AI investments fail to scale because firms treat AI as a broad technology program rather than as discrete, governable decision opportunities; a decision-centric AIPN framework ties AI spending to identifiable, measurable sources of value and prescribes staging and portfolio assembly to improve returns.

Main Finding

Treating enterprise AI as a single technology program drives the "AI-investment paradox"—heavy spending with limited, non-scaling business value. Governing AI instead as a portfolio of discrete, decision-centric investments (AI-Investable Process Nodes, or AIPNs) lets firms assess Expected Net Benefit at the decision point, stage implementations using real-options logic, and assemble investments by risk-return principles. This decision-centric portfolio framework links AI spending to identifiable, governable, and accumulative sources of business value, offering a practical path to scale and sustain AI returns.

Key Points

• AI-investment paradox: persistent high AI spend but many initiatives fail to scale or produce sustained value.
• Root cause argued: governance treats AI as a broad program rather than a set of bounded, investable decision opportunities embedded in workflows.
• AI-Investable Process Nodes (AIPNs): the unit of analysis — bounded decision points where AI can change expected outcomes and where benefits, risks, and costs can be assessed ex ante.
• Node-level value is formalized via an Expected Net Benefit metric (evaluate benefits minus costs/risks at the decision point).
• Staging: individual AIPNs can be implemented in stages using real-options logic (e.g., pilot → scale contingent on realized value), reducing downside and preserving upside.
• Portfolio assembly: AIPNs are combined via risk-return principles to balance diversification, aggregate value, and organizational constraints.
• Outcome: a governance and investment pathway that makes AI spending more transparent, accountable, and cumulative in generating business value.

Data & Methods

• Primary approach: conceptual and formal framework development rather than empirical testing.
• Key methodological elements:
  • Definition and boundary-setting for AIPNs as decision nodes within processes.
  • Formalization of node-level valuation through Expected Net Benefit (ENB) — a forward-looking metric to quantify incremental expected value of applying AI at a node.
  • Application of real-options reasoning to stage investments (sequential investment with information revelation to manage uncertainty).
  • Use of portfolio theory / risk-return tradeoffs to assemble and prioritize multiple AIPNs across the firm.
• Evidence type: theoretical modeling and framework synthesis; the paper appears to offer formal arguments and prescriptive governance constructs rather than large-scale empirical validation.

Implications for AI Economics

• Measurement and valuation: Moves evaluation from vague “AI ROI” to ex ante, decision-level ENB—enabling better cost-benefit comparisons and capital allocation.
• Investment governance: Supports structured investment processes (pilots, option-to-scale triggers) reducing wasted spend and increasing scalability of successful initiatives.
• Portfolio strategy: Enables firms to diversify AI investments across heterogeneous decision nodes, manage aggregate risk, and prioritize nodes with highest marginal value or strategic importance.
• Organizational design: Encourages embedding AI investment decisions within process owners and workflows rather than centralized “AI programs,” shifting incentives and accountability.
• Research agenda: Empirical work can test how well AIPN identification, ENB estimation, and staging predict scaling/sustained value; also invites development of methods to estimate ENB under uncertainty, and to optimize portfolios of AIPNs.
• Policy and investor perspective: Provides a clearer framework for assessing firm-level AI investments and their expected economic returns, which can inform disclosure, benchmarking, and investment decisions.
• Limitations / challenges: Requires reliable identification of AIPNs, credible ex ante estimates of benefits/risks, organizational capability to stage investments and collect outcomes, and potential frictions in reallocating capital across nodes.