In Prague, Tracy Sheen’s speech at the 73rd FCEM World Congress highlighted a still-common confusion. Putting artificial intelligence tools in the hands of employees does not mean they are actually adopted, nor that the company is transforming. Between governance, skills and value creation, the real challenge now lies in leaders’ ability to organize this transition.
By Pascale Caron
Many companies think they have started their transformation through artificial intelligence because they have given their employees access to ChatGPT, Copilot or other generative tools. This is one of the most significant confusions of the moment. Making a tool available creates a possibility. Employees still need to understand why they are using it, what they can entrust to it, the limits they must respect and how this technology fits into their work.
During the second day of the 73rd FCEM World Congress in Prague, Tracy Sheen, founder of UnusualComms in Australia, brought this question back to the terrain of leadership. Her presentation did not seek to demonstrate the power of a new model. Rather, it questioned organizations’ ability to absorb a technology whose uses are evolving faster than their internal structures. This shift is essential: the challenge is no longer just having AI, but knowing what the organization wants to do with it, within what framework and with what skills.
Access, adoption, organizational capacity and transformation
I propose distinguishing four levels. The first is access: the company makes tools available. The second is adoption: employees actually start using them. The third is organizational capacity: the company develops the skills, rules, processes and governance to use AI consistently. The fourth is transformation: these new capabilities actually change the way work is done, decisions are made and value is created.
This framework is neither Tracy Sheen’s nor the AI Act’s. It’s a strategic reading I draw from these discussions. It helps understand why a company can have hundreds of generative AI licenses without having actually transformed its organization. Some employees experiment intensively, others remain cautious, some develop their own methods and managers don’t always have a consolidated view of usage. AI is present, but the company hasn’t yet learned to work with it.
GIST: start with the framework, not the tool
Tracy Sheen structures her approach around a framework called GIST, for Guardrails, Intent, Strategy, Practical Training. On her official website, she emphasizes that order matters: guardrails come first, then intent, strategy and finally practical training. The idea is simple: sustainable adoption cannot rely solely on individual enthusiasm or the availability of a new tool.
Guardrails define what the organization allows and what it refuses. What data can be used? What information must remain internal? What results require human review? Who assumes final responsibility when AI makes mistakes? A clear framework is not necessarily an additional bureaucratic layer. It can instead allow employees to experiment without having to improvise the limits of their use themselves.
Intent then forces a return to an often-neglected question: why do we want to use AI? The starting point should not be the available model, but the problem to solve, the capability to create or the experience to improve. Strategy then organizes trade-offs: which uses deserve to be developed, which risks are acceptable, which skills need to be strengthened and which processes actually need to evolve? Finally, Practical Training reminds us that a policy only becomes operational when teams know what to do concretely in their jobs.
Train before demanding
One of the most relevant messages of the presentation concerns skills. Employees receive new tools, but expectations can increase immediately: produce faster, automate certain tasks, shorten deadlines or integrate AI into activities previously performed manually. If the organization doesn’t support this evolution, a technology meant to reduce workload can become a new source of pressure.
Training doesn’t just mean learning to write better prompts. It’s about understanding the capabilities of tools, their limits, data-related risks, the need to verify certain outputs and the responsibilities associated with decisions. A leader, recruiter, salesperson, engineer or communicator don’t have the same uses or the same risks. AI mastery must therefore be adapted to roles and context.
What the AI Act is already changing in Europe
This question now goes beyond just good managerial practice. Article 4 of the European AI Act requires providers and deployers of AI systems to take measures to develop AI literacy. This obligation concerns their personnel as well as people who use these systems on their behalf. This obligation must take into account technical knowledge, experience, training and context of use. Since the modification introduced by European regulation 2026/1744, the text also specifies that companies are not required to guarantee a determined level of mastery for each individual.
This nuance is important. The AI Act does not impose a uniform curriculum or identical certificate for all employees. However, it does ask concerned organizations to take appropriate measures. The European Commission first recommends ensuring general understanding of AI in the organization. It then recommends identifying the company’s role as provider or deployer, assessing risks related to systems used and adapting awareness or training actions to these elements.
