How we read the evidence
Not all numbers deserve equal weight. We ranked sources in this order, and the briefing leans on them accordingly.
- Official statistics and randomized trials. Representative samples or controlled designs. Slow, narrow, and the most reliable.
- Independent analysts. No implementation work to sell. Often forecasts, and labeled as such.
- Consultancy, vendor and investor surveys. Large and current, but self-reported, drawn mostly from large companies, and published by firms that sell the remedy. Useful for direction, weak for magnitude.
- Executive voices. Essays, interviews and podcasts. Illustrative, never counted as data.
Every figure is given with the question that produced it, because wording moves these numbers more than anything else. Where we could not trace a widely quoted statistic to its publisher, we left it out. Samples with a known skew are flagged where they appear. Observations marked "What we have seen" are ours, drawn from client work with details changed. We are an interested party too. This list is here so you can check us.
What the evidence does not settle
Jobs
Employers cited AI for about 22% of announced US job cuts this year (Challenger, Gray & Christmas). Yet 4% of service firms in the New York Fed's region report AI layoffs, and Yale's Budget Lab sees no economy-wide disruption. Stanford payroll data shows entry-level workers in exposed occupations about 19% below their peers, through slower hiring.
Where the saved time goes
42% of frontline regular users tell BCG they save about a day a week, and 66% get little guidance on what to do with it. Stanford and BetterUp find 40% of desk workers received AI "workslop" in the past month.
Developer productivity
Field trials run from 26% more tasks completed to 19% slower. No clean randomized trial yet tests this year's agents.
The macro effect
US labor productivity is growing 2.4% a year against 1.6% before the pandemic, concentrated in AI-exposed sectors (Dallas Fed). Efficiency-adjusted productivity is flat, and about 95% of what companies say about AI and productivity on earnings calls is forward-looking (St. Louis Fed).
Gaps we could not fill
No published study measures the profit impact of AI in mid-sized companies. Most marquee surveys exclude firms under $500 million in revenue. We found no credible figure for who leads AI in mid-sized firms, and no rigorous measure of how many pilots reach production.
Sources
Government and central-bank statistics
US Census Bureau, Business Trends and Outlook Survey, "Large Firms… Biggest AI Users," May 2026.
Bonney, Breaux, Dinlersoz, Foster, Haltiwanger, Pande, The Microstructure of AI Diffusion, Census CES-WP-26-25 / NBER w35141, Apr 2026.
Federal Reserve Bank of St. Louis, "Measuring AI Adoption among Firms: How You Ask Matters," Jun 2026.
Federal Reserve Bank of St. Louis, Generative AI Adoption Tracker, Sep 2026.
Federal Reserve Bank of St. Louis, "AI and Productivity: What Firms Are Saying on Earnings Calls," Jul 2026.
Federal Reserve Board, FEDS Notes on AI adoption, Mar–Apr 2026.
Federal Reserve Bank of New York, Liberty Street Economics, "Businesses Are Using AI to Transform Work, Not Cut Jobs," Sep 2026.
Federal Reserve Banks of Atlanta and Richmond / Duke, CFO Survey, NBER w34984, Mar 2026 (n=748).
Federal Reserve Banks of San Francisco (Economic Letter 2026-14) and Dallas (Jul 2026), productivity analyses.
Eurostat, "20% of EU enterprises use AI technologies," Dec 2025.
UK Office for National Statistics, "Artificial intelligence in UK businesses: 2023 to 2026," Jul 2026.
Academic studies and worker surveys
Yotzov, Barrero, Bloom, Davis et al., Firm Data on AI, NBER w34836, 2026 (~6,000 executives).
Brynjolfsson, Li, Raymond, "Generative AI at Work," Quarterly Journal of Economics, 2025.
Dell'Acqua et al., "Navigating the Jagged Technological Frontier," Organization Science.
Cui, Demirer, Jaffe et al., "The Effects of Generative AI on High-Skilled Work," Management Science.
METR, developer-productivity trial, Jul 2025, and design update, Feb 2026.
Humlum and Vestergaard, NBER w33777, rev. Mar 2026.
Brynjolfsson, Chandar, Chen, "Canaries in the Coal Mine?", Stanford Digital Economy Lab, Aug 2026 update.
