Jarvis Nelson Osorio on Long-Term Return on Human Capital in the Age of AI

I am Jarvis Nelson Osorio (Jarvis Osorio)—technology entrepreneur, AI innovator, DAO institutional designer, author of The Compliant Mind and Breaking the Centralized Order, and artist. Boards love short-term automation metrics: tickets closed, headcount avoided, minutes saved. I prefer a harder question I raised on LinkedIn as LTRHC—Long-Term Return on Human Capital: How AI Systems Are Shaping the Global Workforce (post).

Automation savings are not a strategy

If AI only removes tasks, the spreadsheet looks green while the organization grows brittle. People stop practicing judgment in the zones the model swallowed. Juniors never see the hard cases. Seniors become prompt operators for systems that dislike dissent. That is a compliant workforce in the sense of my book: fast to accept machine output, thin on the muscles that catch machine error.

Long-term return on human capital asks what the system does to capability over years:

• Do people still rehearse disagreement, estimation and repair?
• Are apprenticeships redesigned, or simply deleted?
• Does the firm measure quality of judgment, or only throughput?
• When the model is wrong, who still knows how to notice?

Design AI as a trainer, not only a replacer

Product and workplace AI can be built to preserve struggle where struggle teaches:

• Scaffolding instead of finished answers for learners
• Mandatory independent review on high-impact outputs
• Rotation so humans still touch edge cases the model rarely sees
• Explicit "challenge mode" so agreement is not the only rewarded path

Those choices rhyme with agent institutions: separation of powers, audit trails, pause-and-reverse. They also rhyme with parenting practice around yes-machines. The workforce is not a separate planet from the household. It is the next chapter of the same training story.

DAO lessons for corporate AI governance

In Breaking the Centralized Order I treat decentralized organizations as institutions that need written roles, limits and change-control. Companies deploying agents need the same seriousness:

• Who may let an agent talk to customers or move money?
• How do policies change—by chat edit or by review?
• Where is the record when an automated decision harms someone?

Without those answers, "AI transformation" is just centralization with better branding.

Metrics that respect the decade

Replace vanity automation counts with a short scorecard:

• Time-to-competent for new hires on core workflows (should not collapse into "they can prompt")
• Rate of caught model errors before customer impact
• Share of high-stakes decisions with an independent human or independent-model review
• Retention of craft skills the firm still needs when the vendor API changes

If those numbers improve while raw ticket volume also improves, you are compounding. If ticket volume soars while judgment metrics fall, you are liquidating human capital and calling it efficiency.

A positive bet

I am optimistic when teams treat people as the compounding asset and models as leverage. The winners will not be the firms that removed the most keystrokes in a quarter. They will be the ones whose humans got sharper because the tools demanded verification, taught craft and left room for refusal.

Explore more: jarvisosorio.com · Telegraph hub.

Measure the decade, not the demo. That is long-term return on human capital.

Originally explored on my LinkedIn. More: https://jarvisosorio.com