Another shortcut must also be avoided: the AI Act does not impose identical human oversight for all uses of artificial intelligence. The specific oversight obligations provided in Articles 14 and 26 concern systems classified as high-risk. Following the 2026 Digital Omnibus, the main obligations applicable to Annex III systems will come into force from December 2, 2027. They notably concern certain uses in employment, education, biometrics or critical infrastructures. High-risk systems integrated into certain regulated products under Annex I follow a timeline extending until August 2, 2028. For leaders, the message is therefore twofold: don’t underestimate the regulatory framework, but also don’t transform every use of AI into a disproportionate compliance procedure.
Saving time is not yet creating value
The other interesting contribution of this reflection concerns productivity. Artificial intelligence is often evaluated by the time it saves. A task performed in three hours can now be done in one hour. Two hours seem to have been gained. But what becomes of this capacity? If it’s immediately replaced by more similar tasks, the company has improved an efficiency indicator without necessarily transforming its activity.
The strategic question then becomes: what do we want to do with the freed time? It can be reinvested in customer relations, reflection, innovation, business development, management or decision-making. From this perspective, AI no longer just serves to do things faster. It allows human attention to be redeployed toward activities where experience, judgment and relationships produce more value. A company can achieve productivity gains without changing its model. Transformation begins when technological gains change the way resources are used and decisions are made.
AI as mediator between different ways of thinking
Tracy Sheen also mentioned a less frequently discussed use. She explains using it to facilitate communication with interlocutors whose way of reasoning differs from her own. This testimony should be presented as a personal experience, not as a demonstrated property of AI. Nevertheless, it opens an interesting avenue for organizations.
Collaboration difficulties don’t always stem from fundamental disagreement. An engineer, leader, financial officer or creative can look at the same problem with different languages and representations. An AI system can help reformulate an idea, change its level of detail or translate it into another profession’s vocabulary. In the case of neurodiversity, this mediation function also deserves to be explored with caution. We talk a lot about AI as an automation tool. We still talk relatively little about its potential as support for the circulation of ideas and, perhaps, collective intelligence.
Human judgment changes role
The more performant systems become, the more another question appears: who validates their results and in what situations? A mediocre answer is relatively easy to challenge. A very convincing but incorrect answer can be harder to detect. Expertise therefore doesn’t disappear with automation. It changes function. It’s no longer just about producing an answer, but knowing how to question it, evaluate it, contextualize it and sometimes refuse it.
This evolution is particularly important in environments where decisions affect people, rights, resources or critical processes. It doesn’t mean every use of AI must be subject to heavy validation. It means the level of control must be proportionate to risk, impact and context. This is precisely one of the structuring principles of responsible AI governance.
Leadership shifts toward the ability to learn
The leader’s role is no longer to identify every new model that appears on the market. It’s to create an environment in which the organization can learn without losing control of its data, responsibilities and priorities. Technology evolves too quickly for a strategy to be based on a fixed list of tools. It must rather rest on decision principles. What problems do we want to solve? What data can we use? What results must we measure? What level of risk do we accept? What skills do we want to retain or develop internally?
This is where the GIST method joins a broader evolution of business strategies. Competitive advantage probably won’t come sustainably from simple access to models, since these capabilities spread quickly. It could depend more on organizations’ ability to integrate them, build their own skills, measure results and modify their processes when experimentation demonstrates real value creation. Adoption then becomes a collective skill rather than a sum of individual uses.
Women must participate in decisions that shape AI
The context of the presentation, devoted to new women’s leadership, finally gives an additional dimension to this reflection. The question is not only whether women use artificial intelligence tools enough. If they truly want to participate in the transformation of the economy through AI, they must be present where decisions are made. This concerns governance, investments, skills development, job design and allocation of created value.
Using a technology and participating in its governance correspond to two different levels of power. The first consists of knowing how to work with the tool. The second allows deciding the place it will occupy in the organization. Women must therefore be present among users and entrepreneurs who experiment with AI. They must also participate in teams that define rules, choose investments, assess risks and decide what should or should not be automated.
From use to transformation
Asking companies whether they use artificial intelligence is becoming progressively insufficient. The real question is now: has their organization learned to work with it? Has it defined its intent, rules and priorities? Do its employees have the necessary skills? Are time gains converted into value? Do responsibilities remain understandable when automation increases?
AI provides a new technological capability. It defines neither the strategy, nor the culture, nor the uses the company wants to make of it. These choices remain human. This is precisely why the adoption of artificial intelligence is, above all else, a leadership issue.
