McKinsey's latest AI report claims that the productivity benefits are genuine but dependent on certain conditions.

      The firm’s latest report titled "AI Productivity Gains and the Performance Paradox" concludes that the majority of existing AI applications "accelerate current work" without fundamentally altering workflows. McKinsey is releasing this report while aiming for parity between its 40,000 human consultants and 40,000 AI agents by the end of the year.

      McKinsey's strategy practice has released an analysis stating that businesses are facing what it refers to as an "AI paradox." While the use of generative and agentic AI is increasing and capital investments are rising, achieving a lasting impact on performance remains challenging.

      The report asserts that most AI applications are "tools that speed up existing work" but largely maintain existing workflows. It suggests that significant productivity improvements will arise only when organizations restructure their processes around AI rather than merely adding it as an extension.

      The report draws a historical comparison to the introduction of electricity in factories. "Initially, many businesses simply exchanged the steam engine for an electric motor, gaining efficiency but keeping the line layout the same," the authors explain. "Real breakthroughs occurred later, when small motors allowed managers to rearrange machines according to workflows, and ultimately when companies redesigned their factories around electricity, leading to new operational models."

      According to McKinsey, "general-purpose technologies rarely generate value in a single wave."

      The report offers three key recommendations for executives: evaluate how AI will transform industry profit pools; strengthen AI-enhanced competitive advantages; and convert speed into a structural benefit.

      Examples cited in the report include JPMorgan Chase’s real-time AI fraud detection, BMW’s computer vision quality inspections, and Siemens’ AI-driven predictive maintenance, which represent the work-acceleration level. These contrast with more profound process redesigns that take firms beyond what McKinsey has termed the "gen AI paradox."

      The context for the report highlights the widening gap between AI investment and tangible returns. The Federal Reserve Bank of St. Louis noted a 1.9% excess in cumulative productivity growth since the launch of ChatGPT in November 2022—a noteworthy figure, yet significantly lower than the rates necessary to warrant current AI capital expenditures.

      JPMorgan's capital expenditure analysis indicated that $650 billion in annual revenue would be needed indefinitely to achieve a 10% return on current AI infrastructure investments, paralleling the telecom fiber buildout of the late 1990s, where infrastructure was established but revenue growth fell short, leading to significant losses for investors.

      Research from MIT Media Lab revealed that 95% of organizations do not see measurable returns from adopting AI. Deloitte's 2026 "State of AI in the Enterprise" report, which surveyed 3,235 leaders from director to C-suite, found that while 66% experienced productivity gains from AI, only 20% reported revenue growth, and just 34% used AI to significantly transform products or processes. PwC's 2026 Global CEO Survey, covering 4,454 CEOs across 95 countries, revealed that 56% felt they derived "nothing" from their AI investments, with only 12% indicating AI had both driven revenue growth and cut costs. Research by Workday indicated that 37–40% of the time AI is expected to save is consumed by reviewing, correcting, and validating AI-generated outputs.

      OpenAI president Greg Brockman's assertion that AI now produces 80% of OpenAI’s code aligns with a February 2026 NBER study showing that 80% of companies actively using AI report no impact on productivity whatsoever.

      The broader picture is complicated by a disparity in expert forecasts that is significant enough to make the question of "is AI working" unanswerable based solely on public data. McKinsey has previously estimated that AI could contribute $4.4 trillion to the global economy, while Nobel laureate Daron Acemoglu has anticipated a more modest 0.5% productivity gain over the next decade. This stark discrepancy—up to a hundredfold difference depending on the estimation used as the lower bound—underlines the uncertainty guiding decisions related to AI capital allocation.

      Adding weight to McKinsey’s skeptical viewpoint is the firm's concurrent deployment of AI. McKinsey CEO Bob Sternfels announced at CES 2026 in January that the firm operates 25,000 AI agents alongside its 40,000 human consultants, expecting to achieve a 1:1 ratio of 40,000 AI agents by year's end.

      In the past year, McKinsey saved 1.5 million hours on search and synthesis work, with back-office output increasing by 10% while employing 25% fewer people. Client-facing roles, including engagement managers and senior consultants, grew by 25%, while research analysts, data processors, and administrative support roles decreased correspondingly.

      McKinsey is not denying AI productivity in any general sense; rather, it asserts that most companies are not realizing the productivity benefits that it itself has achieved because they are not redesigning workflows in the same manner.