Yale Budget Lab, AI and labor-market tracker, 2026.
Stanford HAI, AI Index Report 2026, ch. 4.
MIT Project NANDA, The GenAI Divide, 2025 (52 interviews; not peer reviewed).
BetterUp Labs and Stanford Social Media Lab, "Workslop," Harvard Business Review, Sep 2025.
Bean and Davenport, 2026 AI & Data Leadership Executive Benchmark Survey.
Gallup, AI in the workplace indicator, May 2026.
Pew Research Center, Americans and AI, Mar 2026.
Consultancy and analyst surveys
McKinsey, The State of AI 2026 (n=1,719).
KPMG, Global AI Pulse Q2 2026 (n=2,145).
KPMG, Transforming the Enterprise 2026 (n=1,750).
Deloitte, State of AI in the Enterprise 2026 (n=3,235; fielded Aug–Sep 2025).
PwC, 29th Global CEO Survey, Jan 2026 (n=4,454).
PwC, AI Agent Survey, May 2025 (n=308).
BCG, AI Radar 2026 (n=2,360).
BCG, The Widening AI Value Gap, Sep 2025 (n=1,250).
BCG, AI at Work 2026 (n=11,749 workers).
Accenture, Pulse of Change, Jul 2026 (3,000 executives, 3,000 employees).
EY, US AI Pulse Survey, Wave 5, Jul 2026 (n=534).
IBM Institute for Business Value, 2026 CEO Study (n=2,000).
Capgemini Research Institute, The Multi-Year AI Advantage, Jan 2026, and Rise of Agentic AI, Jul 2025.
Gartner, agentic AI forecast, Jun 2025; 2026 CIO survey.
Forrester, Predictions 2026 and AI jobs forecast.
S&P Global Market Intelligence, Voice of the Enterprise: AI & Machine Learning (n=1,006; fielded late 2024).
Wharton Human-AI Research and GBK Collective, Accountable Acceleration, Oct 2025.
RSM, Middle Market AI Survey 2026 (n=1,030).
McKinsey, "To centralize or not to centralize: gen AI in banking."
National Center for the Middle Market, Mid-Year 2026 Indicator; JPMorganChase, 2026 Business Leaders Outlook.
Investor and market data
Menlo Ventures, 2025: The State of Generative AI in the Enterprise.
Andreessen Horowitz, enterprise CIO survey, Jan 2026 (n=100).
ICONIQ, 2026 State of AI Report.
Battery Ventures, State of Enterprise Tech Spending, Dec 2025.
Challenger, Gray & Christmas, job-cuts report, Aug 2026.
Fast Company / Indeed, forward-deployed engineer postings, Nov 2025.
AI-lab, vendor and regulatory sources
Anthropic, Economic Index, Sep 2025 – Jun 2026.
Anthropic, Blackstone, Hellman & Friedman and Goldman Sachs, and OpenAI: services-company announcements, May–Jul 2026.
SmarterX / Marketing AI Institute, 2026 State of AI for Business (n=2,109; small-business and marketing skew).
NVIDIA, State of AI in Retail and CPG 2026 (vendor-sponsored).
IBM / Ponemon, Cost of a Data Breach 2026; EU AI Act Digital Omnibus, 2026.
Executive voices and company disclosures
Jack Dorsey and Roelof Botha, "From Hierarchy to Intelligence," Block, Mar 2026.
Gill Haus, CIO, Chase, interviewed by ZDNET, Sep 2026.
WSJ CIO Journal, Technology Council Summit coverage, Sep 2026.
Anish Acharya, "Why Companies Are Becoming a Series of Loops," The a16z Show, Sep 2026.
The Enterprise AI Show #1040, "Enterprises Are Concerned about AI Costs, Governance and Trust," Jun 2026.
Jim Kavanaugh, CEO, World Wide Technology, AI Proving Ground Podcast, Sep 2026.
Praveen Neppalli Naga, CTO, Uber, on agentic pods, Jul 2026.
JPMorgan Chase, LLM Suite and AI-spend disclosures, 2024–25.
Klarna, company statements, 2024; CEO remarks reported by Bloomberg, May 2025.