From AI capability to organizational transformation and value creation
Pascale Caron
AI Strategy Advisor | Founder & CEO, Yunova Consulting | Co-author of EntrepreneurIA
It has never been easier to use artificial intelligence in business. In just a few months, tools have multiplied, uses have become commonplace, and experiments have spread across almost all functions. Yet one question remains surprisingly difficult:
What does this adoption actually change for the company?
Because using AI is not yet transforming. And transforming does not necessarily mean creating value. This is probably where the real issue lies today. In its global survey published in March 2025, McKinsey indicated that more than three-quarters of respondents stated that their organization was already using AI in at least one function. But the study revealed a much more nuanced reality when it examined workflow transformation, governance, or reported economic impact.
In other words, technology diffusion is advancing faster than organizational transformation.
This observation aligns with what I have observed for several years in my conversations with leaders and, more particularly, throughout the more than one hundred interviews conducted as part of EntrepreneurIA. Discussions often begin with technology. They almost always end with organization.
Who decides? What process really needs to be transformed? What skills should be preserved or developed? How far should we automate? Where should human judgment remain? And above all, how do we know if AI is actually producing value?
It is from these questions that a conviction has gradually emerged in my work:
Artificial intelligence creates capability. Organizations create value.
This conviction does not claim to invent a new concept. On the contrary, it intersects with an already established field of research around AI capabilities and value creation. But it takes on particular importance today, at a time when access to technology is becoming more widely shared.
My analysis therefore draws on three complementary perspectives: the academic literature on organizational capabilities related to AI, my experience with technological transformations, and the qualitative observations from EntrepreneurIA. These three perspectives lead to reframing the question.
For a long time, companies have been asking:
what AI should we adopt?
They must now ask another:
What organization must we build to transform the possibilities offered by AI into value?
It is precisely this shift from technology to organizational capability that I wish to analyze here.
AI capability is not limited to technology
A company can have high-performing models, solid infrastructure, abundant data, and trained employees without necessarily possessing a mature organizational capability to create value with AI.
This distinction is at the heart of the AI capability concept.
In a study published in 2021 in Information & Management, Patrick Mikalef and Manjul Gupta define and measure AI capability based on AI-specific resources that, when combined, enable the company to leverage this technology. Their framework, grounded in the resource-based theory of the firm, notably distinguishes tangible resources, human skills, and organizational resources. Their empirical study demonstrates a positive relationship between AI capability, organizational creativity, and performance in the studied sample.
A systematic literature review published by Ida Merete Enholm, Emmanouil Papagiannidis, Patrick Mikalef, and John Krogstie in Information Systems Frontiers analyzes how organizations adopt and use artificial intelligence. It also examines the mechanisms by which these uses can generate value for the company. The authors particularly emphasize the factors that facilitate or hinder adoption, forms of organizational use, and first- and second-order effects.
The strategic lesson I take from this is simple. A model is a resource. Data is a resource. Skills are resources. Capability emerges when the company knows how to combine and mobilize them in service of an objective.
This is where the technological question becomes an organizational question.
Adoption, capability, and transformation are not synonymous
We have extensively measured AI adoption by the number of licenses, users, use cases, or prototypes. These indicators remain useful. However, they alone do not allow us to conclude that a company has transformed.
An employee can write more quickly using a generative assistant. A developer can accelerate certain tasks. A marketing team can produce more content. These gains may be real, but their sum does not automatically transform the company’s operations.
I therefore propose distinguishing six stages: Access → Usage → Adoption → Capability → Transformation → Value.
This sequence is a strategic framework, not a scientifically validated maturity scale. Its value is to remind us that between access to a tool and value creation lie several organizational transformations.
The McKinsey 2025 survey provides interesting insight here. Among 25 analyzed attributes, workflow redesign showed the strongest association with the impact on EBIT attributed to generative AI as reported by respondents. CEO oversight of AI governance was also among the elements most correlated with higher reported financial impact. These are correlations from a survey, not a causal demonstration.
The issue is therefore no longer just adoption. It becomes the capability to transform.
What over one hundred EntrepreneurIA interviews have changed in my understanding of AI
With Dr. Yves-Marie Le Bay, I interviewed more than one hundred entrepreneurs, leaders, and experts using artificial intelligence as part of EntrepreneurIA. These interviews do not, in themselves, constitute a representative quantitative survey. However, they form a particularly rich qualitative corpus on how economic actors concretely approach AI.
At the outset, we expected to talk a lot about technology. The most interesting conversations led us elsewhere. They focused on decisions. Why use AI? What problem are we trying to solve? What process needs to change? What skill must be preserved? What can be automated? Where must human judgment remain? How do we measure what actually changes?
The sectors and use cases differed, but one question recurred: what is the organization really trying to transform? This is where my understanding evolved. AI can act as a revealer of the organization. It can make more visible insufficiently structured data, complex processes, ambiguous responsibilities, missing skills, or the absence of indicators to precisely define the expected value.
This observation is not a statistical conclusion drawn from EntrepreneurIA. It is a qualitative interpretation built throughout the interviews, which I now confront with work on organizational capabilities and my transformation experience.
It leads to a leadership question: does the organization know what it wants to transform, why it wants to transform it, and how it will measure the result?
When technology diffuses, organization becomes differentiating
As models, infrastructure, and enterprise tools become more widely accessible, the ability to integrate them into workflows, governance, skills, and decision-making mechanisms can become a factor of organizational differentiation. This idea is explicitly developed by McKinsey in an article published in July 2026 on the operating model. The authors observe that companies using similar technologies achieve different results and argue that the advantage shifts from the tools themselves to organizations capable of deploying them effectively. They also emphasize that an operating model cannot be purchased or instantly replicated: it results from accumulated choices about workflows, governance, talent, data, and decision-making.
