U.S. Census Bureau data · May 3, 2026

State of AI
Implementation

How many U.S. businesses actually use AI, who is furthest behind, and what stops the rest, built on U.S. Census Bureau figures rather than our own estimates.

Source: Business Trends and Outlook Survey (BTOS), December 14, 2025 – May 3, 2026. census.gov— U.S. Census Bureau BTOS, opens in a new tab

17% to 20%

Share of U.S. businesses reporting AI use across December 14, 2025 – May 3, 2026.

20% to 23%

Share expecting to use AI within the following six months.

The adoption gap is about size

The clearest split in the Census data is not industry, it is headcount. Large firms are roughly twice as likely to use AI as the smallest ones, which is exactly the gap most of our clients are standing in.

250+ employees37%

100–249 employees32%

Fewer than 20 employeesunder 20%

Census publishes this as a bound, not a point estimate.

Figures as of May 3, 2026. census.gov— U.S. Census Bureau BTOS, opens in a new tab

By sector

Information39.7%

Finance and insurance33.9%

All sectors (national rate)19.8%

Retail tradeabout 14%

Retail, manufacturing and wholesale trade all sit below the national rate.

Figures as of May 3, 2026. census.gov— U.S. Census Bureau BTOS, opens in a new tab

What stops implementations

What we see repeatedly in client projects. These are our own observations, not survey results, so they carry no percentages.

Data that isn't ready: Records live across a CRM, a spreadsheet and someone's inbox, so there is nothing clean for a model to work from.
No defined success metric: The goal is “use AI” rather than a named number to move, so nothing can be judged a success or a failure.
Buying a tool instead of designing a workflow: A subscription gets bought before anyone maps the process it is meant to replace, and the tool goes unused.
No owner after launch: Automations drift as the business changes and quietly stop matching how the team actually works.

“Most businesses fail because they try to ‘buy AI’ instead of ‘engineering a workflow.’ Success requires process-first thinking.”

What it costs to close the gap

This page used to publish improvement ranges, support response, admin time, lead conversion, with no study behind them. They are gone rather than dressed up.

What we can show you instead is what the components actually cost, with every figure linked to the vendor's own pricing page.

See the AI Agent Pricing Index

Methodology

What this is: Adoption figures published by the U.S. Census Bureau from its Business Trends and Outlook Survey (BTOS), quoted as published and dated. BTOS is a large, nationally representative survey of U.S. businesses.
What this is not: PxlPeak did not conduct this survey and does not run one. Nothing on this page is proprietary research, and no figure here comes from PxlPeak's own client data. The barriers section is explicitly labelled as field observation and carries no statistics.
Citing this data: Census figures are public domain; cite the Census Bureau directly rather than this page. census.gov— U.S. Census Bureau BTOS, opens in a new tab