3 smart ways business leaders can build successful AI strategies - before it's too late

Serg Myshkovsky/Photodisc via Getty Follow ZDNET: Add us as a preferred source on Google. ZDNET's key takeaways Business leaders should create a platform to test AI concepts. Encourage employees to take risks with AI, but proceed with care. Keep one eye on the market for new technologies that might be exploited. Making the most of AI is tough. MIT recently revealed that 95% of enterprises attempting to harness generative AI aren't seeing measurable results in revenue or growth. However, with agentic AI and deep research adding new layers of complexity, the board's demand for the successful exploitation of emerging technology has never been greater. So, what should business leaders do? Also: How AI-enabled autonomous business will change the way you work forever For Kirsty Roth, chief operations and technology officer at business information services specialist Thomson Reuters, the answer is simple: focus on strategy. Her firm's recently released Future of Professionals Survey, which polled 2,275 professionals and C-level executives from over 50 countries, found that firms with a formal AI strategy are twice as likely to experience revenue growth. Those firms are also 81% more likely to experience the benefits of AI. Also: Gartner says add AI agents ASAP - or else. Oh, and they're also overhyped However, the research also found that only 22% of organizations have such a strategy in place. For Roth, this oversight is a big mistake. "Your AI strategy must be clear," she told ZDNET. "From the top of the company, being clear that AI is an important thing you need to be doing, and why, I think, is very helpful. Ensure you have a strategy and it's well-defined for people what that means and what the opportunity is." Here are three ways business leaders can build a successful AI strategy. 1. Create an AI platform to test new ideas Roth explained how Thomson Reuters has created Open Arena, an internal AI platform that allows staff to access major large language models (LLMs) and internal data securely. "We have basically every single one of the world-leading LLMs in there," she said, recognizing that it's unusual for a blue-chip enterprise to provide access to so many models. Roth: "Your AI strategy must be clear." Thomson Reuters Roth and her team try to access as many LLMs as possible to boost software engineering processes. "If I did this role in a corporation that didn't develop software, that approach probably wouldn't be worth doing. You might pick one or two models," she said. "However, given that we need to know what the best LLMs are all the time, because we're putting them in our products, then we have them all in the AI platform." Also: Gen AI disillusionment looms, according to Gartner's 2025 Hype Cycle report Roth said her company's strong partnerships with major LLM providers mean her team gets to test new technologies early, which is helpful. The company also makes acquisitions that can help push its AI platform in new directions. As a result, the AI platform goes beyond Open Arena to internal foundation models and deep research explorations. "We acquired Safe Sign Technologies last year out of Cambridge, UK. They are helping us develop our own foundational models in the legal space right now," she said. "So, we've started to develop our own models, as well as what we've been doing for a while, which is RAG-type solutions and making the best of the public models." 2. Agree on your desired destination Thomson Reuters needed a strong objective for its explorations into AI -- and Roth said that's where use cases came in. "We said, 'Where do we think, given what we know about the tools, we could go and make a difference with AI?' So, whether that was improving sales, improving the way we develop our content, or boosting the processes in the call center." In terms of project leadership, Roth called on a CIO who is focused on internal processes. A CTO, who is the company's head of engineering, considered external, customer-focused products. Her team tried and refined about 200 use cases: "They said, 'These are the things that we think AI should work with.' We then said to the business, 'We can buy you technology that will be helpful. Or we can build something.' And we've put about 70 use cases live." Also: 71% of Americans fear that AI will put 'too many people out of work permanently' Roth described this approach to AI project development as experiment-based rather than theoretical. "We'll try an idea in one team. If it works, great. Then we typically take all the same teams that do the same thing and roll the implementation out widely. And if it doesn't, we won't.'" She continues to encourage her team of hundreds of engineers and the professionals who use the tools her organization implements to embrace risk, but carefully. "If you've built an AI-enabled tool for someone in sales that's going to tell them, 'This is what we recommend is the next thing you should go and do,' they still have a responsibility to use their brain and check the outputs," she said. "I think embracing risk in AI is about getting people comfortable with that human-in-the-loop approach. Understand AI, make it as helpful as you can, but give yourself a sense check. People still have an accountability to go and check the answers are correct." 3. Reimagine processes using fresh innovations Roth said the final phase of Thomson Reuters' AI strategy is about reimagining processes. "So, instead of trying to chop something up and make it better piece by piece, given the tooling today, how could you completely reimagine the way that you do something internally? And I think that's going to be a huge focus for us in the next two years," she said. The pace of change with AI is so fast, said Roth, that business leaders must keep one eye on the market and the other on how new technologies might be exploited. "If you'd asked me last year if the first half of this year was going to be all about deep research and agentic, I'm not sure I would have predicted that," she said. "There's new stuff being shown to us every day." Also: Stop using AI for these 9 work tasks - here's why Thomson Reuters is already exploring agentic AI. Roth said the aim is that these emerging technologies will work autonomously, helping professionals in her company and at client organizations to work quickly and effectively. "As we roll forward, the agents are going to make the process of working out the context of what you're trying to do much easier," she said. "What's clear is there will be a big push for agentic solutions coming out. And in the spaces that we've looked at, I think the agentic solutions are helpful and very powerful." The company is also exploring other areas of AI. It recently released CoCounsel Legal, which includes deep research and agentic-guided workflows. The technology, which is grounded in Westlaw, the company's legal research tool and proprietary database for lawyers and legal professionals, creates citation-backed reports. Also: 5 ways automation can speed up your daily workflow - and implementation is easy Roth said the move toward deep research is representative of how professionals will start to use AI like a trusted teammate. "We're all learning how to be good prompt engineers now. By using deep research, you'll get a good summary, with all the depth, and far more insight than you would if you asked a model question by question," she said. "So, deep research is a bit like packaging up your work in one go -- that's a good way of thinking about it. And if you think about lawyers, they may want a timeline of what happened in a case, and all the details summarized, or all the precedents together." Want to follow my work? Add ZDNET as a trusted source on Google.