[01]Article

Census Data Reveals AI's Real Winners: Not Your Typical Tech Functions

New BTOS survey shows customer service and operations beating engineering in productivity gains from AI adoption.

James Roycroft-Davis··2 min read·For operators

The Census Bureau's new AI supplement data dropped this week, and it's making venture capitalists uncomfortable. While VCs have been pouring billions into AI tools for developers and data scientists, the actual productivity gains are showing up in decidedly unglamorous places.

According to the 2026 Business Trends and Outlook Survey, 17.3% of U.S. businesses now use AI — but the distribution across functions tells a different story than Silicon Valley's pitch decks.

The Productivity Paradox

The survey's most striking finding: customer service departments report the highest productivity gains from AI adoption, followed by operations and supply chain management. Engineering and R&D? They're seeing adoption but minimal productivity impact.

This tracks with what operators are seeing on the ground. Customer service teams using AI for ticket routing and response drafting report 30-40% efficiency gains. Operations teams deploying AI for demand forecasting and inventory optimization cite similar numbers.

Meanwhile, engineering teams struggle to quantify productivity improvements from GitHub Copilot and other coding assistants. The tools are popular — adoption is high — but the measurable impact on output remains fuzzy.

Small Firms Lead the Charge

Another surprise: the smallest businesses are outpacing their larger peers in AI adoption. Self Employed reports that firms with fewer than 10 employees show higher AI usage rates than mid-size companies.

The reason is simple: small firms focus on immediate ROI. They're buying $20/month ChatGPT subscriptions for customer service, not $100K enterprise AI platforms for "digital transformation."

Large enterprises report extensive AI pilots across multiple functions. Small businesses report actual deployment in one or two areas with clear payback periods.

The Federal Reserve Connection

Federal Reserve data confirms the pattern: worker-level AI use is outpacing firm-level adoption metrics. Employees are bringing their own AI tools to work, especially in customer-facing roles.

This bottom-up adoption explains why customer service and operations see bigger gains. Individual workers can immediately apply AI to repetitive tasks without waiting for IT approval or integration projects.

The Census data also reveals that 20% of firms plan to adopt AI in the first half of 2026. But based on current patterns, the real action will be in function-specific tools solving narrow problems, not platform plays promising to "revolutionize your business."

For operators, the message is clear: stop looking for AI transformation in the places VCs point you. The real productivity gains are happening where the work is most repetitive and the implementation is simplest.

Your customer service team's ChatGPT subscription might deliver more value than your engineering team's fancy code completion tool.

[02]Sources

  1. The Microstructure of AI Diffusion
  2. Federal data shows AI adoption is broad, but uneven - TechInformed
  3. The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks | NBER
  4. Census Bureau Releases New BTOS Data on AI Adoption | GovPing
  5. Census Data Shows Smallest Firms Leading Small-Business AI Adoption - Self Employed

Ready to put this into practice?

Get a Human in Residence
Build your team →