A growing number of professionals are beginning to question whether AI is delivering meaningful value to their companies, according to a report titled "2026 Future of Professionals" by Thomson Reuters. The report is based on a survey of 1,800 professionals from various industries, revealing a widening gap between the high expectations of AI and its actual implementation. Nearly 91 percent of respondents said their companies are not fully utilizing AI’s potential. According to Kirsty Roth, director of operations at Thomson Reuters, the situation has shifted significantly from 18 months ago, when employees were excited about experimenting with AI and company leaders were supportive. Today, professionals are using AI more freely, but leaders are increasingly worried about rising IT costs. Last year, a study by MIT found that 95 percent of AI projects fail to deliver tangible value, contributing to companies’ growing hesitation about AI investments. AI tools and models have become highly fragmented, with a wide range of options available to professionals. Steve Lucas, CEO of Boomi, notes that employees are now familiar with many technical terms that were not common just a few years ago, such as "leading models," "private models," and "agent frameworks." Roth suggests that companies should focus on two main areas: conducting well-founded explorations and identifying specific, practical use cases that can be implemented in real-world settings. The survey also highlights what professionals expect from AI tools. Over 96 percent believe AI should protect confidential data, 94 percent expect it to base results on authoritative content, and 90 percent want it to provide explainable and defensible reasoning. However, 41 percent of professionals using AI at work say they don’t have access to high-quality tools. Even when companies have an AI strategy, implementation often lags. Roth points to the "tool blast" — a situation where companies offer many AI services without clear goals — making it hard for employees to see the benefits unless costs rise significantly. Gartner predicts that 40 percent of companies will reduce or disable their autonomous AI agents by 2027 due to doubts about their value. Roth suggests that business leaders should allow professionals to explore emerging technologies without taking excessive risks. At Thomson Reuters, the company has taken an open approach to generative AI, letting employees in areas like marketing, sales, and development test tools they find useful. Instead of focusing on cost targets from the start, the company encourages experimentation, with employees testing AI tools and identifying better ways of working. Roth notes that some employees needed more encouragement, but this approach has worked well for the company. A key part of this strategy is evaluating which AI tools deliver real benefits and discontinuing those that do not. Thomson Reuters initially tested various tools for about six weeks, allowing employees to use AI depending on their roles. If the results were positive, the tool was rolled out to other teams. If not, the project was abandoned. Roth emphasizes the importance of starting with access to tools and testing them early. Companies that are leading in generative and agent-based AI are those that successfully convert experiments into operational services. Roth says the most successful organizations define their use cases clearly and adapt their business processes to new ways of working. At Thomson Reuters, specific AI use cases are focused on five areas: engineering, customer service and success, marketing, editorial and content operations, and basic technology operations. In these areas, AI tools are identified, tested, and deployed, with their effectiveness measured over time. Currently, 87 percent of Thomson Reuters employees use AI tools in their daily work. Roth cites the example of customer service and sales support, where staff can use the company’s internal AI platform, called Open Arena, as well as the leading model Claude. Instead of spending hours gathering information from salespeople and the Salesforce platform, staff can use approved AI services to get answers to their questions in seconds. With Claude, employees can enter a query to extract information, allowing the AI to write a summary, identify key opportunities, detect potential risks related to an account, or determine if a customer has recently contacted customer service and whether they were dissatisfied with something. This preparation for meetings is much quicker and more efficient. Thomson Reuters employees also use AI for market research, document writing, and tracking the profitability of products and services. Roth emphasizes that the main lesson learned from deploying AI was that business leaders should work to overcome the fears of professionals. Humans are naturally resistant to change, and it was essential to demystify AI and give people the opportunity to try out these tools without fear. The company is now reaching a more advanced stage of maturity, and Roth believes that the direction taken and the consistency of the approach are just as important as the technology itself.