Cutting to invest in AI is not enough: ROI is built on skills

Cutting staff to finance artificial intelligence may free up budget in the short term, but it does not guarantee returns. Value emerges when the company invests in skills, governance, processes, and the TCO of adoption.

The problem is not that AI is too expensive. The problem is that many companies still interpret its cost as if it were only the visible part: tokens, licenses, vendors, APIs. In reality, the bill is much broader and includes integration, training, supervision, change management, data quality, and operational risk. And that’s where ROI is determined.

The Gartner research highlighted by AI4Business should be read in this light: layoffs may free up budget, but they do not, on their own, produce a return on investment. As Helen Poitevin of Gartner summarized, “workforce reductions may create budget room, but they do not create return.” Value comes when AI is embedded in an operating model capable of strengthening human work, not weakening it.

The cost you don’t see

In the AI debate, a rather persistent simplification continues to circulate: fewer people, more automation, more margin. It’s an intuitive formula, but an incomplete one. Artificial intelligence does not work as a linear cost‑saving lever; it is a transformation that shifts work, redistributes responsibilities, and introduces new oversight needs.

When an AI system truly enters business processes, the initial cost is only the starting point. You need integration with existing systems, policy definition, data stewardship, team training, and the ability to manage exceptions. Without these elements, adoption remains fragile and, over time, does not generate value: it generates complexity.

This is where TCO becomes the right metric. The Total Cost of Ownership allows companies to look beyond the price of the tool and evaluate the full cost required to make it truly productive. In an AI context, ignoring it means underestimating the project from the start.

What Gartner says

Gartner’s message is direct. Among organizations piloting or deploying autonomous business capabilities, about 80% reported workforce reductions. But those reductions do not appear to translate into better ROI.

The most important data point is not the presence of cuts, but the fact that they do not meaningfully distinguish those achieving high returns from those with modest or negative results. Eliminating people, in other words, is not enough: it may free up resources, but it does not build capability.

The report’s conclusion is clear: the companies that improve ROI are not those that reduce the need for people, but those that invest in skills, roles, and operating models capable of enabling humans and autonomous systems to work together. Gartner calls this approach human-amplified business, likely the most useful framework for understanding AI adoption at this stage.

Why cutting staff does not create value

The appeal of headcount reduction lies in its immediacy: numbers improve quickly, the budget lightens, and the project seems more sustainable. But this is a bookkeeping victory, not necessarily a strategic one. This explains why TCO is so relevant: AI does not function as a simple cost reduction, but as an organizational capability that requires human infrastructure.

The more autonomous a system is, the more oversight it needs. The more it automates, the more exceptions it generates, edge cases to validate, and anomalies to correct. Productivity does not arise from the disappearance of human work, but from its reconfiguration into higher‑value functions. If this reconfiguration does not occur, the company risks weakening precisely the people needed to make AI work.

Without redesigning the operating model, the initial savings may be absorbed by indirect costs: errors, slowdowns, loss of knowledge, lower quality, and greater dependence on vendors.

Value lies in people

Perhaps the most important part of Gartner’s message is this: returns come when the company invests in people. Not for cultural reasons, but because AI produces value only if someone guides it, controls it, and integrates it into daily work.

This implies at least three steps:

Creating new roles related to orchestration, governance, and automation oversight
Investing in upskilling for teams impacted by the change
Rethinking processes, because AI does not fit well into an organization that remains identical to before

The real question for CEOs is not how many positions can be cut, but which operating model makes AI sustainable over time. ROI does not depend on reducing headcount, but on the quality of the organizational architecture that supports it. A more skilled workforce makes AI more effective, safer, and more sustainable.

From adoption to return

The Stanford AI Index 2026 report adds a useful element: AI adoption is growing much faster than fully visible economic returns. In 2025, AI usage in organizations reached very high levels: 88% of organizations report using AI in some form, but AI agents are still in an early phase in most business functions. Adoption is accelerating, but value is not yet distributed evenly.

This is precisely where the importance of TCO becomes clear: if a technology enters processes before it enters operating models, the bill arrives immediately while benefits take time to emerge. Stanford also shows that productivity gains are stronger in structured and measurable tasks, while in activities requiring deep reasoning the impact is more uncertain. This confirms that AI ROI is not automatic: it depends on the type of work, the maturity of adoption, and the company’s ability to truly integrate the technology into its operating model.

Where AI truly creates value

AI does not produce the same results in all contexts. The clearest benefits appear in structured, repetitive, and easily measurable tasks: customer support, software development, marketing output. Where work requires deeper reasoning, complex coordination, or greater decision‑making capacity, the impact is less linear and often harder to translate into immediate returns.

This reinforces a central point: introducing technology is not enough to achieve efficiency. You must understand where it can truly amplify human work and where it risks adding complexity. This is also why TCO must be evaluated function by function, not as a single abstract figure: in some areas AI can generate value very quickly, while in others it requires more oversight, more integration, and more time to become sustainable.

A market in transition

AI adoption is growing rapidly, but organizational maturity is not advancing at the same pace. This gap explains many current tensions: companies want quick benefits, but AI requires time to be absorbed, governed, and made productive. It is a transition that produces costs first and only later, potentially, returns.

Success stories are not those that simply replaced people with software, but those that redesigned work: they defined new responsibilities, introduced controls, strengthened skills, and created a more sophisticated balance between automation and oversight.

The market is therefore entering a more mature phase: less storytelling and more execution, less abstract enthusiasm and more attention to how AI actually fits into processes. This is an important shift because it makes the debate more useful for those who must decide budgets and priorities.

The risk of AI without governance

One of the most frequent mistakes is disorganized adoption. If AI is used individually, without shared rules and without centralized oversight, the company is not building a strategy: it is accumulating risk. And when risk grows, costs almost always follow — and they do arrive.

Without governance, outputs become inconsistent, data becomes fragmented, responsibilities blur, and quality tends to degrade. The project may seem efficient at first, but in the medium term it can create more problems than it solves.

This is why TCO is also a matter of control. It’s not enough to ask how much the technology costs: you must ask how much it costs to make it governable. Often this is precisely the difference between an interesting experiment and a truly scalable corporate capability.

AI ROI depends on skills

The point is not whether AI is expensive. The point is understanding how much it really costs to adopt it well. The price of the tool is only part of the story. Value depends on everything around it: people, processes, data, governance, and operational capability.

Gartner shows that layoffs may free up budget, but they do not generate returns on their own. ROI comes when the company invests in the conditions that allow AI to truly work: skills, roles, control, and integration into human work. AI does not reward those who cut fastest. It rewards those who build better.

Sources: