The 5 Types of AI Investment–and How to Capture Their Value

Corporate leaders are starting to worry about the returns—or lack thereof—on their recent AI investments. McKinsey’s 2025 Global Survey found that 88% of organizations use AI in at least one business function, but only 39% report any impact on EBIT, and even among those, the impact is typically less than 5%. BCG’s analysis reveals that 60% of companies investing in AI generate no material value, and only 5% create substantial value at scale. Deloitte’s survey of nearly 2,000 executives finds that satisfactory ROI on a typical AI use case takes two to four years, which is much longer than the seven-to-twelve-month payback typically expected for technology investments.

Firm-level research adds to the puzzle. A study of 1,950 American firms found that a 10% increase in AI investment correlates with only a 0.04% increase in firm growth. In the typical J-curve pattern of organizational adjustment, short-term productivity often dips before recovering. And firms that invest seriously in AI restructure their workforces in ways that go far beyond automation: they flatten hierarchies, shift toward higher-skilled labor, and fundamentally reorganize how decisions get made.

These findings are routinely interpreted as evidence that AI investment is failing. I believe they are evidence of something different: we are treating AI as a commodity. But AI’s most valuable effects are not commodity-like at all. They are inherently local; embedded in specific companies’ workflows, shaped by proprietary data, and inseparable from institutional context. I argue that AI investments come in five different flavors, and only two can be seen as commodities. The other three create durable competitive advantage. And none should be weighed against standard ROI metrics.

The Five Types

The first two types are tactical investments that sustain your position:

Type 1: Competitive Parity

The most common reason companies invest in AI is the simplest: competitors are. When other organizations in your industry are deploying AI-powered customer service, automated underwriting, or predictive analytics, matching those capabilities is not a strategic choice. It is a survival tactic.

Consider the banking industry. Bank of America’s virtual assistant Erica has surpassed three billion customer interactions, now averaging more than 58 million conversations per month. Nearly 50 million clients have used the platform, which resolves 98% of inquiries without human intervention. That is impressive, but it is not a source of competitive advantage, because JPMorgan Chase, Wells Fargo, Citigroup, and Capital One have all introduced similar agents. The technology differs at each institution, but the competitive effect is identical: no bank gains an edge, and any bank that fails to deploy falls behind.

• The financial logic: Competitive parity is a cost-avoidance investment not a return-generating one. “What is the ROI?” is the wrong question. The right question is “What is the cost of not doing this?” Measure it by benchmarking against peer capabilities and quantifying the competitive gap cost (customer churn, market share erosion, or talent flight, for example) that would result from falling behind.
• The financial strategy: Limit your investment to the industry median. Any dollar spent above parity in this category is wasted unless it migrates the initiative toward one of the strategic types.

Type 2: Option Value

The second type of AI investment is more subtle. It recognizes that AI is not a plug-and-play technology and calls for new capabilities that can leverage the potential of the AI system. With this perspective, leaders recognize that every dollar spent on AI today builds institutional knowledge that better positions the organization. This is option value: the spending might not pay off immediately, but it opens the door to future AI-based opportunities that would otherwise be inaccessible.

Moderna illustrates this vividly. In 2023, the company deployed to its 3,000-person workforce an internal platform called mChat (built off ChatGPT Enterprise). The initiative allowed Moderna’s employees to work with and develop trust in the AI system, and 80% of the employees began to use it. By 2025, Moderna had built more than 750 custom GPTs spanning clinical data analysis, contract review, regulatory submissions, and mRNA sequence design. No single tool produced a breakthrough. But collectively, they built institutional fluency with AI across every function. CEO Stéphane Bancel stated that Moderna now aims to bring 15 new products to market in five years with 6,000 staff, a goal that would traditionally require 100,000 employees. The return is not in any individual GPT. It is in the absorptive capacity—the ability to recognize, assimilate, and exploit new knowledge—that the organization builds through sustained experimentation.

• The financial logic: The value of the AI investment should be determined using real-options thinking, not traditional ROI. The relevant metrics are absorptive capacity indicators: How fast can the organization adopt new AI capabilities when they emerge? What is the pilot-to-production conversion rate? How many functions have built working AI fluency?
• The financial strategy: Allocate a fixed percentage of revenue as a learning budget, similar to R&D spending, and measure adoption velocity rather than immediate payoff. The value of the option lies in the future capabilities it makes accessible, and the cost of the option is the investment in the experiments.

The next three types are strategic investments that build durable advantage:

Type 3: Unique Integration

Integrating AI into the distinctive workflows, customer relationships, and institutional processes that define a particular company is the first step in ensuring that it creates strategic long-term competitive advantage.

Amazon’s supply chain is a powerful example. The company has deployed over a million robots across more than 300 fulfillment centers worldwide, coordinated by AI systems that forecast demand for hundreds of millions of products daily. Its new foundational AI forecasting model factors in regional demand patterns—such as ski goggles in Boulder, Colorado, during peak snow season—and has improved long-term national forecasts by 10% and regional forecasts for popular items by 20%. In 2024, Amazon delivered more than 9 billion same-day or next-day packages around the world. That advantage was created not by AI alone but by Amazon’s integration of it into decades of supplier relationships, logistics infrastructure, and operational culture. No competitor—not even ones with access to the same AI—could replicate that system.

