The hype around artificial intelligence in the enterprise space is red hot. Many companies report that they plan to deploy AI in the near term, or that they are already using the technology.
But according to new data from the U.S. Census Bureau, fewer than 10% of companies in the country were actually using AI as of late 2018. Most of the firms that reported they were taking advantage of the technology (measured via machine learning adoption) were larger firms.
The results of the survey, one of the broadest attempts thus far to measure AI adoption by enterprise, were presented in July at a virtual conference held by the National Bureau of Economic Research. Its findings are counter to some earlier surveys that estimated much higher AI adoption by enterprise — for example, McKinsey previously reported that nearly 30% of executives responding to its survey were piloting AI in their organizations.
It’s not entirely surprising to see those low numbers, said Chao He, founder of Swenson He. “Businesses are naturally resistant to change, and the amount of effort required to learn and implement new processes that utilize AI is almost inversely proportional to perceived gains over time,” he said.
U.S. organizations aren’t slow on adoption of only AI-associated technologies, according to the census survey. The survey did find that of the advanced technologies it measured, machine learning was well down on the list for adoption. However, even though companies were more than twice as likely to say they were using touchscreens, for example, as they were to say they were using machine learning, it was still just 5.9% for the former and 2.8% for the latter.
The Census Bureau asked about the use of other AI-associated technologies, including autonomous vehicles, machine vision and natural language, but fewer than 2% of companies reported the use or testing of these tools.
The results differed based on the size of the company. Nearly a quarter of firms with at least 250 employees reported they were using AI-associated technologies, compared to 7.7% of those with 10 or fewer employees.
“While large companies have the budgets and resources to hire experts and build out custom AI technology, small and medium businesses often feel left behind on adopting the latest tech and have concerns about the return on investment,” said Ryan Lester, senior director of Customer Experience Technologies at LogMeIn.
Smaller businesses should focus on single-use cases for artificial intelligence, Lester said, and look for existing technology that addresses that case specifically.
“Once more businesses start to understand that they can deploy AI in a more incremental fashion, without having to do a major technology overhaul or costly investment, we’ll start to see adoption increase in the short term,” he said.
Looking to the Future
In the coming years, the numbers from the Census Bureau survey may go up thanks to the push on automation provided by COVID-19 and the increasing advancement of artificial intelligence technology. As we get further from AI breakthroughs, it becomes more likely that products reflecting those advances will exist, said Jay Srinivasan, CEO and co-founder of atSpoke.
“The major breakthroughs in AI, especially with models for text — which are essential for enterprise — were only a few years ago,” Srinivasan said. “And since those breakthroughs, we’ve gotten a lot better at turning those breakthroughs into great products.”
It’s also possible that some companies are using AI every day and just don’t realize it. Machine learning and other AI-associated technologies are, for example, part of the Microsoft and Google suites of products, are at use on LinkedIn, and are part of how we unlock and take photos from our phones.
“Ultimately, businesses should remember that adoption of AI is a marathon, not a sprint,” said Peter Wang, co-founder and CEO of Anaconda. It’s still early days, Wang said, and growth is being seen across sectors — from finance and insurance to auto manufacturing and insulation.
“In very short order,” he predicted, “everything you see, touch and buy will be improved by machine learning processes somewhere along the way.”