The McKinsey and BCG reports published last fall compile all the weak signals needed to understand what is really happening with AI in organizations, but they only hint at it (Adoption and impact of AI: lessons (and limitations) from the latest McKinsey and BCG studies). In fact, a careful reading between the lines shows that value does not depend on models or tools, but on how managers redefine work, decision-making, daily practices, and how the business as a whole conceives of its own functioning. In other words, we are not talking about a technical subject, but a subject of business design.
Businesses are equipping themselves, multiplying pilot projects, and quickly training their teams, but without really expressing a clear intention, even though trust, which is essential for adoption, stems from precisely that: saying what AI should enable, and also saying what it should not become (Is AI the new Lean?). This is the logic behind approaches that refuse to use operational improvement as a pretext for downsizing, and it is this transparency that transforms the way teams embrace the tools (Lean Without Layoffs: The Commitment That Makes Continuous Improvement Work).
In this context, the role of the manager is changing profoundly, not because they need to “manage differently”, but because their environment, responsibilities, and the decision-making mechanisms around them are no longer the same. I recently asked myself how to rehabilitate the role of the manager at a time when the function is widely criticized (Let’s restore the role of the manager), and the release of these two reports allows me to add to what I was saying from a new angle, even if it ultimately confirms many of my ideas on the subject.
In short:
- The value of AI in business depends primarily on how managers redefine work, processes, and decision-making, thereby transforming their role into that of designers of organizational functioning.
- Managers must clarify the intentions behind transformations, orchestrate collaboration between humans and AI, identify operational friction, and rethink workflows to give meaning to the uses of AI.
- The integration of AI changes the nature of decisions: managers must establish rules for delegation between humans and AI, creating a hybrid decision-making system based on clarity, reversibility, and discussion.
- Middle management plays a central but often underestimated role in the adoption of AI; it must acquire skills in process analysis, continuous improvement, and AI risk management.
- Managers become responsible for data quality, consistency between strategy, technology, and business, as well as managing the cognitive and emotional impacts of AI on teams.
From control to work design
Both reports emphasize this point without always getting to the bottom of it and drawing all the conclusions: the value of AI depends directly on the ability to redesign workflows. This change transforms the managerial function, with managers no longer acting as overseers or supervisors, but rather as creators of the context in which work takes place (To manage is to design).
Clarifying the intention behind a process is the first part of this role. As long as teams do not know why a process exists or what value it is supposed to produce, AI merely automates ambiguity. Managers must therefore restore meaning to each process and operating procedure by explaining what we are trying to achieve, how, and why.
They must then orchestrate the interaction between humans and AI and identify operational frictions, those that we tolerate out of habit or even weariness and which become natural entry points for AI (“Work about work”: when the reality of work consists of making things that don’t work work).
Organizations that truly derive value from AI are those where managers adopt this role of work designer, a role in which they shape the context that enables everyone to succeed, make the best use of their abilities, and progress, rather than controlling the effects (People Are Complex – Our Systems Are Not and How to love control and not be a burden to yourself and your teams?).
Implication: managers become responsible for clarifying the intention behind transformation initiatives, orchestrating the roles between humans and AI, and identifying operational friction and workflow consistency.
Decisions redistributed with a different level of granularity
The arrival of agents is changing decision-making. McKinsey describes them as a lever in the experimental phase, while BCG already notes that they account for a significant share of the value captured. In any case, they introduce a new mechanism: decisions are no longer solely human, but they are not yet entirely entrusted to machines.
Managers must therefore define the boundary between what AI decides and what it merely facilitates. This boundary must not be fixed, as it depends on the level of risk, the quality of the data, and the maturity of the teams.
They must then establish delegation thresholds, specifying the areas where AI takes the initiative, those where it merely contributes, and those where it must remain in observation mode in order to learn. This enables the emergence of validation, discussion, or correction mechanisms, without which we move beyond the realm of “augmented decision-making” and into that of opaque decision-making (Augmented governance: AI as a lever for collective lucidity).
Finally, managers must build a hybrid decision-making system in which human contribution is not overwhelmed but repositioned where it has the most impact. It is no longer a role of authority in a vertical system. It is a role of decision architect.
Implication: managers put in place the limits, thresholds, and mechanisms that transform decisions into augmented decisions.
Middle management must evolve or it will become a bottleneck
Both reports emphasize that organizations want to move forward but remain stuck in their pilot projects. The most likely cause is unsurprising: middle management has not been prepared for what it means to integrate AI into the reality of work, and without this upskilling, the entire adoption process is blocked.
Middle management must master process analysis, because without it, it is impossible to rethink workflows or decide how AI can be integrated into them. It must understand how to integrate AI into daily work routines, because most of the gains are found in these micro-routines that are not formalized anywhere, the famous “work about work”?
