AI adoption

For the 24th time in history, the U.S. is trying to count every person in the country. You probably knew that, but did you know the Census Bureau has also counted businesses that use AI?

In late 2018, the Census Bureau and National Center for Science and Engineering Statistics polled 850,000 U.S. businesses on “advanced technology” adoption and received 583,000 responses. Last month, a team of researchers published a paper dissecting the data (h/t Wired).

It’s the first 10,000-foot view of AI adoption for all shapes and sizes of U.S. businesses. Typically, adoption reports come from consulting firms interviewing the C-suite of Fortune 500 companies.

Unsurprising results first

The vast majority of businesses haven’t messed around with neural nets. Here’s the percentage breakdown of companies that reported adoption of a given tech (former = usage, latter = testing):

  • Touchscreens: 5.9% / 0.9%
  • Machine learning: 2.8% / 0.7%
  • Voice recognition: 2.5% / 0.7%
  • Computer vision: 1.7 / 0.4%
  • Robotics: 1.3 / 0.3%
  • Natural language processing: 1.2% / 0.4%
  • RFID: 1.1% / 0.3%
  • Augmented reality: 0.8% / 0.4%
  • Automated vehicles: 0.8% / 0.2%

Size and age seem to be the main determinants of tech adoption, according to the researchers. Large companies can devote more resources to R&D and tech spend. And while it’s true that tech startups have higher adoption rates, it’s also true that few small, young businesses are tech startups.

  • For reference from a similar time period, PwC found that 20% of the executives it surveyed planned to implement AI at scale in 2019.

Real adoption rates

I think they’re at least a few percentage points higher, for two reasons: 1) these numbers are a snapshot of the past, since the data is a few years old and 2) not every company responded. Those that did may have been confused about the survey's methodology.

Bottom line: U.S. businesses are still in the early innings. There’s a reason we say “emerging” technology.