Nagesh Nama, CEO, xLM. Nagesh Nama is a seasoned technology executive with over 30 years of experience in life sciences.
Most life sciences companies are still asking the wrong question about AI. The issue is not whether token capital will replace human capital. The real issue is whether we can build a learning loop in which human expertise improves digital systems, and digital systems strengthen human judgment, without stripping the organization of the very capabilities that made it competitive in the first place.
That distinction is critical. Microsoft CEO Satya Nadella’s framing of token capital versus human capital is useful because it forces leaders to focus on where long-term value actually accumulates. Token capital is the AI capability a company builds and owns: models, agents, workflow memory, traces, evaluation frameworks and internal learning loops.
Human capital is the harder asset: scientific judgment, regulatory intuition, tacit manufacturing knowledge and the accumulated experience that allows seasoned people to recognize what matters before the data fully proves it. His central argument is that the true moat is ownership of the learning loop that allows human capital and token capital to compound over time.
In life sciences, I think that argument is broadly right. But it becomes real only when we confront the sector’s constraints honestly.
In a consumer SaaS environment, continuous self-improvement sounds great. In a GxP environment, the same idea can quickly become a validation problem if model changes, workflow behavior and decision outputs are not documented and governed. Life sciences requires a regulatory-specific architecture for token capital, one built around traceable oversight from the beginning.
The Operating Model Question
I view AI in life sciences primarily as an operating model issue. Right now, too many organizations are mistaking AI procurement for AI strategy. They buy access to copilots, drug discovery partnerships and tools, then declare themselves transformed. At best, that is localized acceleration; at worst, it is institutional leakage. If your scientists are generating the learning signal but the accumulated intelligence is ultimately enriching an external platform, then the enterprise is financing someone else’s advantage.
This matters far more in life sciences than in most sectors because the underlying knowledge base is unusually valuable. A company’s advantage lives in decades of submission history, batch knowledge and accumulated regulatory judgment. These are expensive assets, and once they are encoded into third-party systems, the company has effectively given away hard-won differentiation at a fraction of its creation cost.
So yes, token capital and human capital can absolutely work together. But only if the organization is deliberate about what is being captured, how it is governed and which human capabilities are being preserved and developed.
That last point is where many discussions become sloppy. Not all human capital is equally important. Large enterprises are good at producing process-compliant, committee-efficient operators. In regulated industries, those traits often exist for legitimate reasons: They help maintain control, consistency and discipline. But the human capital that matters most in an AI learning loop is different. It includes the people who can translate tacit expertise into machine-usable structure, challenge weak outputs and connect scientific and operational realities into a coherent decision model.
In life sciences, those capabilities are exceptionally scarce. Regulatory judgment cannot be reduced to a static rules library because regulators do not behave like deterministic software. Manufacturing tacit knowledge cannot be captured fully in SOPs because real process behavior often lives in pattern memory built through years of batch history, deviations and corrective action. These forms of expertise should be amplified through governed AI systems that preserve institutional memory and make expert reasoning more scalable.
A strong learning loop therefore has to be reciprocal. Human experts should define the evaluation logic and adjudicate edge cases. Token capital should reduce repetitive workload and preserve high-value reasoning trails, as well as surface patterns at a scale no human team can match and create continuity that survives turnover. When those two sides are connected properly, the enterprise urns individual expertise into institutional judgment that compounds.
The Apprenticeship Problem
There is a harder problem that most executives are not confronting aggressively enough: If AI takes over too much junior-level work, who becomes the next generation of senior experts?
Life sciences already faces a knowledge-transfer problem as experienced professionals retire faster than replacements can be trained. If entry-level analytical work, deviation support, submission drafting and record review become increasingly automated, then the industry risks removing the very apprenticeship layers through which judgment is formed. That is a structural threat to the future stock of human capital.
For that reason, the learning loop in life sciences must be designed not only for performance, but also for capability formation. Junior professionals still need exposure to the reasoning behind conclusions. If AI eliminates that developmental pathway, the organization may gain near-term throughput while hollowing out long-term expertise. That would be a catastrophic trade.
The companies that will win in life sciences are the ones that build the most effective governed loop between owned digital intelligence and irreplaceable human judgment. They will treat token capital as institutional memory, pattern recognition, workflow intelligence and scalable decision support. They will treat human capital as interpretation, challenge, discretion and developmental continuity. Most important, they will design the architecture so that the machine learns from the human without commoditizing the human out of existence.
Token capital and human capital can work together in a learning loop. But they do so only when leaders start treating AI as an organizational design problem under regulatory constraints. In life sciences, the real moat is not the model alone, and it is not expertise alone. It is the compliant structure capable of compounding both.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
