The firm’s new ‘AI productivity gains and the performance paradox’ report concludes that most current AI applications ‘accelerate existing work’ without redesigning workflows, a finding McKinsey is publishing while targeting 1:1 parity between its 40,000 human consultants and 40,000 AI agents by year-end.
McKinsey’s strategy practice has published a new analysis arguing that the corporate world is grappling with what it calls an ‘AI paradox’: adoption of generative and agentic AI is growing, capital investment is accelerating, but ‘sustained impact on performance is elusive.’
The report, ‘AI productivity gains and the performance paradox’, argues that most current AI applications are ‘tools that accelerate existing work’ but ‘largely preserve underlying workflows’, and that the larger productivity gains will only emerge once organisations redesign processes around AI rather than simply bolting it on top.
The report’s central historical analogy is electricity in factories. “When electricity first arrived in factories, many businesses simply replaced the steam engine with an electric motor, capturing efficiency gains but leaving the line-shaft layout unchanged,” the authors write.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
“The breakthrough came later, when small motors enabled managers to rearrange machines around workflows, and ultimately when companies redesigned their factories around electricity, creating new operating models.”
General-purpose technologies, McKinsey argues, “rarely create value in a single wave.”
McKinsey’s recommendations to executives reading the analysis are three: assess how AI will reshape industry profit pools; build or strengthen AI-powered competitive moats; turn speed into a structural advantage.
The report cites JPMorgan Chase’s real-time AI fraud detection, BMW’s computer vision quality inspection, and Siemens’ AI-coordinated predictive maintenance as examples of the work-acceleration tier, and contrasts them with deeper process redesigns that move companies beyond what McKinsey has elsewhere called the ‘gen AI paradox.’
The skeptical evidence McKinsey is publishing into
The report lands in a context where the gap between AI investment and measurable returns has become impossible to ignore. The Federal Reserve Bank of St. Louis has measured 1.9% excess cumulative productivity growth since ChatGPT launched in November 2022, a meaningful figure but well below the rates required to justify current AI capital spending.
JPMorgan published a capex analysis warning that $650 billion in annual revenue would be needed ‘into perpetuity’ to deliver a 10% return on current AI infrastructure investment, and drew a direct parallel to the late-1990s telecom fibre buildout, where the infrastructure was laid, the revenue never arrived fast enough, and the investors who funded it were wiped out.
MIT Media Lab research has found that 95% of organisations see no measurable returns from AI adoption. Deloitte’s 2026 ‘State of AI in the Enterprise’ report, surveying 3,235 director-to-C-suite leaders, found that 66% report productivity gains from AI but only 20% report revenue growth, and only 34% are using AI to deeply transform products or processes.
PwC’s 2026 Global CEO Survey, covering 4,454 CEOs across 95 countries, found that 56% say they have gotten ‘nothing out of’ their AI investments, and only 12% report AI both growing revenues and reducing costs. Workday’s 2026 research found that 37–40% of the time AI is supposedly saving gets eaten up by reviewing, correcting, and verifying AI-generated output.
OpenAI president Greg Brockman’s claim that AI is now writing 80% of OpenAI’s code sits alongside a February 2026 NBER study finding that 80% of companies actively using AI report no productivity impact at all.
The macro picture is complicated by a divergence in expert estimates large enough to render the ‘is AI working’ question unanswerable from public data alone. McKinsey itself has previously estimated that AI could add $4.4 trillion to the global economy.
Nobel laureate Daron Acemoglu has projected a ‘modest 0.5% productivity gain over the next decade.’ The gap between those two figures, a hundredfold difference, depending on which is taken as the lower bound, is the gap inside which every enterprise AI capital allocation decision is being made.
McKinsey’s own AI deployment
What gives McKinsey’s skeptical framing particular force is the firm’s own simultaneous AI deployment. McKinsey CEO Bob Sternfels said at CES 2026 in January that the firm runs 25,000 AI agents alongside its 40,000 human consultants, and expects to reach 1:1 parity, 40,000 AI agents, by the end of this year.
McKinsey saved 1.5 million hours in search and synthesis work last year alone, and back-office output increased 10% with 25% fewer people. Client-facing roles, engagement managers, senior consultants, strategic advisors, grew by 25%, while research analyst, data processor, and administrative support positions shrank by the same proportion.
The firm is not arguing against AI productivity in any abstract sense. It is arguing that most companies are not capturing the productivity gains it is itself capturing, because most companies are not redesigning workflows the way McKinsey has.
This makes McKinsey simultaneously the most credible mainstream voice on the AI productivity paradox and the firm whose consulting work most directly depends on selling enterprises a solution to that paradox.
The recommendations in the new report, redesign workflows, build AI-powered moats, develop a structural advantage in speed, map almost exactly to the kind of multi-year transformation engagements McKinsey sells.
Whether that is a sign that the firm has independently identified what works, or a sign that the framing has been calibrated to position McKinsey’s services at the centre of the answer, is a question the report itself does not address.
The McKinsey analysis lands the same week the largest AI infrastructure spenders in the world disclosed their Q1 2026 results. Combined 2026 capex across the five major hyperscalers is now on track to exceed $650 billion.
Google Cloud grew 63%, AWS grew 28%, and Meta raised its full-year capex guidance to $125–$145 billion. The capex commitments at this scale are themselves only justifiable if the productivity gains McKinsey describes as elusive eventually materialise across the corporate sector that buys hyperscaler services.
Of the McKinsey high performers, defined in the firm’s November 2025 ‘State of AI’ report as the roughly 6% of respondents attributing 5% or more EBIT impact to AI, the distinguishing characteristics were not better technology choices but better organisational practices: redesigning workflows, scaling faster, embedding AI into business processes, tracking KPIs for AI solutions, having senior leadership demonstrably committed to the work.
The latest report extends the same finding into a strategic playbook. The implicit question for executives reading it is whether their organisations look more like the 6% or like the 80% in the NBER paper that report no productivity impact at all.
What this means for the AI capex thesis?
The McKinsey report does not predict that the AI capital cycle will fail. It argues, on a careful reading, the opposite, that AI will eventually generate substantial value, that the value will accrue disproportionately to early movers who redesign workflows ahead of competitors, and that the strategic risk for executives is moving too slowly rather than too aggressively.
The report’s recommendation that companies turn speed into a structural advantage is, on its face, an argument to spend more, not less.
But the framing of the paradox, from the firm whose CEO has publicly committed to a 1:1 ratio of AI agents to human consultants, marks a notable rhetorical shift. Until very recently, the dominant mainstream consulting voice on enterprise AI was a story of accelerating value capture.
The new McKinsey position is more careful: that value is real, but it is concentrated, conditional, and not arriving on the timeline that current capex commitments imply. For the corporate boards approving AI infrastructure spending against expected returns three to five years out, that distinction is the entire investment thesis.
Whether the AI capital cycle resembles the long-tail value creation of railroads and electricity, or the wipeout of late-1990s telecom fibre, depends on which side of McKinsey’s paradox a given company ends up on.
