Beyond Headcount: Rethinking Artificial Intelligence ROI

Image generated using AI with ChatGPT

If Michael Jackson were still alive (some people believe he is), he might feel that his title as the “King of Pop” was under threat. Today, in almost every conversation, corporate event, professional conference, publication, and debate, Artificial Intelligence is “the topic.”

The hype surrounding artificial intelligence has created a collective anxiety that is pressuring executives to demand rapid results from their technology teams. Over the past three years, a growing perception has emerged that organizations failing to invest in AI would inevitably fall behind and be pushed to the margins of history. Budgets were redirected, strategic priorities shifted, and now the time has come to demonstrate results and justify those decisions.

In this context, laying off large numbers of people and pointing to reduced labor costs appears to be an obvious ROI metric. But is it really?

This article does not question the potential of AI. Quite the opposite. It starts from the recognition that we are facing one of the most transformative technologies of recent history. The real point of reflection is different: AI itself is not the problem. What we need to discuss is whether we truly understand the value we intend to obtain from it, and whether what we are actually achieving as outcomes corresponds to what we originally intended.

Join the discussion and leave your comment for this article on LinkedIn. Also available in Portuguese: “Além do Headcount: Repensando o ROI da Inteligência Artificial?“.

The Problem with ROI in Business Transformation

“I bought a hair clipper for $40 and now I no longer need to spend money going to the barber. In four months, it pays for itself.”

The payback logic seems simple. But it does not take much reflection to realize something is wrong with that calculation. A $10 barber’s work is not simply about running a clipper over someone’s head. There is skill, experience, aesthetic judgment, improvisation, human connection, and professionalism that do not come embedded in the machine. It is not an equivalent exchange.

With AI, many organizations seem to be falling into a very similar trap.

Traditional ROI calculations tend to privilege tangible and immediate metrics. Monthly cost reduction is easier to measure than strategic value, innovation capability, knowledge retention, customer satisfaction, or brand strengthening. Indicators that rarely appear in short-term spreadsheets.

According to recent research from PwC, many CEOs still struggle to identify concrete financial gains from generative AI initiatives despite the enormous volume of investment already made [1]. At the same time, studies associated with MIT suggest that a large portion of AI initiatives still fail to generate measurable P&L impact, often due to poor integration with business processes and organizational objectives [2].

The superficial way some organizations are attempting to generate value from AI may itself be the greatest barrier to achieving meaningful results.

The Hidden Cost of Artificial Intelligence

Even when considering cost alone, the hair clipper example remains far too simplistic compared to adopting AI in a real business environment.

AI is not an off-the-shelf product that can simply be purchased and deployed overnight. Corporate AI solutions require investment in analysis, development, integration, testing, deployment, governance, and security. And once implemented, the costs continue.

AI systems require ongoing human supervision to handle situations that cannot be fully automated. Teams must monitor errors, correct hallucinations, address security risks, review biases, and continuously fine-tune models. On top of that, there are computational inference and token consumption costs, which are far from negligible and, worse still, often unpredictable.

Recent research on AI governance and operational economics has already begun discussing the need for “AI FinOps” practices, bringing together IT, finance, and business teams to balance speed, cost, and performance — especially because model consumption behavior can vary dramatically depending on context, volume, and even how users interact with the system.

This does not mean AI is “not worth it.” It simply means that the total operational cost is frequently underestimated or ignored in overly optimistic business cases.

When Efficiency Destroys Organizational Capability

Once again: the hair clipper is not the barber.

When companies conduct layoffs to replace people with automation, they risk throwing the baby out with bathwater. What disappears is not merely operational cost. Tacit knowledge, organizational memory, market relationships, long-built trust, and corporate culture are lost as well.

Relationships between customers, suppliers, and organizations often depend more on people than on processes. A company is not merely a collection of automatable workflows. It is also composed of collaboration networks, historical context, informal communication, and collective intelligence.

Researchers such as Ikujiro Nonaka and Thomas Davenport have spent decades exploring the value of tacit knowledge and organizational memory as strategic assets that are extremely difficult to replace [3][4].

And several recent cases seem to reinforce this concern.

Klarna attracted global attention after announcing significant replacement of human customer service with AI. Later, the company partially revised its strategy and resumed hiring people for certain support and relationship-management functions, acknowledging important limitations related to customer experience and the need for human interaction in more complex situations [5].

Something similar happened with Duolingo. After announcing an “AI-first” strategy and gradually replacing contractors through automation, the company faced strong public backlash, damage to brand perception, and criticism regarding both quality and the human consequences of those decisions. Later, executives publicly clarified that AI was intended to amplify human capabilities rather than simply eliminate people [6][7].

