Dive Brief:
- As many as 80% of enterprise tech leaders attribute AI project failures to a lack of visibility or oversight, a report from Solvd found this week. The data comes from a survey of 500 U.S. CIOs and CTOs at large enterprise companies conducted by the AI advisory firm in December.
- More than half of companies say it’s somewhat likely they’ll shut down a pilot for poor performance this year. As adoption unfolds, AI budgets are under scrutiny, with more than four in five CIOs and CTOs saying their board is questioning the amount they are spending on AI.
- Some enterprises are taking a portfolio approach, piloting several AI experiments at once with the assumption that some will fail, Solvd CEO Mike Hulbert told CIO Dive.
Dive Insight:
Enterprise AI spending is growing quickly, but a lack of infrastructure that supports projects, like frameworks and governance, is keeping the technology from meeting expectations or proving ROI in 2026.
Solvd found most of these failures were related to a lack of visibility with projects, coordination of experiments and poor management, rather than the technology itself. Resolving these issues for teams usually requires establishing clear ownership over projects and transparency across an organization.
Although 71% of leaders reported they plan to somewhat increase their investment into AI initiatives in 2026, AI projects will also be more heavily scrutinized this year than in the past. Tech leaders reported they’re receiving more pressure to support AI projects whose results are backed by data. Most leaders expect only AI projects that pay back will survive in 2026.
Many companies are still in the early days of AI experimentation, Hulbert said, but the industry has moved to a more mature phase than the repeated failures coming out of the hype phase of 2022. Hulbert cautioned it can take multiple attempts to form successful pilots, as CIOs and CTOs may need to assess if they’ve gone too narrow or too broad in their approach.
“While it’s painful in some cases, the right approach is to just throw that away and start over with all the great learning that the team has, but with a more robust architecture and a broader scope of use cases,” Hulbert said.
Starting over might feel like a gut punch to tech leaders who need to report back on projects, but it’s better than continuing to spend on projects that aren’t delivering, something that 52% of tech leaders reported happening at their organizations.
This approach runs counter to how technology has historically operated, Hulbert said.
“There’s always been this idea that we build a foundation, and we make these fundamental investments that then we build on top of over time,” he said. “And that’s not what we’re aspiring to do with AI. The underlying technologies are moving so quickly that there are times that the right answer is just to throw it away.”