• The financial logic: Unique integration is measured at the process level, not the enterprise level—which is why it is invisible to conventional ROI analysis. Identify the specific workflows where AI has been embedded in your distinctive capabilities and measure the performance delta: cycle time reduction, defect rate improvement, and customer retention on those specific processes versus pre-integration baselines. Amazon doesn’t measure “AI ROI.” It measures fulfillment speed, inventory carrying cost, and customer reorder rates—metrics that happen to be driven by AI integration.
• The financial strategy: Invest where AI deepens your existing competitive moats. Measure value by how much you increase moat strength not by efficiency improvements.

Type 4: Data Flywheels and Lock-In Ecosystems

When a company deploys AI in real operational contexts, it generates proprietary data that feeds back to improve the performance of the AI system. This, in turn, generates better data, which improves performance further. When this flywheel creates switching costs for customers or ecosystem partners, it produces durable competitive advantage.

John Deere’s precision agriculture platform is the most compelling current case study. The company’s See & Spray technology uses 36 cameras and machine learning to identify individual weeds in real time, reducing the use of herbicide by up to 67%. But the flywheel is what happens next: every spraying session generates millions of new data points that feed back into the platform. Each season, the system gets smarter about that farmer’s specific fields, microclimate, and weed populations. New model year 2025 tractors ship “Autonomy Ready,” and retrofit kits turn existing machines into new data-generating nodes that feed Deere’s Operations Center, the cloud hub that becomes indispensable to the farmer’s daily decisions, which makes switching to a competitor extraordinarily costly.

• The financial logic: Flywheels are valued by their compounding rate not their current output. The right metrics are flywheel velocity (how fast is the AI improving per cycle of operational data?), customer switching costs (how much would a customer lose by leaving?), and lifetime value growth (is LTV increasing as the flywheel turns?).
• The financial strategy: Invest in closed-loop systems where proprietary operational data feeds back into the AI. Measure by compounding rate and lock-in depth.

Type 5: Organizational Capability Building

This is the most important type of AI investment—and the most overlooked. If Type 2 is absorptive capacity for AI tools, Type 5 is absorptive capacity for organizational transformation—the ability to change not just what you use but what you are.

Walmart shows how this can work at massive scale. The company has not merely deployed AI tools; it has restructured how 2.1 million associates work, learn, and make decisions. In 2025, Walmart equipped 1.5 million store associates with AI tools through its proprietary Element platform, while simultaneously reskilling 50,000 frontline employees into entirely new roles and jobs—drone technicians, robot supervisors, AI agent developers—that did not exist two years ago. CEO Doug McMillon framed the ambition in a conversation with HBR: “You have to set yourself up to change all the time, not just once.” The company built a unified AI architecture organizing all agents into four “super agents” designed to prevent fragmentation, hired a dedicated AI transformation leader whose role explicitly combines technology and change management, and established Walmart Academies to continue building AI fluency at every level.

Cross-functional collaboration, an experimental culture, and the capacity for continuous role reinvention become organizational muscles that are the rarest and most valuable assets a company can possess. They satisfy the textbook criteria for sustainable competitive advantage. And yet they are invisible to any analysis that looks for advantage in the tech rather than in what the adoption process builds inside the organization.

• The financial logic: This type of AI investment should be treated as a capability premium—an option on all future organizational capabilities not just current AI. The metrics are leading indicators such as time-to-adapt when new situations emerge, velocity of both decision-making and decision-change, cross-functional collaboration indices, and new role creation rate. This is why the persistent finding that 70% of enterprise AI value comes from people, process, and culture should not be read as a critique of AI. It is a map of where the real returns are hiding.
• The financial strategy: Beyond AI tools, invest in transformation infrastructure: reskilling programs, cross-functional team architectures, and change management capabilities. Build strategic agility, the ability to make quick and wise decisions. The company that builds these muscles is investing in the capacity to exploit whatever comes after AI.

The AI Investment Diagnostic

The five-types-framework provides a practical diagnostic for senior leaders. Map every significant AI initiative onto the tactical-to-strategic spectrum, classify it by which of the five types it primarily serves, and, most critically, apply the financial logic appropriate to that type to assess the return on that investment.

In my work with Fortune 500 executive teams, I’ve found that at most companies, 70% or more of AI spending clusters at the tactical end, and when evaluated with standard ROI metrics, the cost-benefit analyses look disappointing. The strategic investments are meanwhile both underfunded and being measured with the wrong instruments.

Organizations that instead learn to allocate appropriately between the five different types of AI investment and correctly measure the value created by each are the ones who can not only justify the spending to their stakeholders, but also better position themselves for the future. The CEO who can say, “Our AI spending on competitive parity is capped at industry benchmarks but our investment in integration uniqueness and organizational capability is generating returns that no competitor can replicate” will be heard differently by a market starved for this kind of strategic clarity.

While many have predicted that AI will soon be a utility like electricity or cloud computing, the five-types-framework explains why this prediction is mistaken. AI’s most valuable applications are local, contextual, and deeply embedded in the specific institutional fabric of each organization. Walmart cannot replicate Amazon’s AI-driven supply chain in total. Caterpillar cannot transplant John Deere’s flywheel into its operations. Target cannot just copy Walmart’s organizational transformation. AI technology may become a commodity, but the integration, data ecosystems, and capabilities it builds never will.