|
The trillion-dollar question of the AI boom is whether businesses’ spending on the technology pays off. But measuring that return is often uncertain, confusing or incomplete.
That’s because it takes time for AI to show a return on investment that can be demonstrated in a company’s income statement, as we’ve written.
A new study from the University of Pennsylvania’s Wharton School and GBK Collective, a consultancy, echoes this idea, showing that most companies are still tracking ROI by measuring whether it makes their workers more efficient rather than whether it is driving more revenue. The study, based on online surveys of 1,400 U.S.-based employees from firms that have at least 1,000 staff and generate more than $50 million annually, showed that 40% of managers measured AI’s impact on their business in terms of “hours saved per employee” and 37% tracked how AI reduced their workers’ errors.
Only about 27% of managers said they track the revenue they generate from AI features, apps and tools they provide to their customers.
Perhaps worryingly, 44% of managers said that over the next one to two years, AI’s primary role will be to improve productivity and just 8% said it would be to drive new revenue or create new products.
Of course, improving productivity isn’t something to sneeze at, especially when it means that companies are using AI to replace labor for things like external contractors, customer service or responding to IT support tickets.
Still, Jeremy Korst, one of the Wharton study’s authors, said it was “very difficult” to find “actual, materially measurable financial impacts” from AI spending.
Today, measuring returns is more about vibes and managers’ “perception” of a return, rather than hard numbers, he said. AI is still in the early rollout phase, he said, so this level of return is “normal” or in line with previous technology adoption cycles.
Some 80% of managers in the study reported positive returns from using AI, and many of them work in IT, finance or HR.
But the 28% of managers who went a step further to say their organization was showing “strongly positive” returns were also more likely to work at companies that measure AI returns based on “growth outcomes” such as new revenues generated, not just cost savings from worker productivity, Korst said. (In comparison, 35% of The Information subscribers—which tend to be from technology businesses and investment firms—said in a recent survey that they were getting a “multiple” return on their AI spending, though only 52% of respondents said they were getting a positive return overall.)
In our own reporting, examples of non-tech firms getting revenue gains from AI seem to be few and far between. Novo Nordisk previously told us it built AI agents to audit and coordinate clinical trials, with expected additional revenue in the tens or hundreds of millions of dollars a year. And Walmart CEO David Guggina said during the Goldman Sachs technology conference in September that customers using its AI assistant Sparky spent 40% more on their orders on average. Most other examples cited in this column over the years showcase labor and time savings.
The Wharton study suggests that corporate spending on AI will keep rising quickly only if companies show a strong return on that spending, linking it to growth outcomes such as new revenue from AI and higher output, such as how much equipment a manufacturer produces, rather than cost-cutting.
For now, though, spending is rising at a decent clip. About 45% of managers said their companies are set to spend $10 million or more on AI this year, up from 39% in 2025.
But substantial increases in AI budgets may not be as plentiful in 2027, the study suggested. Just 10% of managers in the survey said they expect their company’s AI spending to rise “substantially” in the coming 12 months, compared to 19% of managers who said so in 2025.
One bright spot: Companies may be getting better at tracking their AI spending. Some 40% of managers in the survey said they were “very” confident in their company’s ability to track such spending, and another 39% said they were “fairly confident.” Just 17% were “somewhat confident.”
Workday Joins Salesforce in Providing ‘Headless’ Option
Workday on Tuesday revealed new tools that let customers use AI agents made by firms such as Anthropic, Microsoft and OpenAI to automate tasks in their Workday accounts more effectively.
The tools are similar to Salesforce’s SIforce, which last month was known as AIforce and, before then, Headless 360.
As I reported in detail last week, some employees and investors worry that Workday and Salesforce’s new tools are effectively a capitulation to the new crop of AI providers, and that the software firms will become “dumb” infrastructure, or glorified databases. But Workday and Salesforce leaders say they’ll make plenty of money this way.
“I would basically say we‘re giving customers choice,” CEO Aneel Bhusri said in a press briefing ahead of the company’s customer conference in Las Vegas. “We do have a user experience that is direct through Workday, through Sana [Workday’s AI platform] that is absolutely world class. But if a customer wants to choose headless”—meaning, using another company’s AI to access Workday—“that’s fine,” he said.
“I'd always prefer if they use our user interface, but headless will be equally good for Workday.”
Workday will charge customers based on how often they use AI agents to tap connectors and perform actions in Workday starting in February of 2027, at the end of the company’s “grace period” for customers to use such tools without extra fees.
|
