Enterprise AI spending split

Companies are putting almost all of their enterprise artificial intelligence budgets into tools, not into the workers expected to use them: 93% for technology and 7% for people, according to Deloitte Chief Technology Officer Bill Briggs. (youtube.com) Briggs said the 93/7 split covers technology and tooling on one side and culture, change management, and learning on the other. Fortune reported the figure on December 15, 2025, after interviewing Briggs about enterprise artificial intelligence rollouts. (iheart.com) (fortune.com) The imbalance is showing up as companies move from pilots to production. Bain said 74% of executives ranked artificial intelligence as a top-three strategic priority by the third quarter of 2025, up from 60% a year earlier, but only 23% could tie initiatives to new revenue or lower costs. (bain.com) Artificial intelligence spending is also getting folded into ordinary corporate budgets instead of one-off experiments. Andreessen Horowitz said enterprises it surveyed spent an average of $7 million on foundation model application programming interfaces, self-hosting, and fine-tuning in 2023, and nearly all planned to raise that spending by two to five times in 2024. (a16z.com) At the same time, firms are reporting that the human side is still thin. Bain said 44% of executives reported that a lack of in-house expertise was slowing adoption, and another Bain survey said data security, talent shortages, and output quality were the top concerns holding deployments back. (bain.com 1) (bain.com 2) That gap is not just about classroom training. Bain said companies scaling artificial intelligence need business users to understand and adopt the systems, and one conference speaker told the firm some companies may need to spend twice as much on training and education as on the technology itself. (bain.com) Some surveys now show what heavier people investment looks like. Bain said its “artificial intelligence front-runners” were companies that had integrated generative artificial intelligence into core processes, improved efficiency by at least 10%, and made human resources very involved in adoption. (bain.com) Deloitte has been making a parallel point in its own public materials. In a January 15, 2025 video, the firm said early adopters were spending not just on generative artificial intelligence tools but also on cybersecurity, data management, and cloud capacity needed to make those tools usable inside large companies. (youtube.com) The spending split does not mean companies should buy less software. It means the bill for enterprise artificial intelligence is larger than model access, because the work also includes redesigning processes, setting rules, and teaching employees when to trust the system and when to override it. (bain.com 1) (bain.com 2) The next test is whether 2026 budgets move money from licenses and infrastructure into learning and workflow change. If they do not, companies may keep buying more artificial intelligence and still struggle to get more work done with it. (iheart.com) (bain.com)