How Tacit Knowledge and Token Capital Shape a Company’s Lifeline

As large models and AI agents rapidly enter business scenarios, most enterprises remain at a shallow, cost‑saving tool level, which fails to unlock technical value and risks leaking long‑term tacit knowledge and decision logic, thereby eroding core competitiveness.

Satya Nadella argues that the current AI wave is not merely an efficiency boost but a fundamental reshaping of asset structures and operating logic; competition now centers on tacit knowledge and token capital, while scaling agents introduces collaboration and risk‑control challenges. Managers must move beyond tool thinking, create private AI closed loops, safeguard knowledge assets, and adapt to agent‑centric collaboration.

"Tacit knowledge" refers to the professional skills, judgment, and processes accumulated over years and residing in employees’ minds, which are hard to document fully. AI can capture work‑behavior traces and convert this intangible knowledge into machine‑readable content.

Nadella contrasts the common view of AI as a ubiquitous utility with a banking example: in the mobile‑internet era, firms only needed to keep applications running, whereas AI now directly touches deep‑seated credit‑approval experience and risk assessment—unique tacit knowledge that, if exposed, can dismantle competitive advantage.

To prevent such leakage, companies should adopt a "re‑foundation" approach using a controllable "climbing machine"—a private, closed‑loop AI learning system that continuously trains proprietary models on internal data, turning tacit knowledge into token capital and building a distinct competitive moat.

Regarding model selection, Nadira advises against blindly chasing frontier large models. For routine, predictable tasks, lightweight, domain‑specific models combined with human‑in‑the‑loop reinforcement learning can lower compute costs and often outperform cutting‑edge models in efficiency.

Frontier models, being expensive, are better suited for breakthrough research areas such as new material development or fundamental scientific exploration.

Looking ahead, the AI supply chain will become a central concern. Enterprises must construct an end‑to‑end, self‑controlled AI supply chain that links data, models, training, and applications, enabling continuous amplification of competitive advantage.

An illustrative case shows that retail firms’ decades‑long negotiation tactics can be exposed when employees use public AI tools like ChatGPT, allowing competitors to learn the same strategies and eroding the firm’s unique edge.