• L’AI Manager Club, présenté comme le premier réseau national des responsables IA en entreprise, sera inauguré le 6 novembre à Paris.
• sur plus de 2 500 profils LinkedIn identifiés comme actifs dans l’IA.
• La création du club répond clairement au besoin d’accompagner une fonction émergente, souvent isolée, appelée à structurer les usages de l’IA au sein des organisations
• Pourquoi limiter la communauté à ces seuls 1 000 professionnels identifiés, sur des fonctions très techniques ou stratégiques ?
• créer un entre-soi élitiste, là où le véritable défi réside dans la diffusion des usages auprès des managers de proximité et des métiers opérationnels.
• mais former les managers tout court – RH, commerciaux, chefs de projet, responsables qualité, DAF – devient aujourd’hui indispensable
• Ce biais structurel, concentrer l’effort sur une élite technologique de l’IA, peut nuire à l’adoption généralisée
• Or les freins se situent souvent plus bas dans l’organigramme, comme la méconnaissance des cas d’usage, la peur de l’obsolescence, la résistance culturelle, l’absence de support dans les décisions opérationnelles.
• pour catalyser une adoption réellement organisationnelle de l’IA, l’effort devra s’élargir.

7 more annotations...

• Inspired, Curious, and Unbossed, isn’t a set of slogans but a daily guide for how work gets done.

      
   
  • Inspired means connecting everything back to patient impact.
      
  • Curious means learning continuously, across functions and roles.
      
  • Unbossed means trusting people to take ownership rather than managing through control.
   
      
• Novartis created behaviors in action—specific examples that help employees discuss, observe, and measure how culture shows up.
• t’s not just about why people join, but why they stay,”
• “Let your team do their work because they know their work better than everybody.”

• Leadership at Novartis now centers on three expectations:

      
   
  1. Lead and develop people.
      
  2. Navigate complexity.
      
  3. Deliver collective impact.
• “Thirty percent of what you do should be leading your people,”

• Data Starts the Conversation

      

  While AI tools are making leadership analytics more sophisticated, Rob warns CHROs against letting data replace human judgment. Metrics should start the conversation, not end it.
• the real value comes from what happens next: dialogue, reflection, and action.
• a blueprint for where HR is heading: not toward automation for its own sake, but toward augmentation
• the future of leadership belongs to those who make work more human, not less.

8 more annotations...

• Work is a way all of us keep roofs over our heads, offers social mobility and other opportunities and for many (and dare I say this?) in my experience how especially men define their purpose in life. What will happen when that changes and are we ready?
• One week, headlines announce mass layoffs, the next, companies are rehiring after being let go (Klarna and IBM to name but two).
• an economy where machines outperform humans not only in physical labour, but also in cognitive and creative tasks.
• . Professions and career paths as we know them across many sectors are predicted to disappear and some already have.
• Elon Musk recently said, “AI and robots will replace all jobs. Working will be optional, like growing your own vegetables instead of buying them from the store.”
• echoes Karl Marx’s utopian vision of post-labour freedom: “To hunt in the morning, fish in the afternoon, rear cattle in the evening, and criticise after dinner.”
• Automation and AI aren’t just changing the nature of work; they’re changing how we think about life.
• too many leaders seem to be taking a “FAFO” approach (F** Around and Find Out) rather than developing real strategies for a post-work world
• . In the United States, with its tech concentration and lighter employment protection structures, the the shift is already accelerating
• Some forecasts say by the 2040s we will probably see work becoming something people choose rather than need
• which will require safety nets on a wide scale
• , much to the frustration of the US tech sector, Europe, has a slower rate of transition: lower productivity growth, but stronger social systems mean it is estimated that its version of the post-work future will likely unfold between the 2040s and 2050s.

• 1. Tier One: The Tech Bros

  At the top we will have Tech Barons, “deviced- up,” sporting wearables from head to toe, living on adrenaline until there is a power outage or a drop in network signal. Scott Galloway warns that while technology creates immense wealth, it tends to concentrate that wealth in fewer hands, making us fragile.
• 2. Tier Two: Resilient Workers
• Highly-skilled technologists: these will be top level engineers who design, programme and maintain the AI and robotics systems.
• Supervisors of machines, humans giving oversight to these systems monitoring and managing
• AI resilient professionals: artists, caregivers, educators, therapists, performing artists, crafts people, skilled trades etc, the kind of AI resilient work that requires empathy, imagination, and physical human presence.
• Ethicists, policymakers, governance and social architects who ensure automation aligns with human values

• 3. Tier Three: The Displaced

  The dumped. Potentially millions (or billions) of people displaced by automation, living in an economy where working is optional or opportunities are scarce. This group will define the true human cost of AI-driven progress.

• Our job is to stop that divide from becoming permanent.

  That means:

  • Building AI literacy across the workforce, not just in the IT department.

  • Identifying future skills early and creating pathways for employees to evolve alongside technology. This should not be based on age and we will need the roles of the CLOs (Chief Longevity Officers) with strong policies for managing multi-generational workforces with inbuilt Longevity Literacy, not just AI literacy

  • Redefining performance metrics to include adaptability, ethical judgement, and collaboration with AI systems.

  • Designing trauma-informed offboarding, replacing today’s brutal mass layoffs with humane transitions that preserve dignity and possibility.

  • Lifelong purpose-building: which is possibly a new concept for redundant employees, preparing them for a world of No work.
• We can’t afford “AI slop, poorly trained systems or poorly trained people.
• We need strategies for a job-less world and I wonder how far along we are with this.
• Do we finally see the rise of Universal Basic Income funded by the very corporations profiting from automation?
• What happens when we lose the sense of purpose that work provided?
• Loneliness , (already a feature of our current workplaces) and loss of purpose have already been cited as key concerns.
• “The greatest threat of AI, hands down, in my view, is loneliness. I think we are on the precipice of disaster, combining synthetic relationships and AI”.
• The workless future isn’t science fiction anymore. It’s coming fast and we need to get ready. If work once gave us purpose, perhaps the next great task is to learn how to build meaning without it.

