Week of September 7, 2026
Most AI use is still narrow, not deep
Adoption is wide but shallow. Most AI-using businesses have it running in just a few functions and on just a few tasks.
65%
Among companies that use AI, 65% limit it to three or fewer tasks and 57% keep it inside three or fewer business functions. This means AI is mostly running as a contained pilot rather than something woven across the org.
Source: U.S. Census Bureau — analysis of ~1.2 million U.S. employer businesses
“The teams pulling ahead aren't the ones who adopted AI hardest, but the ones who wired it together, so their agents pass work to each other instead of waiting for a human to copy-paste between tabs. Broad adoption is a headcount stat. Deep adoption is an unfair advantage.”
Holly Hale, Senior Manager, SEO/AEO, ActiveCampaign
Analysis
What the numbers actually mean.
Most companies trialed AI like a tool. Few rebuilt a workflow around it.
Among businesses using AI, most keep it boxed into a few tasks and a few corners of the org. Two thirds run it on three or fewer tasks, and more than half confine it to three or fewer functions. Adoption has spread wide, but inside most companies it is still a contained pilot, not something woven through the work.
Shallow use is where the value stalls.
The gap between adopting AI and getting value from it lives right here. McKinsey's read is blunt: value comes from redesigning one workflow end to end around a few good tools, not from piling up more of them. A three-task pilot proves AI works, but rarely changes the economics, which is why teams feel helped and still can't name what moved.
Depth on one workflow beats breadth across ten.
The move from shallow to deep is about depth on one process, not breadth across many. Push AI all the way through a single workflow, from first draft to final measurement, and the returns compound as each stage feeds the next. A team that commits to one process fully will pull ahead of one spreading AI a task at a time.
Subscribe to the newsletter to be alerted when new tools drop
Related
More data from the AI Lab.
