Through experimentation and as organizations mature, we now know pretty much everything we shouldn’t expect from AI. We know that use is not enough (AI adoption does not replace productive appropriation), that productivity does not spontaneously become value (Technologies sell productivity, but businesses want revenue), and that gains do not automatically transform into something useful and valuable (Without governance, the gains from AI are virtual), and that adoption does not start at the top or the bottom, but where the business agrees to make the connection between what is happening on the ground and what is decided collectively (Back office or front office? Where is the potential of AI greatest?).
However, one very basic question remains: once AI has been introduced, once its uses have been established, once the initial gains have been observed, what will actually become of the business?
In short:
- Some organizations compensate for the lack of impact on productivity (P) by acting on employment (L), through cost reductions or reorganizations.
- The perceived failure of technology often masks a governance problem rather than a genuine technical failure.
- The effects of AI are neither unique nor predictable, and the differences between businesses stem from how they use the technology, not from the technology itself.
- The adoption of AI, productivity gains, and value creation do not follow an automatic sequence, but depend on collective decisions and trade-offs.
- Without this collective transformation, AI remains limited to a role as a convenience tool or becomes a source of internal tension.
When AI improves everyday life without ever changing the business
In many businesses, AI is used every day. Teams use it to write faster, prepare summaries, rephrase, and analyze. No one disputes the value of these uses, which make work flow more smoothly, sometimes more enjoyably, and give the feeling that things are finally moving a little faster.
However, when we ask what this has changed for the business, we sense a certain unease. No thought has been given to how the time saved has been used, and its use has not been identified. It has simply been absorbed by the system, as is always the case with available margins in organizations that are already under pressure.
The work is being done better, but it is being done in the same volumes, with the same objectives and the same constraints. Costs are not falling and revenues are not increasing in any identifiable way. At the P&L level, it is difficult to attribute anything to AI, even though everyone recognizes that it has become useful (AI from productivity to P&L: nothing happens by chance) .
This scenario can last a long time, precisely because it creates neither conflict nor urgency. AI is neither questioned nor really defended. It becomes part of the backdrop, appreciated but politically fragile when the question of value is asked head-on.
When productivity turns into pressure
In other businesses, the question of value is raised very early on. Gains are measured, sometimes even before things have really stabilized, the use of time saved is clearly defined, and AI is clearly presented as a performance lever with a clear intention: to produce a visible effect quickly.
At first, the results are there. The indicators are changing and even the P&L shows positive signs. But pretty soon, something starts to feel tense. Teams spend more time checking, rereading, and justifying what they produce. The cognitive load increases. Some of the initial gains are gradually offset by invisible costs that, in a way, add to the “work about work” (Work about work: when the reality of work consists of making things that don’t work work).
AI then ceases to be perceived as a support and becomes a tool for putting pressure on people. Its use continues, but confidence erodes. Teams then seek to protect themselves, to circumvent it, and ultimately, reported productivity grows less quickly and eventually plateaus.
This scenario can produce short-term results and is often appreciated for its clarity. But it quickly runs out of steam if the organization does not take a more in-depth look at how work is organized.
When AI becomes an economic lever
This scenario exists, but it is rare, not because it is out of reach, but because it forces us to ask uncomfortable questions.
Uses may emerge, but only within a governance framework that defines what we do with AI, what we do with the gains it brings, and how we leverage them. The business accepts that productivity will increase before the value is reflected in the figures, and this gap creates discomfort for those who only have eyes for the P&L and margins.
Some gains are reinvested in transforming the work, others are converted into cost reductions or used to absorb more volume or modify the value proposition. Nothing is left to chance, everything is under control, and just because something is possible does not mean it will be implemented if it is not in line with the business’s intention (Taking back control of enterprise design: intention before tools and AI First is not AI Only: clarify your intentions before transforming your business).
The P&L evolves, often later than expected, but in a more sustainable way, and the value created circulates sufficiently to produce an overall effect. This scenario relies less on management than on responsibility and assumes exemplarity in decisions and trade-offs, not in usage.
It is a reference point rather than a model because it forces organizations to make choices that many prefer to avoid on issues that they sometimes do not even consider at the start of their AI project.
When everything stops due to a lack of visible value
Finally, there are businesses where there is a lot of experimentation, where uses are multiplying, where the rhetoric is ambitious, but where structural decisions never come (Stabilize to move forward: why experimentation alone does not produce results with AI). Gains remain local, the P&L does not change, or changes too slowly, and the question of value eventually becomes pressing.
Gradually, projects end up being stopped, not because they don’t work, but because they don’t produce anything defensible on a business-wide scale. AI is then seen as an overpromise.
With no impact on P, some organizations end up acting on L. Cost reductions and other reorganizations compensate for the fact that productivity gains have not been reflected in the books. This scenario is often described as a technological failure, when in fact it is almost always the result of a governance problem.
To answer your questions…
Because productivity gains often remain localized and are not translated into collective decisions. The problem stems less from technology than from governance, which prevents these gains from being incorporated into the income statement.
AI generates information and gains, but without clear trade-offs, it does not transform the organization. The lack of structural decisions prevents AI from having a global impact.
Using AI is not enough. Productivity requires organizational choices and the ability to transform local gains into collective rules and decisions.
Businesses often compensate with cost-cutting or reorganizations, which creates tension and fuels a false narrative of technological failure.
It is not the tools used, but the decisions made based on what AI produces. Value depends on the ability to make choices and sacrifices.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
Who is handling your artificial intelligence projects? Probably not the right people.
AI from productivity to P&L: nothing happens by chance
Two interpretations of AI in the workplace: agentic uses and capability overhang
Adoption and impact of AI: lessons (and limitations) from the latest McKinsey and BCG studies
Without governance, the gains from AI are virtual
Back office or front office? Where is the potential of AI greatest?
