 The Cost Gradient of the Build: How Differential Commoditization Reshapes Entrepreneurship and Valuation - A Layer-Decomposed Risk Premium for the Post-AI FirmArthur de Miranda Neto ()
Working Papers from HAL Abstract: The diffusion of generative artificial intelligence since late 2022 has triggered a debate over whether intelligence, expertise, knowledge work, and scientific production are becoming commodities. This work argues that the debate is held at cross purposes and that answering it well requires changing the unit of analysis. Commoditization is not a property of artificial intelligence, considered as a single object; it is a property of distinct layers of the knowledge-production stack, which move at different speeds and in opposite directions. The work operates this thesis through a seven-layer decomposition that maps the direction and velocity of commoditization at each layer and develops four applications of the resulting structure. First, the canonical iterative methodologies of innovation-driven entrepreneurship — Customer Development (Blank, 2013), Lean Startup (Ries, 2011), and Business Model Generation (Osterwalder and Pigneur, 2010) — are not refuted, but the cost gradient across the stations of their iterative cycle has steepened by one to two orders of magnitude on the build station, a new pre-build station of automated investigation has emerged, and the Minimum Viable Hypothesis — the terminology is not new — is proposed as a positive substitute for the Minimum Viable Product. Second, a layered theory of post-AI defensibility shows that the classical sources of competitive advantage do not commoditize jointly; strategic value migrates upward through the stack, from the artifact to the proprietary learning loop to institutional embedding. Third, Damodaran's scalar key-person discount is generalized to a vector of signed, layer-specific exposures, in which the discount can carry a negative sign even when the key person remains a value-creator. Fourth, a seventh layer is offered as a tentative hypothesis: the cross-border knowledge regime, captured by a knowledge-integration coefficient K7 bounded in [0, 1], with an illustrative same-bloc collapse threshold near K7 = 0.45 and a further contraction of approximately 38 percent in the inversion premium when frontier-model dependencies cross blocs. These thresholds are illustrative explorations of the hypothesis, not forecasts. The valuation application warrants closer treatment because its mechanism and its empirics are quantitative. In its canonical form Damodaran's discount is a single number (typically 10–25 percent) applied at appraiser discretion; Damodaran's own treatment already allows the underlying key-person value to be negative when the individual has become a value-destroyer (Damodaran, 2023). The present work isolates a distinct mechanism that no scalar can represent: the discount can carry a negative sign even when the key person remains a value-creator, once the technical labor surrounding that person becomes substitutable by frontier-model services and is therefore re-priced as a commodity input rather than a defensible asset. The scalar is replaced by a vector of signed, layer-specific exposures — commoditizing layers add risk, anti-commoditizing layers subtract it — of which the classical positive discount and the negative-sign case are both special collapses. The generalization is jurisdictionally structured: high-wage jurisdictions such as the United States produce the largest absolute substitution premium — despite lower statutory labor-cost multipliers — because the absolute base salary, not the multiplier, dominates the dollar value of the substitution. Because of the canonical scalar averages over this layered structure, a substantial fraction of early-stage valuations across jurisdictions may be systematically misspecified, a concern sharpened by venture markets that, in 2026, pay record premiums for technical teams and founder pedigree. The substitution argument is given temporal form: a migration analysis introduces the AI-orchestrator function as a permanent labor input and identifies a team-size threshold below which migration to AI-augmented operations is net-negative regardless of jurisdiction — a structural barrier for small regulated firms that the steady-state arithmetic conceals. The framework is then applied to science itself, treated as a case study in which the codified-output layer is increasingly performed by AI systems while the institutional-trust layer becomes the scarce, anti-commoditizing input. The argument is supported by secondary empirical evidence, including large-scale field experiments (Dell'Acqua et al., 2023, 2025; Brynjolfsson, Li and Raymond, 2025), payroll-level employment data (Brynjolfsson, Chandar and Chen, 2025), MBA placement statistics (Bloomberg, 2025), and bibliometric and policy analyses on the cross-border knowledge regime (Wagner and Cai, 2022; Okamura, 2025; Zhang et al., 2026; UNESCO, 2026; ASPI, 2025; Quincy Institute, 2025). It is accompanied by a simulation environment (de Miranda Neto, 2026) that materializes the framework, implements the jurisdictional analysis and the sensitivity to K7, and permits replication and extension. Eight appendices (A–H) extend the analysis: from the layered DCF and two contrasting hypothetical case companies (A), through the phase-conditional reformulation of CAPM, WACC, EVA, ROI and the Gordon perpetuity under the post-AI double-valley dynamic — a second valley of disillusionment, extending the Gartner Hype Cycle, that a firm faces around months 24–36 as competitors close the technical gap by integrating frontier-model capabilities, and which the classical formulas, assuming constant parameters, leave invisible (B) — and an eight-step operational manual (C), to consumer-price and cross-bloc fiscal implications (D), the dynamic analysis of the two case companies (E), the upstream AI value chain (F), the distributional, stewardship and epistemic-justice dimensions (G), and a conceptual glossary (H). The work is written to be read by distinct audiences — investors and acquirers, founders and operators, corporate innovation leaders, strategy and M&A advisors, research-funding agencies and project evaluators, regulators and accreditation bodies, policymakers, undergraduate and graduate students, recent graduates entering the labor market, research scientists and professors, technologists, workforce-planning professionals, and analysts in the defense sector — and a reader's guide in the Introduction indicates the entry point for each; a closing agenda identifies where economics, data science, law, ethics, political science, management, and education research can each strengthen the framework. ISBN 978-65-02-13475-7 . Keywords: accounting substitution; unit of analysis; venture valuation; post-AI defensibility; migration dynamics; AI orchestrator function; entrepreneurship; cross-border knowledge regime; jurisdiction; key-person discount; valuation; Lean Startup; Customer Development; Minimum Viable Hypothesis; defensibility; knowledge-production stack; commoditization; generative AI; generative AI commoditization knowledge-production stack defensibility Minimum Viable Hypothesis Customer Development Lean Startup valuation key-person discount jurisdiction accounting substitution cross-border knowledge regime entrepreneurship AIorchestrator function migration dynamics post-AI defensibility venture valuation unit of analysis (search for similar items in EconPapers) Downloads: (external link) Related works: Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text Persistent link: https://EconPapers.repec.org/RePEc:hal:wpaper:hal-05631380 Access Statistics for this paper More papers in Working Papers from HAL |
Is your work missing from RePEc? Here is how to contribute.
Questions or problems? Check the EconPapers FAQ or send mail to .
EconPapers is hosted by the
School of Business at Örebro University.