Even as companies spend more on artificial intelligence each year, a majority continue to miss their related cost-savings targets, according to new research by Bain & Co.
In Bain’s recent survey of 951 global companies, among those that measured cost savings from the use of AI tools, 37% of them targeted savings of 11% to 20%, but only 29% achieved that level. And while 17% anticipated savings of 21% to 30%, only 10% reached that threshold.
Further accentuating the low-return pattern, 25% expected savings of 10% or less, but 40% actually fell into that bucket.
For the majority, while the technology worked, “the value didn’t arrive,” Bain wrote in its research report. “And rather than pause to understand why, 90% of those same companies are now increasing their budgets again.”
But this time, many companies’ spending increases will be for AI agents that, according to Bain, will operate with even greater autonomy and complexity.
While many executives foresee a future of autonomous systems handling complex decisions end-to-end, that’s far from today’s reality. Only 7% of those surveyed said they’re currently running fully autonomous agents in production.
“If your agentic AI business case was built on the economics of full automation and the reality is a system routing a significant share of decisions to a human queue, the CFO approved one set of numbers and the organization is living with another,” Bain wrote.
Failing to hit cost-savings thresholds can perpetuate more of the same, as 44% of companies said they’re planning to fund investments in agentic and generative AI with savings from prior automation programs. Such a funding model “works only if the prior returns are real,” Bain said.
It’s not a new phenomenon, though. Bain noted that sizing an investment case based on “projections rather than actuals” is a pattern the firm has documented across every recent wave of automation.
What did respondents themselves see as the major barriers to AI progress? The top response — inadequate data access and integration across systems — was cited by 41% of survey respondents.
Bain noted the irony of this result, considering that it follows a decade of investments in data modernization that have cost “well into” the hundreds of billions of dollars globally.
The research report counseled that companies should use AI to solve the data problem, rather than waiting for the data problem to be solved first.
“The fastest path to value is often automating one repeatable, high-value workflow — where humans are currently pulling data manually, consolidating spreadsheets and producing reports — and replacing that entire sequence with AI,” Bain wrote.
