AI’s Hidden ROI: Measuring Second and Third-Order Effects for Board Decisions

Boards are accustomed to evaluating investment proposals through a narrow financial lens: a focus on reduced costs, faster processes, and fewer errors. In other words, the immediate and easily quantifiable gains. Yet the organisations that extract the greatest value from artificial intelligence (AI) know this is only the beginning. The most transformative benefits often emerge much later — and in places that the original business case never anticipated.

This “hidden ROI” of AI is not a mystical bonus; it is the result of cascading effects that ripple across the enterprise once AI takes root. When AI assumes responsibility for routine or repetitive work, it doesn’t simply free up capacity. It reshapes workflows, reallocates human effort, and unlocks possibilities that were previously out of reach. These are the second and third-order effects that compound over time, but they are rarely captured in traditional ROI frameworks.

The danger for Boards is clear. If you measure AI only by its immediate gains, you are almost certain to undervalue it — sometimes by orders of magnitude. Worse, you risk abandoning promising initiatives prematurely. Gartner has already warned that more than 40% of agentic AI projects are likely to be cancelled by 2027, not because the technology fails, but because decision-makers lose patience when early returns appear modest. In an environment where competitors are already building for the long term, that is a strategic mistake.

Why the numbers look low — at first

There is plenty of evidence to suggest that early-stage ROI figures understate AI’s true potential. A March 2025 survey by BCG of 280 finance executives found that the median reported ROI from AI was just 10%, well below the 20% that many leadership teams target. Yet a small group — about one in five organisations — are already exceeding that target. These out-performers tend to have one thing in common: they understand that AI’s value creation is rarely linear and that the biggest gains often arrive in the second and third waves of impact.

PwC’s own analysis supports this view. Its 2025 AI Predictions study found that while adoption is accelerating, the benefits are not evenly distributed. Early adopters that invested heavily in AI talent, governance, and infrastructure are now seeing compounding returns, while others remain stuck in pilot purgatory. The gap between these two groups is widening, creating a two-speed economy in which those who recognise — and measure — AI’s cascading value pull decisively ahead.

The cascade effect in action

The pattern is not unique to AI. History shows the same thing happening with other general-purpose technologies. The internet’s initial commercial returns came from obvious efficiencies, but the real transformation emerged years later in the form of e-commerce, social media, and platform economies — developments few had predicted in the early 1990s.

AI exhibits the same trajectory. The first wave of benefits comes from doing familiar things more efficiently: automating document processing, improving demand forecasts, optimising scheduling. The second wave emerges as new capabilities and data streams enable innovations in products, services, and processes. The third wave — the most transformative — occurs when AI is woven into the organisation’s strategy and operating model, creating entirely new sources of revenue and competitive advantage.

McKinsey’s research into workplace AI adoption reinforces this, noting that companies which successfully scale AI across functions capture up to four times more value than those that confine it to isolated pilots. The difference is not simply in the quantity of AI deployed, but in the breadth of its integration and the organisation’s ability to act on the new possibilities it reveals.

Understanding the orders of effect

First-order effects are the easiest to measure because they are direct and immediate. An AI-powered claims processing system that cuts average handling time from ten minutes to three produces a measurable labour saving. These figures fit neatly into spreadsheets and Board packs.

Second-order effects take longer to emerge. In the claims example, faster processing may lead to higher customer satisfaction, which in turn drives repeat business and reduces churn. Staff freed from repetitive tasks may retrain into roles that add more value, building organisational capabilities in data analysis or customer engagement.

Third-order effects represent the most significant — and often the most surprising — shifts. They involve fundamental changes in business models, market positioning, or industry structure. The insurer that began with AI-assisted claims processing might evolve into a provider of real-time risk mitigation services, using live data feeds to prevent claims before they occur. This is a complete redefinition of the value proposition, and it is unlikely to have been in the original business case.

Why traditional ROI misses the mark

Several factors conspire to keep these later-stage benefits out of sight.

The first is time horizon mismatch. Most business cases and ROI analyses look no further than three years ahead, yet many second and third-order effects mature over longer cycles. Eaton Vance’s Consilient Observer notes that second-order effects, by their nature, are often invisible until they are well underway, and that investors systematically undervalue them as a result.

Attribution is another challenge. When AI in supply chain management improves product availability, and customer satisfaction rises as a consequence, which initiative gets the credit? Without integrated measurement frameworks, such effects are lost in departmental silos.

Finally, many of the most valuable outcomes are unpredictable. AI’s impressive technical capabilities frequently mask the fact that business value lags behind, emerging only after organisations adapt their processes and culture to exploit new possibilities. This unpredictability is both a challenge and an opportunity: a challenge because it complicates planning, and an opportunity because it can create entirely new markets.

Measuring what matters

Recognising the existence of hidden ROI is one thing; capturing it is another. From my work developing AI adoption mechanisms such as Well-Advised and the AI Stages of Adoption (AISA), I have seen how Boards can make these effects visible and measurable.

One effective approach is to structure measurement around leading, lagging, and predictive indicators. As I’ve previously detailed in my article on measuring AI ROI, leading indicators — such as early customer satisfaction improvements or innovation pipeline velocity — provide early signals of future value creation. Lagging indicators — like market share growth or increased profit margins — confirm the results of past actions. Predictive indicators — such as AI-generated demand forecasts or churn probability models — anticipate future performance. By embedding all three types into Board dashboards, organisations can see beyond the immediate, and make informed decisions about scaling AI initiatives even before the full financial impact is visible.

This integration also makes it easier to track non-financial value. Customer satisfaction can be measured alongside operational metrics. Innovation activity can be monitored as a driver of long-term revenue. Skills and capability development can be tracked as part of human capital reporting. When these indicators are linked to the AISA stages of maturity, they help Boards understand when to expect certain types of value, and avoid prematurely abandoning projects whose most significant returns lie ahead.

Evidence from industry

The financial services sector provides a clear illustration. InterVision has reported on banks using AI for fraud detection, where the initial ROI was calculated on reduced fraud losses and investigation costs. However, as the systems matured, the same banks found they could launch new, lower-cost products to customer segments previously considered too risky, opening entirely new revenue streams.

In retail, PwC’s work with global brands shows that AI-driven personalisation often starts with modest uplifts in conversion rates. Over time, the data generated by these systems reveals new customer segments and product opportunities, leading to the creation of subscription models and loyalty programmes that significantly extend customer lifetime value.

Healthcare has seen similar trajectories. According to Hypestudio’s 2025 review, hospitals implementing AI scheduling tools initially reported reduced no-shows and improved utilisation. Two years later, some of those same institutions had used the underlying data infrastructure to launch telemedicine services and integrated care pathways, generating new revenue while improving patient outcomes.

The competitive imperative

The risk for Boards is not just undervaluing individual projects, but falling behind in the broader race to embed AI across the enterprise. BCG’s AI Radar 2025 notes that one in three companies plans to invest more than $25 million in AI this year, but only a quarter report significant value capture to date. The leaders are not necessarily those spending the most, but those who have built the governance, data infrastructure, and skills to identify and exploit cascading benefits.

Balancing opportunity with responsibility

Boards should approach hidden ROI evaluation through the lens of their six primary concerns: strategic alignment, ethical and legal responsibility, financial and operational impact, risk management, stakeholder confidence, and safeguarding innovation. I explain each in depth in my article on the Board’s AI governance concerns, but the essence is ensuring that as cascading benefits emerge, they are captured responsibly, transparently, and in line with the organisation’s strategic priorities.

Recognising the hidden ROI of AI should not obscure the challenges that accompany it. Ethical considerations, from bias in algorithms to the responsible use of customer data, demand careful governance. Data privacy regulations impose constraints that can slow implementation, while integration and change management costs can be substantial. These are not reasons to avoid AI investment, but they are factors that must be accounted for in both business cases and oversight mechanisms if the promised cascading value is to be realised responsibly.

Looking ahead

The question is not whether second and third-order effects exist — the evidence is overwhelming — but whether Boards have the vision and the governance structures to capture them. Measuring only first-order ROI is like judging a book by its opening chapter; you may miss the real story entirely.

The choice facing Boards is stark. You can optimise for quick wins, bank the short-term gains, and risk being overtaken by competitors who are building for compounding advantage. Or you can take a portfolio approach to AI investment, balancing immediate returns with initiatives designed to unlock future possibilities.

In an era where the half-life of competitive advantage is shrinking, the latter path demands courage. But it is also the one most likely to deliver sustainable value — and to position your organisation not just to adapt to the future, but to shape it.

About the Author

Mario Thomas is a transformational business leader with nearly three decades of experience driving operational excellence and revenue growth across global enterprises. As Head of Global Training and Press Spokesperson at Amazon Web Services (AWS), he leads worldwide enablement delivery and operations for one of technology's largest sales forces during a pivotal era of AI innovation. A Chartered Director and Fellow of the Institute of Directors, and an alumnus of the London School of Economics, Mario partners with Boards and C-suite leaders to deliver measurable business outcomes through strategic transformation. His frameworks and methodologies have generated over two-billion dollars in enterprise value through the effective adoption of AI, data, and cloud technologies.