91% of professionals say their organization still isn’t realizing AI’s potential

A global Thomson Reuters survey of 1,800 professionals across sectors reveals a widening gap between AI aspirations and on-the-ground adoption. 91% of respondents said their organization has not yet fully realized the technology, and Christie Roth, chief operating officer at the company, told ZDNet the conversation has moved from “which tools to buy” to the classic change-management stage of how to actually alter processes to generate value.

Roth says organizations are beginning to realize these technologies are expensive and don’t always deliver immediate ROI. Steve Lucas, CEO of Boomi, described the situation as “hyper-fragmented and hyper-siloed.” Professionals now must differentiate between foundation models, private models, domain-specific models, open-weight models, agency frameworks, agency lifts and agency loops—terms that did not exist a few months ago.

Survey respondents were clear about requirements: 96% demanded protection of confidential information, 94% wanted outputs anchored in authoritative content, and 90% required explanations that could be justified and defended. Yet 41% of those already using AI at work reported lacking access to high-level tools. Even when a declared strategy exists, execution falters: only 35% of professionals in organizations with a defined AI strategy said it appears in their daily work.

Roth labeled the phenomenon “tool explosion,” where organizations push a wide array of AI services to employees without a defined business result. “I’ve heard people say: ‘They gave me all these things. But what am I supposed to do with them?’” she explained. Without clarity on which tools to use, improvements are hard to see beyond rising software spend. Conversely, savvy managers allow controlled exploration of emerging technologies without taking excessive risk.

Thomson Reuters adopted an open approach: marketing, sales and development staff who wanted to try a tool received a six-week license. “We told them: ‘Here are the tools. Reimagine what you can do,’” Roth described. Some succeeded more, some needed a push, but the principle was clear—test quickly and drop what doesn’t deliver value. That, she says, enabled the organization to spot stable production use cases amid the noise.