$725 Billion Later, 91% Say Organizations Still Aren't Realizing AI's Full Value

News

1. Ford Rehires 350 Engineers After Finding AI Alone Wasn't Enough

2. $725 Billion Later, 91% Say Organizations Still Aren't Realizing AI's Full Value

3. AI Dispatch | AI's Transformation Phase

4. OPAQUE and TII Launch Open Standard for Verifiable AI Governance

5. e& UAE Launches OneWork, a Unified Communications Platform Hosted Entirely in the UAE

6. Alibaba Used 29 Million Fake Exchanges to Extract Claude's Intelligence, Claims Anthropic

Artificial intelligence has largely cleared its first hurdle of adoption. The next phase is proving far more difficult as organizations are discovering that deploying AI tools is not the same as embedding them into business operations, creating measurable value, or earning customer trust.

Two new reports released this week indicate that the AI economy is entering a new phase where success depends less on access to models and more on execution quality.

According to Thomson Reuters’ 2026 Future of Professionals report, the technology is no longer the primary bottleneck. Instead, organizations are struggling to translate AI investments into operational outcomes. Based on a global survey of 1,800 legal, tax, audit, and compliance professionals, the report found that while 74% of respondents use AI every week, 91% believe their organizations are failing to realize its full potential.

The consequences are becoming tangible. Thomson Reuters estimates that professional services firms in the United States have as much as $143 billion in client revenue at risk as customers increasingly evaluate providers on AI capabilities. Nearly one-third of corporate clients said they are likely to reassess their relationships with service providers within the next year based on how effectively AI is being deployed.

Yet organizations appear unable to keep pace with those expectations. While 78% of clients consider AI-enabled improvements in quality to be important or essential, only 6% believe that most providers are currently delivering meaningful benefits.

The disconnect is also reshaping workforce behavior. One-third of professionals admitted to using AI applications that have not been approved by their employers, a figure that rises to 41% among employees who believe their organizations are moving too slowly. Nearly one in four professionals who perceive a significant gap between AI’s capabilities and their company’s implementation are considering leaving within the next two years.

The findings suggest that AI implementation has become as much a leadership and governance challenge as a technology one. Thomson Reuters argues organizations now require what it calls “Fiduciary-Grade AI”—systems built on trusted data, transparent reasoning, human oversight, and enterprise-grade security rather than simply access to increasingly capable foundation models.

That emphasis on execution comes as another question surrounding AI is beginning to find an answer: whether the economics of the technology can justify unprecedented infrastructure spending.

A separate report by research firm Exponential View found that AI revenues have reached a tipping point where the industry’s massive investments may be becoming economically sustainable. 

The analysis suggests that revenue growth is now beginning to cover the enormous capital expenditures technology companies are making, although profit margins remain narrow. Depreciation charges still consume more than two-thirds of AI-related revenue, leaving limited room to absorb other operating costs, including power, labor, and financing.

The report addresses one of the defining uncertainties surrounding the generative AI boom. Technology giants, including Meta, Alphabet, Microsoft, and Amazon, are expected to collectively spend as much as $725 billion on capital expenditures this year, much of it directed toward AI chips, cloud infrastructure, and data centers.

At the same time, the competitive landscape is becoming more fragmented. Data from OpenRouter shows demand is shifting beyond the dominant frontier model providers. The share of inference tokens requested from Google, OpenAI, and Anthropic models fell to 33% in June 2026 from 72% a year earlier, reflecting growing adoption of open-weight models and Chinese alternatives such as DeepSeek.

Taken together, the reports point to an inflection point in the AI market. Infrastructure investment is no longer the only determinant of competitive advantage, nor is access to the largest models. As model capabilities become increasingly commoditized and customer expectations rise, organizations face the challenge of converting AI experimentation into measurable business performance.