Companies Going ‘All In’ on AI, Appen Study Says
AI budgets are up significantly over the past year as companies compete to survive and grow market share during the global pandemic, according to Appen, which published its State of AI and Machine Learning report this week. The study also detected a correlation between AI budget size and the likelihood that AI projects will actually be deployed on the one hand, and budgets and the use of external data providers on the other.
Now in its seventh year, Appen’s State of AI seeks to generate a broad snapshot of AI investments across the United States. The company contracted with Harris Poll to investigate various aspects of AI investments and project management at 500 companies, all of which had at least 100 employees.
The growth in AI budgets was perhaps the most compelling result to come out of the study, which had a margin of error of ± 5%. According to the study, the number of companies with budgets ranging from $500k to $5 million increased by 55% compared to last year. What’s more, Appen found that only 26% of survey respondents reported having an AI budget of less than $500,000, down significantly from last year, when 40% of survey respondents indicated that level of spending.
This points to a significant uptick in spending on AI, says Appen CEO Mark Brayan, who said the results “clearly indicate that companies are going all-in on AI.”
“When we first set out to understand the state of AI and machine learning seven years ago, organizations globally were eager to start AI projects but weren’t making meaningful progress,” Brayan stated in a press release. “Today, in a time of transition during the pandemic, AI is considered core to survival, and we see this in the results of the survey – increased budgets, more technologists in charge, and increased reliance on external training data providers.”
Larger companies tended to have higher AI budgets, according to the study, which is not surprising. The use of an external provider of training data also correlated with having a higher budget, irrespective of company size. That indicates that use of external data is a mark of maturity of AI projects, Appen suggests.
“High-quality data is the foundation for success for every AI project, and no amount of tweaking of algorithms can negate the consequences of bad data,” Wilson Pang, CTO at Appen, stated in a press release. “Companies that want to deliver better results faster understand that third-party training data is a must.”
Interestingly, small and mid-size companies showed a higher rate of using third-party training data and data management providers (such as Appen) than their larger brethren, according to the study. Ninety percent of small businesses used these external data providers, while 92% of mid-size businesses did, the study shows, but only 85% of large enterprises used them.
“Companies who use external data providers are 1.5 times more likely to say their company is ahead of others in AI deployment, and 4 times less likely to say that they are lagging,” Appen states in its report. “This is likely tied with the fact that companies that use external data providers deploy substantially more projects than those that don’t, as well as to achieving meaningful ROI.”
The study also suggests that companies with higher AI budgets also tend to have a higher rate of deploying AI projects into production. That may sound obvious, but considering the significant AI project failure rate in recent years, the fact that deployments are trending with budgets is a positive sign for CFOs under pressure to deliver a return on AI investment.
However, the deployment rate is still not great. Even among companies with AI budgets that exceed $3 million per year, only 49% of AI projects made it into production. The deployment rate was almost identical (48%) for companies with AI budgets between $1.1 million and $3 million, while it dropped considerably (to 38%) for companies with AI budgets between $500,000 and $1 million. Companies that spent less than half-a-million-dollars per year on AI saw one-third of their projects go live, the study found.
Appen also noted a significant decrease in involvement from the C-suite. In 2020, C-level executives were responsible for 71% of AI projects, a figure that dropped to 39% in 2021, the study found. “AI responsibility has trended away from the C-suite and into the rest of the organization, making AI projects more operational,” the company stated in the study.
So, what are companies doing with AI in 2021? According to the Appen report, it’s a wide mix. Support for internal IT operations was the number one use case among all cohorts (small, mid-size, and large businesses), while achieving better understanding of corporate data was number two for small and mid-size businesses. Improved productivity and efficiency of business processes was the second leading use for AI amid large companies, and was number three for small and mid-size businesses. Cutting costs was the third most-popular use case for large businesses, but that only ranked eighth and ninth for small and mid-size businesses, respectively.
“Right now, we’re seeing a lot of companies dramatically increase their machine learning investments, with special focus on use cases surrounding process automation and customer experience,” the study quotes Diego Oppenheimer, the CEO of Appen-partner Algorithmia as saying. “This makes a lot of sense because these use cases offer both cost savings and top-line business growth, which are both essential investments during times of economic uncertainty.”
Finally, the report also detected some differences of opinion in regards to ethics and transparency, which is an emerging concern for AI practitioners. Specifically, the report found that business leaders and technologists don’t yet agree in areas like ethics and interpretability. It found that technologists have a greater concern for ethics (41% versus business leaders’ 33%), while business leaders are more concerned with interpretability (47% versus technologists’ 38%).
You can access the 2021 State of AI and Machine Learning Report here.