Enterprise AI Adoption in 2026: Large vs Small Firms

If you lead marketing at a mid-size firm, the useful question is not whether your team uses AI. It is whether larger rivals have moved from individual tools to systems used across the company.

Here, enterprise scaling means moving AI beyond isolated teams or pilots so it is used across the organization. Scaling AI agents means deploying agents beyond experiments in at least one business function.

The agent comparison rests on 595 respondents at organizations with $1 billion or more in annual revenue and 1,035 respondents at organizations under $1 billion in annual revenue.

“The share scaling agents in one or more functions increased from 27 percent to 40 percent, while adoption among smaller organizations remained essentially flat, at 22 percent.”McKinsey, The state of AI in 2026: On the road to ROI

From regular use to scaled agents

Regular use is common, but the ladder narrows at each stage. McKinsey's survey shows nearly 9 in 10 respondents regularly used AI in at least 1 business function. The share scaling AI across the enterprise reached 44%, up from 38% a year earlier, while about 2 in 10 respondents reached the agent scaling phase. That difference is why “uses AI” and “has scaled AI” should not be treated as the same thing.

McKinsey's 40% versus 22% size split measures organizations scaling agents in one or more functions. The about 2 in 10 figure measures respondents scaling agents across the organization.

The survey also shows wider use inside organizations. The share of respondents using AI in at least 3 business functions rose from 51% to 56%.

Large versus smaller organizations

The two surveys use different size ladders and measure different stages. McKinsey reports how far respondents' organizations have scaled AI. The Census Bureau reports whether US businesses use AI in their operations.

AI adoption and scaling by organization size, with each source kept separate.

Source: US Census Bureau. Across all US businesses, AI use hovered between 17% and 20% from December 2025 to May 2026. Use did not change significantly among firms with fewer than 20 employees, while it rose among firms with at least 20 employees.

The McKinsey rows show acceleration rather than a one-year snapshot. Larger organizations moved quickly from experimentation toward scaling, while the smaller-organization share stayed essentially flat on agent scaling.

Those figures describe respondents' organizations, not a market share of all businesses. McKinsey's full survey included 1,719 participants in 97 nations and was weighted by each respondent's nation's contribution to global GDP.

Our example (illustrative): A 3-question self-benchmark. A made-up firm has 150 employees and annual revenue under $1 billion. Its leaders answer these questions:

• Do we use AI regularly in at least 1 function? Yes. McKinsey's benchmark is nearly 9 in 10 respondents. The firm's Census size row is also 32% of businesses with 100 to 249 employees using AI.
• Are we scaling AI across the enterprise? No. The McKinsey benchmark for organizations under $1 billion in annual revenue is one-third.
• Are we scaling agents in any function? No. The McKinsey benchmark for smaller organizations is 22%.

Our suggestion: The first yes means the firm has crossed the adoption threshold. The 2 no answers show where to investigate one repeatable workflow before claiming enterprise or agent scale.

Why large firms are pulling ahead (our analysis)

Possible reason 1, a wider enterprise base. Our reading: The 54% enterprise-scaling share at larger organizations, against one-third at smaller organizations, gives large firms more company-wide workflows in which an agent could operate.

Possible reason 2, in-house build capacity. Our reading: The coding-agent scale at larger enterprises may signal more capacity to build the integrations and controls that agent workflows need.

Possible reason 3, building instead of buying. Our reading: The build-instead-of-buy response may signal that procurement and deployment are moving closer together.

Our suggestion: A smaller firm should not chase agent scale before regular AI use is in place in its own functions and one business result is clear.

Coding agents are changing the build-or-buy decision

McKinsey's survey puts the share scaling software coding agents at about 2 in 10 respondents. Among larger enterprises, the share was 31%.

Nearly one-third of respondents, 32%, said their organizations had decided against purchasing at least 1 software product or feature because they could build the functionality in-house with agentic coding tools.

That does not prove that internal builds are cheaper or better. It does show that agentic coding tools have entered a real procurement decision, not just an employee productivity test.

What the Census count of all US firms shows

The US Census Bureau offers a different view. Its Business Trends and Outlook Survey tracks businesses across the economy, including very small firms. It should be read beside McKinsey's respondent survey, not merged into it.

“For example, 37% of firms with at least 250 employees reported using AI in their business operations.”US Census Bureau, AI Use at U.S. Businesses by Size and Sector, May 26, 2026

The two sources answer different questions. McKinsey shows how far surveyed organizations have moved from regular use into enterprise and agent scaling. The Census Bureau shows the share of US businesses reporting AI use in their operations.

Get your free AI visibility check

In McKinsey's survey, 47% of respondents said their organizations were scaling chatbots across the enterprise.

In our view, as rivals scale chatbots and agents, more buyers meet brands through AI answers.

Xtrusio's free check shows when ChatGPT and Gemini mention you, which competitors appear instead and what to work on next.

Start with Xtrusio's free visibility check.

Sources

1. McKinsey & Company, The state of AI in 2026: On the road to ROI, August 2026; read November 6, 2026.
2. US Census Bureau, AI Use at U.S. Businesses by Size and Sector, May 26, 2026; read November 6, 2026.