The integration of artificial intelligence into the corporate environment is entering a new phase. As recent studies show, the decisive factor for success is not so much the power of computing resources, but rather the ability of companies to extract and formalize the tacit knowledge of their employees—those practical skills and intuitive decisions that are developed over years but are rarely documented.
This concept, tracing back to philosopher Michael Polanyi and his principle that "we know more than we can tell," is now gaining practical significance. Universal AI models indeed show impressive results in areas with large volumes of open data and clear verification criteria. However, in tasks where a company's internal context and expert judgment are critical, written instructions alone prove insufficient. The situation resembles engineer Frederick Taylor's attempts in the early 20th century to systematize the experience of factory workers—only now the tool is not a stopwatch, but neural networks.
Bridgewater: Fine-tuning Models on Expert Data
A telling example is the joint study by Bridgewater AIA Labs and Thinking Machines Lab, published on June 30. The teams tested leading AI models (including Claude Opus 4.6, 4.8, Gemini 3.1 Pro, GPT-5.4, and GPT-5.5) on six financial tasks, including analyzing news and central bank documents. Using standard instructions, model accuracy ranged from 45.6% to 50.1%. However, after applying prompts developed by Bridgewater experts, the figure soared to 74.3–78.2%.
The key breakthrough came during the fine-tuning stage: the Qwen3-235B model, trained on data labeled by Bridgewater specialists, achieved an average accuracy of 84.7%. According to developers' calculations, the number of errors decreased by 29.8% compared to the best universal model, while the cost of processing tasks dropped by a factor of 13.8. This approach, termed "differentiated intelligence," involves tailoring individual models to the specific processes and requirements of a particular organization.
Ford: Bringing Back Veterans to Train AI
Another striking example is Ford's strategy. Over the past three years, the automaker has hired, promoted, or rehired about 350 experienced technical specialists. Their task is to review designs, identify potential faults, train young engineers, and, crucially, participate in improving automated quality control tools. Ford's Vice President of Hardware Development, Charles Poon, stated directly: "Artificial intelligence is a great tool, but it is only as good as the information used to train it." The company acknowledges that it previously did not systematically preserve the knowledge of experienced engineers, and some specialists left Ford before their expertise could be fully transferred into internal processes.
My analysis: We are witnessing a fundamental shift in corporate AI strategy. Instead of chasing ever more powerful universal models, companies are beginning to understand that true value lies in their own data and expertise. Those who can effectively "digitize" the tacit knowledge of their employees will gain a sustainable competitive advantage. This is not just a technological trend—it is a rethinking of the role of human capital in the age of artificial intelligence.