It must also embrace the concept of continuous improvement as a way of constantly adjusting the human-AI hybridization (Improving a team’ s work: a story of continuous improvement). Finally, it must understand the risks associated with AI, whether they be hallucinations, biases, or security issues.
Without this evolution, middle management becomes a hindrance because the business lines never embrace the overall logic.
Implication: middle management must evolve in its positioning and posture, otherwise it may become a bottleneck in the AI transformation.
Managers must orchestrate collective learning
High-performing organizations redesign workflows, invent new ways of working, and reinvest the gains (Without governance, the gains from AI are virtual). This dynamic does not happen spontaneously and must be driven by managers.
Creating learning routines around use cases is also essential. As long as experiences remain isolated, teams relearn the same things on their own (95% of Enterprise AI Pilots “Fail”–Just Like Lean? Not So Fast and The experimental organization). Managers must therefore structure this learning, make it collective, and commit to it over the long term.
They must institutionalize the sharing of experiences, document emerging practices, stabilize workflows to prepare for the arrival of agents, and translate business constraints into requirements for AI teams.
Here, they play a true leadership role. It is not, or no longer, an executive role, but rather an integrative role in which managers ensure that what the organization learns becomes a shared asset.
Implication: the manager becomes the architect of collective learning.
Managers must manage the cognitive and emotional impact of AI on teams
This is an angle that reports mention without really exploring it, but the figures are quite clear: fear, misunderstanding, and cognitive overload are hindering adoption much more than technology. AI is changing professional identities, redistributing skills, and, in doing so, blurring the lines.
Managers must reassure teams about the intended purpose. If they leave doubt about the implications of AI, confidence will fade and teams will turn to parallel uses or workarounds. A clear, explicit, and assumed intention, on the other hand, creates a climate where adoption becomes possible.
Managers must avoid at all costs an overabundance of tools that creates more confusion than value. They must help teams distinguish between what is useful and what is noise, support professional change by enabling everyone to understand how their role is changing, and finally combat uncontrolled uses, because “shadow AI” creates a risk that undermines trust.
Implication: the managerial impact of AI is as much cognitive as it is operational.
The manager responsible for ensuring consistency between value, technology, and organization
The businesses that truly capture the value of AI, the famous 5% identified by BCG, are those where managers play a translating role. They link strategic intent, operational constraints, and technological capabilities, and bring consistency to the whole.
Managers must translate the business’s ambitions into concrete operating procedures. They must work with IT to build shared governance, because no transformation can succeed if technological choices are disconnected from business realities.
This is precisely where most organizations fail today, because they separate strategy, operations, and technology, when the value of AI depends on their articulation.
Implication: managers become the interface between intention, reality, and capability.
Bottom Line
AI requires a change in approach that goes beyond a new management model. Managers are shifting from control to work design, from decision-making power to the architecture of a hybrid decision-making system, from individual expertise to the orchestration of human-AI hybridization, from compliance to collective learning, and from hierarchical management to systemic consistency.
The bottom line is that the success of AI depends heavily on the manager, not the model, which is why the adoption of AI is a matter of business design before it is a matter of technology.
To answer your questions…
The McKinsey and BCG reports show that AI only creates value if work is redesigned. Without a clear purpose, AI automates vague processes and amplifies existing dysfunctions. Managers therefore play a central role in clarifying objectives, redefining workflows, and creating a framework of trust. AI then becomes a lever for improving the functioning of the business, rather than a simple technological add-on. Without this organizational design work, investments remain limited.
Managers are shifting from a supervisory role to a role as work designers. They no longer primarily monitor execution, but create the conditions in which work can be done well. They clarify the meaning of processes, identify operational friction, and organize the articulation between humans and AI. This evolution transforms the managerial function into one of consistency, context, and collective value creation rather than hierarchical supervision.
AI introduces hybrid decisions that are neither entirely human nor fully automated. Managers must define what AI can decide, recommend, or simply observe. These boundaries depend on risk, data quality, and team maturity. By setting thresholds and validation mechanisms, managers avoid opaque decisions and build an augmented decision-making system where humans intervene where their judgment is most useful.
Middle management is often the sticking point in AI transformations. Without an understanding of processes, work routines, and continuous improvement, it is impossible to integrate AI into everyday life. Middle managers must also be able to manage the risks associated with AI. If they do not evolve in their role and skills, they will slow down adoption by teams and prevent scaling.
Trust is based on clear and explicit intentions. Managers must state what AI should enable and what it should not become. They limit tool overload, combat parallel uses, and support the evolution of professional roles. By reassuring teams on a cognitive and emotional level, they create a climate where AI can be adopted without fear or circumvention.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
How can you prepare your organization for AI?
McKinsey identifies 6 key sucess factors for enterprise 2.0
Enterprise 2.0 and social business : what to expect in 2012 ?
AI in the workplace: going beyond augmentation to actually transform
To manage is to design
Adoption and impact of AI: lessons (and limitations) from the latest McKinsey and BCG studies