These cases do not mean AI has failed. They just show that overly simplistic automation strategies can create significant side effects when cost-reduction metrics become confused with sustainable value creation.

The Macroeconomic Question: Productivity for Whom?

There is nothing wrong with prioritizing AI investments. Nor is there anything wrong with pursuing productivity or financial return. Research and technological development have always played a fundamental role in transforming society and the economy.

The problem begins when the only strategic compass becomes headcount reduction to justify quarterly targets and executive bonuses, and when innovation is understood purely in terms of cost reduction and profit growth without considering its broader social impact.

This short-term mindset can create a dangerous distortion: treating human beings merely as replaceable operational costs.

Recently, even the Holy See entered this debate through the encyclical Magnifica Humanitas issued by Pope Leo XIV [8].

The document warns about the risk that advances in artificial intelligence may be conducted primarily for the benefit of a few economic groups, increasing inequality and reducing human dignity to metrics of efficiency and competitive advantage.

“It is certainly desirable for technology to relieve humans of arduous, repetitive or dangerous tasks and to provide intelligent support for human activity. Yet, the protection of employment opportunities and the irreplaceable role of the individual must remain the general rule. The pursuit of greater profits cannot justify choices that systematically sacrifice jobs, because the human person is an end, not a means, and the economic order must remain subordinate to human dignity and the common good.”

(Pope Leo XIV)

The Pope’s letter does not condemn AI. Instead, it recognizes its enormous transformative potential. But it calls on governments, companies, professionals, and society to participate consciously in this transformation, placing the human person at the center of technological decisions.

Economists such as Daron Acemoglu have also warned that automation does not always generate benefits distributed equally across society. The question is not merely how to produce more, but who benefits from those productivity gains and what social and economic capabilities may be destroyed in the process [9].

OECD reports similarly warn about increasing inequality and labor-market polarization if governments and organizations fail to invest in adaptation, reskilling, and more sustainable mechanisms for economic transformation [10].

It is important to remember that we are not facing an unavoidable natural disaster. AI advances are driven by human decisions. That means executives, investors, technology leaders, and business analysis professionals hold direct influence and responsibility over how this transformation unfolds.

Toward a More Human and Intelligent Model for Artificial Intelligence ROI

We need a more mature way to evaluate the success of AI initiatives.

Reducing labor costs can absolutely be a legitimate benefit of automation. In many contexts, replacing repetitive work with technology makes sense and creates real gains. The problem emerges when this becomes the only (or primary) metric used to justify complex transformation investments.

AI should not be viewed merely as a substitute for human capability. Its greatest potential may lie precisely in expanding that capability.

Organizations must define, upfront, the outcomes expected from each initiative so that investments can be guided strategically.

Possible indicators for evaluating AI ROI include:

customer satisfaction;
quality;
risk reduction;
innovation speed;
knowledge retention;
quality-of-life improvements (for customers, employees, or society);
organizational intelligence;
adaptability;
analytical capability;
long-term business sustainability.

This type of discussion requires professionals capable of connecting technology, strategy, people, and business value in an integrated way. It requires a less mechanistic and more systemic view of organizational transformation [11].

Some initiatives are already beginning to explore different paths.

A recent report from the Brookings Institution and AIPI on digital sovereignty among Native American tribal nations presents a very different perspective from the one commonly found in traditional business environments [12].

Within these communities, AI is being discussed as a tool for strengthening local institutions and expanding human capability, and not primarily as a mechanism for replacing workers. The explicit goal is to preserve community jobs, strengthen autonomy, and ensure that technological gains remain aligned with the values and dignity of the people affected by the transformation.

It seems to me that this perspective has something important to teach the rest of the market. Prioritizing the preservation of jobs over immediate profit maximization may sound strange within traditional capitalist logic, but it becomes far more understandable when human dignity and social cohesion become part of the equation, especially in a world increasingly connected as a global tribe.

Conclusion

Traditional ROI models designed for process automation may no longer be sufficient for guiding decisions in the age of artificial intelligence.

Viewing AI merely as a mechanism for replacing labor is an extremely limited interpretation of a technology with far greater potential. We need to develop ways of directing implementation that are more intelligent, more human, and more sustainable.

If we truly want AI to lead us toward a better stage of economic and social development, we will need responsible leaders, more mature metrics, and organizations capable of looking beyond short-term results.

After all, productivity should not simply mean producing more with fewer people. It should mean creating more value for customers, companies, professionals, society, and the environment.

From AI4BA to BA4AI: Empowering Analysts to Lead the AI Revolution – Presentations and workshops on how to do business analysis for AI transformations