Over the past two years, publications on artificial intelligence in business have multiplied, offering us an increasingly dense but also increasingly fragmented landscape. Two main types of reviews dominate.
On the one hand, technology and model vendors describe a massive potential that is still largely untapped: capabilities that grow quickly, advanced uses reserved for a minority, a growing gap between what AI can do and what is actually used (Two interpretations of AI in the workplace: agentic uses and capability overhang).
On the other, consulting firms make a more prosaic observation: low value creation at the business level, a minority of truly successful organizations, the majority of initiatives confined to pilot projects, dispersion of efforts, and chronic difficulty in scaling up (Adoption and impact of AI: lessons (and limitations) from the latest McKinsey and BCG studies).
Taken in isolation, these two discourses seem almost contradictory: the former talks about power and potential, while the latter talks about limited value and relative failure. However, this opposition does not reflect a fundamental disagreement about AI itself, but rather reveals a profound disconnect in what is observed, measured, and how it is interpreted.
Vendors talk about what AI enables, consulting firms talk about what it brings in, but neither seriously looks at what it does at work. The gap between their narratives does not reveal a divergence on AI, but rather a persistent inability to think of work as a central object of transformation.
And, before going any further, I repeat my usual warning: ask yourself how the author of a report earns a living and you will often know what the report says in broad terms without having to read it. And when talking about the future, always be careful not to confuse predictions with forecasts (AGI, employment, productivity: the great bluff of AI predictions).
In short:
- Technology vendors highlight the untapped potential of AI in business, while consulting firms emphasize the low value created on a large scale, revealing two complementary but different perspectives: capabilities vs. performance.
- These two discourses converge on several points (organizational barriers, polarization of uses, importance of structured integration), but differ in their framing: vendors see a gap between capabilities and uses, while consultancies see a gap between investments and value.
- Work, as a complex system, remains conspicuously absent from these analyses: it is either a simple framework for use or a lever for performance, but rarely a central object of transformation.
- The limitations of current analyses stem from their respective frames of reference, which are influenced by their economic interests: vendors value individual use, while consulting firms value governance, without really questioning the design of work.
- True transformation lies neither in the tool nor in the strategy, but in the ability to rethink work itself. As long as this remains a blind spot, the gaps between potential, usage, and value will persist.
Two points of observation, two realities
Vendors observe AI based on model capabilities, individual uses, and adoption behaviors. They have access to an impressive amount of data, sometimes empirical, sometimes very detailed, but always focused on the interaction between a user and a technology. The problem is then framed in terms of appropriation, agency, skill development, and the gap between capacity and actual use.
In this interpretation, AI is above all a cognitive infrastructure, a term which, as we will see in a future article, is completely meaningless: it is progressing faster than the ability of individuals and organizations to absorb it. The core of the diagnosis is therefore not failure but delay. Value is possible, as demonstrated in certain advanced uses, but it remains under-exploited on a large scale.
Consulting firms, on the other hand, observe AI from within organizations themselves. Their entry point is P&L, governance, the ability to drive complex transformations, mobilize resources, and produce measurable effects on performance. The problem is then framed in terms of value captured, return on investment, transformation at scale, or even greater competitiveness (Technologies sell productivity, but businesses want revenue and The great illusion of technological productivity gains (including AI)).
In this reading, AI is not primarily a cognitive infrastructure but a strategic lever. The question is not what AI can do, but what it actually transforms in economic models, processes, overall performance and, above all, if we want to take things in the right order, work.
These two points of observation automatically produce two different narratives, not because they describe incompatible realities, but because they do not observe the system from the same place.
Compatible but not identical findings
When comparing these interpretations, however, there are strong similarities.
Both recognize that technology is no longer the main limiting factor. The capabilities are available, accessible, and industrializable, and the obstacles are organizational, cultural, and structural.
Both show that value emerges when AI is integrated into structured, repeatable processes, rather than when it remains confined to peripheral or exploratory uses (Collective appropriation of AI: the only condition for tangible impact). I would add that thinking about the governance of gains even before they appear is a very healthy precaution (Without governance, the gains from AI are virtual).
Both describe a dynamic of polarization: a minority of players capture a growing share of the value, while the majority remain in superficial, unproductive uses (AI adoption does not replace productive appropriation).
But these convergences mask an essential difference in framing. Vendors describe a gap between capabilities and uses, while consultancies describe a gap between investments and value.
These are two different projections of the same phenomenon, but neither focuses on the transformation of work itself.
Work is the blind spot of AI
In both discourses, work appears as a context for observation, rarely as an object.
For vendors, work is a usage environment, a framework in which AI is mobilized, more or less effectively, by individuals or teams. Conversely, for consulting firms, work is a performance lever: a set of processes to be optimized, automated, and transformed to produce value.