However, we must avoid too hasty a conclusion. Technology remains strategic. Choices of models, architecture, security, data, and vendors can produce significant differences. But access to high-performing technology does not, in itself, guarantee value creation.
Two companies can use comparable technologies and integrate them very differently. One can add them to existing processes. The other can rethink these processes. One can measure usage. The other can seek to measure value. One can multiply experiments. The other can select, industrialize, and stop. Differentiation then occurs in the quality of organizational orchestration.
The operating model becomes an AI topic
During the early years of generative AI, the conversation focused heavily on use cases: which professions can use AI, which tasks to automate, which tools to deploy? These questions remain relevant, but they become insufficient when AI intervenes in complete workflows, in knowledge production, or in certain decisions.
Who decides? Who controls? What decision can be delegated? Where does human intervention remain necessary? How does data flow? Who assumes responsibility when a system acts? How do we organize work between humans, traditional software, and AI agents? We are now entering the operating model.
The Organizational Transformation in the Age of AI report, published by the World Economic Forum in March 2026, draws on contributions from over 450 leaders in its AI Transformation of Industries community. It identifies five principles to support AI adoption at scale. Human accountability, end-to-end operating model redesign, talent systems capable of scaling, trust based on transparency, and disciplined experimentation.
This convergence between academic literature, business surveys, and field feedback reinforces an idea: scaling AI is not merely technical deployment. It touches on how work, responsibilities, and decisions are organized.
The Yunova framework for AI Capability
To help leaders analyze this transformation, I use at Yunova Consulting a framework structured around six dimensions.
This is not an academically validated model. The Yunova framework is a strategic synthesis informed by literature on organizational capabilities, my experience with technological transformations, and qualitative observations from EntrepreneurIA.
Its objective is operational: to help executive leadership identify the conditions that can enable converting the possibilities offered by AI into organizational and economic results.
1. Strategic Intent
Linking AI to strategic priorities and the value sought.
Before talking about technology, you need to know what the organization is trying to accomplish. Reducing costs, improving customer experience, accelerating innovation, increasing quality, reducing risk, improving decisions, or developing a new service do not lead to the same choices. Without explicit strategic intent, use cases can accumulate without forming a coherent trajectory.
2. Data & Technology
Building appropriate technical and informational foundations.
Data, architecture, security, interoperability, and model choices constitute essential foundations. But they represent only one dimension of the system. High-performing infrastructure compensates for neither fuzzy strategy, nor inadequate processes, nor insufficient governance.
3. Governance & Decision Rights
Defining who decides, who controls, and who assumes responsibility.
Governance should not be reduced to compliance. It organizes decision rights: which uses require validation, what level of autonomy can be granted to a system, when human oversight is necessary, who accepts the risk, and who assumes responsibility. With agentic systems, this decision architecture becomes even more important.
4. Skills & Human Capability
Developing skills, judgment, and human-AI collaboration capability.
Training on tool usage is not enough. The organization must develop AI literacy, understanding of system limitations, domain skills, and judgment capability. Leaders are directly concerned: they do not need to become technical experts, but must understand enough about capabilities, limitations, and risks to make informed decisions.
5. Operating Model & Workflows
Rethinking work rather than simply adding AI to existing processes.
Adding AI to an existing process can improve certain tasks without questioning the process design. Therefore, we must sometimes start from the work itself: which steps remain necessary, where is human value located, which tasks can be automated, which workflows can be redesigned, and how to organize exceptions and supervision.
6. Value & Measurement
Linking new capabilities to observable results.
The number of licenses, prompts, or prototypes does not measure value. The organization must link its initiatives to relevant indicators: revenue, costs, quality, time, risks, customer satisfaction, innovation capacity, or decision speed. One question deserves particular attention: what do we do with the capacity freed up by AI? Time savings are not automatically economic gains; their value also depends on how this capacity is reallocated.
A capability is a system
The six dimensions of the Yunova framework do not function in isolation. An organization can have quality data and lack skills. It can have a clear strategy and governance incapable of supporting it. It can train its teams without modifying their workflows. It can automate numerous tasks without knowing how to measure what this automation actually produces.
The framework can therefore be read as a system: Strategic Intent + Data & Technology + Governance & Decision Rights + Skills & Human Capability + Operating Model & Workflows + Value & Measurement.
Its value is not to assign a universal maturity score. It is to identify the imbalances that prevent a technological possibility from becoming a genuinely exploitable organizational capability.
The question executive committees should now be asking
For leaders, the shift in perspective is significant. The question is no longer just: “What AI should we adopt?” It becomes:
“What organization do we want to build with the new capabilities that AI makes available to us?”
This question leads to decisions that fall directly under executive leadership: which capabilities to develop, which processes to rethink, which decisions to delegate, where to preserve human authority, which skills to strengthen, what level of risk to accept, and what value to seek.
These decisions touch on strategy, human capital, work organization, resource allocation, and accountability. They therefore cannot be left solely to technology departments.
The academic literature does not allow us to affirm that a single organizational model will guarantee value. Business surveys do not establish a simple causality between a practice and a financial result. However, sources converge on one point: value creation with AI depends on resources and organizational mechanisms that extend far beyond technology itself.
This is why I return to this conviction:
artificial intelligence creates capability. Organizations create value.
Technology opens the field of possibilities. The organization determines what it does with them.
And for executive committees, the question becomes:
does our organization truly possess the capabilities necessary to transform the possibilities offered by AI into value?
