What did the Federal Reserve say about AI and economic uncertainty?

Michael Barr, the Federal Reserve's vice chair for supervision, recently stated that artificial intelligence is not a guaranteed productivity shortcut nor a clear path to lower interest rates. While AI is often described as efficiency-enhancing and disinflationary, Barr cautioned that in the short term, it may disrupt labor markets, strain energy systems, and even contribute to inflation. This perspective challenges the assumption that AI adoption automatically translates into macroeconomic stability. For M&E professionals, it reinforces the need to separate technological potential from measurable outcomes.

• ▹ AI may increase short-term inflation due to transition costs.
• ▹ Labor market disruptions could precede productivity gains.
• ▹ Energy systems face new strains from AI infrastructure.

Why does AI uncertainty matter for Monitoring & Evaluation?

AI uncertainty matters for M&E because it exposes the gap between innovation narratives and on-the-ground realities. Evaluators are trained to ask: who benefits, who bears the costs, and under what conditions? The Federal Reserve's warning validates that AI impacts are contingent, not automatic. In development contexts, where systems are already fragile, early disruptions can deepen inequality. M&E frameworks must therefore track both short-term harms and long-term gains, ensuring that policy decisions are guided by evidence rather than technological optimism.

• ▹ Disruption may hit vulnerable workers and informal economies first.
• ▹ Attribution of outcomes becomes more complex with AI interventions.
• ▹ Evaluation provides accountability in uncertain transitions.

How does AI challenge traditional productivity assumptions?

Traditional economic models assume that technological adoption yields linear productivity improvements. AI challenges this assumption because its effects are mediated by infrastructure, skills, and regulatory capacity. The Federal Reserve's analysis highlights that productivity gains may be unevenly distributed across sectors and geographies. For international development, this means that AI investments in health, agriculture, or education may not produce immediate returns. M&E professionals must design evaluations that capture delayed effects, unintended consequences, and the redistribution of economic opportunities.

• ▹ Productivity gains depend on complementary investments in digital literacy.
• ▹ Sectoral variations require tailored evaluation indicators.
• ▹ Short-term costs may outweigh benefits in early adoption phases.

What are the key M&E challenges in an era of AI uncertainty?

AI uncertainty introduces several specific challenges for M&E practitioners. First, the opacity of AI systems makes it difficult to trace causal pathways from inputs to outcomes. Second, the speed of AI evolution demands continuous monitoring rather than periodic evaluations. Third, distributional effects—such as labor displacement or energy consumption—require new types of data and interdisciplinary methods. The Federal Reserve's emphasis on short-term disruption underscores that evaluators must be equipped to detect harms early, even when long-term benefits are anticipated.

• ▹ Causal attribution: separating AI effects from other factors.
• ▹ Real-time data: monitoring labor and inflation impacts as they emerge.
• ▹ Equity metrics: tracking whether vulnerable groups bear disproportionate costs.

How can evaluation frameworks address AI-related disruptions?

Evaluation frameworks can address AI disruptions by incorporating principles from developmental evaluation and complexity-aware methods. Theories of change must explicitly map assumptions about how AI leads to productivity or welfare improvements—and what could go wrong. Indicators should track not only efficiency but also resilience, equity, and environmental sustainability. The Federal Reserve's caution about energy strain, for example, points to the need for evaluating the carbon footprint of AI systems. Adaptive management approaches allow programs to pivot when early evidence shows negative side effects.

• ▹ Develop theories of change that include disruption pathways.
• ▹ Use mixed methods to capture quantitative and qualitative impacts.
• ▹ Build feedback loops for mid-course corrections.

What role do international organizations play in evaluating AI?

International organizations such as the World Bank, OECD, and UNESCO are increasingly shaping how AI is evaluated in development contexts. They provide normative frameworks, technical guidance, and funding for M&E capacity building. The OECD's AI Principles emphasize transparency and accountability, while UNESCO's recommendations focus on ethical impact assessments. These institutions can help standardize indicators for cross-country comparisons and ensure that evaluation findings inform global policy. The Federal Reserve's intervention adds macroeconomic credibility to the argument that evaluation must be central to AI governance.

• ▹ World Bank: funds AI-for-development projects with embedded M&E.
• ▹ OECD: develops indicators for trustworthy AI.
• ▹ UNESCO: promotes ethical impact assessments.

Frequently asked questions about AI uncertainty and evaluation

Authoritative resources and further reading