But in both cases, it is rarely treated as a complex system, structured by trade-offs, constraints, organizational legacies, implicit norms, power dynamics, routines, and compromises. It is viewed in an overly simplistic manner that minimizes its importance in the face of technological advances, a kind of denial of its complexity and even its natural intricacy.
Yet this is precisely where the transformation is taking place.
The gaps observed, whether between potential and usage or between investment and value, are not only problems of adoption, management, or maturity, but are also the product of organizations designed for another work, onto which new technology is projected without any structural overhaul of the framework in which this work is carried out.
AI is then integrated into systems that are already saturated in terms of procedures, coordination, reporting, fragmentation of responsibilities, and organizational debt (AI Reasoning Is Cool, But First How Can We Tackle Organizational Debt and How to Tackle the Biggest Threat to Your Team’s Growth). In this context, it does not transform work but is added to it.
Two discourses, one structural limitation
Vendors, by design, cannot fully see this dimension. Their observation system focuses on uses, capabilities, and adoption trajectories, and objectively taking it into account would be marketing suicide for an industry that, despite raising a lot of money, struggles to get its customers to pay.
They see very clearly what AI can do, but few care that organizations are preventing it.
Consulting firms, for their part, see very clearly the macro effects (dispersion of initiatives, difficulty scaling up, low value capture, concentration of profits), but their reading remains largely structured by management categories: governance, leadership, strategy, transformation, organizational playbooks. As with vendors, unsurprisingly, their interpretation is tailored to their business and marketing.
In both cases, the transformation of everyday work remains largely under-analyzed and often reduced to an adjustment variable.
This is not an accidental oversight but a structural limitation of the dominant analytical frameworks, which continue to think of digital transformation as a question of tools, strategy, or value, rather than a question of work design(Prepare the business and work before integrating AI and How management let systems do the thinking for them).
A false opposition
Pitting the discourse of vendors against that of consulting firms is therefore tantamount to creating a false controversy, because they do not describe two incompatible realities, but two different levels of the same system: one observes capabilities and uses, the other observes organizations and value, but neither directly observes the transformation of work as a system.
What appears to be a contradiction is in fact a difference in focus.
AI can be both technologically mature and organizationally inefficient, powerful but unproductive, and can transform certain tasks without transforming work systems.
What the comparison between the views of vendors and consulting firms reveals
While comparing these two types of discourse does not allow us to determine who is right, it does help us understand why the same technologies produce such diverse effects.
It shows that the central issue is neither adoption, strategy, nor even value, but rather the ability of organizations to absorb a technology that changes the very nature of work without rethinking their structures (Taking back control of enterprise design: intention before tools and If your business isn’t designed for AI, it will end up being designed by AI).
As long as AI is thought of as a tool to be integrated into unchanged organizations, it will produce discrepancies, and as long as transformation is thought of as a project rather than a restructuring of work, the gaps between potential, use, and value will continue to widen.
Bottom line
Vendors talk about what AI enables, and consulting firms talk about what it brings, but the transformation is happening elsewhere: in what it does at work, in the way it reconfigures tasks, roles, trade-offs, coordination, and responsibilities.
The gap between their reviews does not reveal a disagreement about AI, but rather a persistent difficulty in thinking about work as an object of transformation in its own right.
As long as this issue remains peripheral, reviews will continue to diverge and the gaps observed will continue to appear as anomalies, when in fact they are primarily structural consequences.
To answer your questions…
They are not talking about the same thing. Vendors describe the capabilities and potential uses of AI, while consulting firms analyze the actual value created at the organizational level. One looks at the potential, the other at the results. This difference in focus produces different reviews without revealing any contradiction about the technology itself.
They agree on several key findings. Both recognize that technology is no longer the main obstacle and that the barriers are organizational. They also show that value emerges when AI is integrated into structured processes and that a minority of players reap most of the benefits.
In both discourses, work is treated as a context or a lever, rarely as a complex system in need of transformation. Its reality, made up of routines, constraints, and trade-offs, is simplified. Yet it is precisely this complexity that determines the real impact of AI on performance.
Because it is often added to organizations designed for a different type of work. Without overhauling structures, responsibilities, and coordination methods, AI improves certain tasks but does not transform work systems, which severely limits the value captured.
The central issue is not the adoption of tools, but the ability to rethink work itself. As long as AI is integrated into unchanged organizations, the gaps between potential, usage, and value will persist. Transformation requires a restructuring of work, not just a technological project.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
Adoption and impact of AI: lessons (and limitations) from the latest McKinsey and BCG studies
Two interpretations of AI in the workplace: agentic uses and capability overhang
How can you prepare your organization for AI?
Generative AI: your organization is worth more than your tools
AI won’t kill consulting, but it will force it to grow
Enterprise 2.0 does not tolerate halve measures
