Leveraging AI in business has never been easy. A recent study from MIT found that up to 95% of businesses implementing AI for generation purposes have not seen clear results in terms of revenue or growth.
With AI agents and 'deep research' technologies becoming increasingly sophisticated, pressure from leadership to effectively leverage AI is also growing.
So where should businesses begin?
According to Kirsty Roth, Chief Operating Officer and Technology Officer at Thomson Reuters, the answer lies in strategy.
A global survey by this company found that businesses with a clear AI strategy are twice as likely to increase revenue and have an 81% chance of leveraging the benefits of AI. However, only 22% of organizations actually have such a strategy.
This reveals a core reason: the biggest mistake is the lack of a strategy.
Building an AI platform to test ideas.
One of the key steps is to create an internal AI testing environment.
At Thomson Reuters, the Open Arena platform was built to allow employees to access various large language models (LLMs) and internal data in a secure environment.
Accessing multiple models allows the engineering team to continuously evaluate and select the best technology to incorporate into the product. For businesses that don't develop software, this approach can be simpler—just choose a few suitable models.
Additionally, businesses can expand their platforms by combining public and private models. For example, Thomson Reuters has developed its own models in the legal field while continuing to leverage solutions like RAG to utilize data more effectively.
Clearly define the objectives and "use case".
An AI strategy cannot survive without specific goals. Instead of deploying AI in a general way, businesses need to answer the question: What will AI improve? Will it increase sales, optimize internal processes, or enhance the customer experience?
At Thomson Reuters, the team tested approximately 200 different use cases, then selected and implemented about 70 in real-world scenarios. Notably, they didn't follow theory, but rather a practical testing approach: testing on a small scale, scaling up if effective, and stopping if not.
Another crucial element is 'human-in-the-loop' — humans still need to check and be accountable for the results. AI can suggest actions, but the final decision still requires human evaluation.
Redesign the process instead of just improving it.
A common mistake when deploying AI is trying to optimize only small parts of the existing process. Instead, businesses should ask: if we had AI, how could we redo this process from scratch?
This approach helps to maximize the power of new technology, rather than simply 'patching up' old systems. In the next one to two years, this is considered a crucial direction for many businesses.
Furthermore, the pace of AI development is extremely rapid. Trends like AI agents or deep research can emerge and become the standard in a short period of time. Therefore, businesses need to constantly monitor the market and be ready to adjust their strategies.
New solutions, such as CoCounsel Legal – Thomson Reuters' AI-integrated legal research tool – demonstrate that AI is gradually becoming a 'digital colleague' that can assist in analyzing, synthesizing, and providing insightful data.
Currently, AI is no longer an option, but a mandatory element in business development strategies. However, success doesn't come from hastily adopting the technology, but from building a clear strategy, conducting controlled testing, and continuously redesigning processes.
In the current climate, businesses that correctly understand and implement AI will gain a significant competitive advantage — while those that lag behind may pay the price.