realised benefit. Maintain separate records for observations, assumptions and proposed actions. This makes a decision easier to revisit when later evidence becomes available and allows uncertainty to remain visible to the people reviewing the proposal.
Connect evaluation with people decisions This editor recommends asking how a proposed workforce change affects training, employee support and access to benefits. A technology business case should explain its people assumptions. Identify who reviews those assumptions and how employees can raise practical questions about changing work. Those responsibilities should be agreed before a programme moves from a proposal into an operational change.
Evaluation needs comparable measures before and after implementation, with limitations stated. Document other changes that may affect the result instead of attributing every movement to technology. Qualitative feedback can help explain a measure if collected through an appropriate process. The reporting team should know how findings will be assessed and what would justify further investigation.
The survey offers a reason to be precise. An employer can recognise growing adoption while requiring separate evidence for claims about productivity, staffing or financial returns in its own organisation. That evidence should support a defined decision rather than simply reinforce an existing expectation.
Sources: Source de référence [1]
Share this:
- More [3]