• The future of work is being rewritten but people must stay at the centre of the story.

27 more annotations...

• How to Reduce the Gap Between Knowing and Doing

   
  1. Start  Small, Start Now
   
   
  • Break  large goals into quick, low-risk actions you can take immediately.
   
  • Example:  Instead of analyzing a process problem for weeks, run a quick trial  solution on one workstation.
   
   
  2. Make  It Safe to Try—and Fail
   
   
  • Reward  initiative, not just outcomes.
   
  • Treat  mistakes as learning investments, not career-limiting moves.
   
   
  3. Set  Short Feedback Loops
   
   
  • Replace  long, drawn-out project cycles with rapid check-ins and adjustments.
   
  • Quick  learning cycles make it easier to see progress and maintain momentum.
   
   
  4. Measure  Action, Not Just Ideas
   
   
  • Track  “execution metrics” like number of experiments run, pilot projects  launched, or issues resolved—not just meetings held or plans made.
   
   
  5. Recognize  and Celebrate Movers
   
   
  • Publicly  acknowledge employees who move projects forward, even in small ways.
   
  • Stories  of action create peer pressure to act.
   
   
  6. Simplify  the First Step
   
   
  • Remove  unnecessary approvals, overly complex templates, or ambiguous ownership  that slow down action.
   
   
  7. Model  It From the Top
   
   
  • Leaders  must be the first to move from idea to test. When the team sees action  modeled at the top, it becomes part of the culture.

• D’un côté, l’explosion de l’infrastructure et ses conséquences géopolitiques (data, énergie, eau…), et de l’autre, l’objet de fascination que représente l’espace latent (relire ma newsletter).

• la majorité de l’argent levé, prêté ou échangé s’est fait dans l’infrastructure.
• une partie du “seed money” de Microsoft à OpenAI s’est faite en crédits Azure pour le compute.
• Le crédit en compute est fascinant car il attache la destinée d’un acteur à un opérateur de cloud et d’IA spécifique
• Nvidia a une valorisation en termes de ratio P/E plus basse qu’il y a cinq ans. C’est-à-dire que ses bénéfices ont rattrapé son prix.
• La bulle des télécoms était surtout une bulle de la fibre noire. Bien qu’on ait surinvesti et surinstallé à la fin des années 1990, ce réseau de câbles était tout à fait utilisable des années après, quand la demande est arrivée.
• Aujourd’hui, le surachat en puces Nvidia, dont la durée d’amortissement est très faible, interroge. De plus, comme tout produit, dès qu’on l’ouvre pour l’installer (même si les puces Nvidia sont livrées par palettes déjà préconfigurées, un cloud dans le cloud, NDR), elles perdent une partie de leur valeur.
• C’est donc là que se pose la vraie question : l’investissement colossal en puces doit être amorti sur une durée courte, un an, deux ans, trois ans, cinq ans. Est-ce possible ?
• le deal entre OpenAI et Oracle (ainsi qu’avec AMD) s’est fait hors des circuits classiques des banques d’affaires. Les accords récents d’OpenAI (avec Oracle, AMD, CoreWeave…) sont des deals de CAPEX, pas des deals produits ou de financement traditionnels, et s’apparentent davantage à des échanges de compute, de crédits et de stock-options
• C’est sur cette base qu’OpenAI aurait signé ses engagements faramineux (jusqu’à 1 000 milliards $ d’infrastructure) sans disposer de revenus correspondants,
• • L’offre pléthorique et la capacité de compute n’ont jamais été aussi importantes dans l’histoire de l’humanité.
  • Mais pour que ça marche, il faut de l’eau, de l’eau douce évidemment, qu’il faut prélever aux autres industries et aux usages quotidiens, mais aussi de l’énergie.
• a véritable raison pour laquelle ces acteurs avaient besoin de lui n’était pas l’argent — qu’ils ont déjà —, ni le pouvoir — qu’ils ont déjà —, mais bien l’accès à l’énergie et une accélération des process par le gouvernement fédéral américain pour permettre la construction de centrales nucléaires
• le boom de l’IA aux États-Unis offre à Trump une véritable couverture politique : les retraites des Américains (et celles par capitalisation des Français) restent, pour l’instant, garanties par la flambée du NASDAQ.
• sans la dépense massive liée à l’intelligence artificielle — en particulier dans les data centers, les puces et le cloud —, l’économie américaine serait aujourd’hui proche de la récession, voire déjà en récession.
• En 2025, les investissements dans l’IA devraient représenter près de 40 % de la contribution à la croissance du PIB américain
• ce boom n’est pas soutenable : il ne pourra se poursuivre que si les dépenses technologiques deviennent « paraboliques », un scénario jugé hautement improbable
• si le CAPEX IA ralentit, la dynamique de soutien à la croissance s’essoufflera rapidement.
• Tech et administration américaines se tiennent désormais par la barbichette. La question qui se pose, c’est de savoir si la viabilité de l’économie américaine et, par extension, celle du reste du monde ne repose pas entièrement sur un pari sans plan B :
• Pourquoi vivons-nous dans un monde où, le même jour, OpenAI annonce travailler sur des chats érotiques, autrement dit des sexbots, tandis que Google et l’université de Yale utilisent une IA pour découvrir une nouvelle méthode permettant au corps de mieux repérer et attaquer certaines tumeurs cancéreuses ?
• les revenus des abonnements en Europe sont en baisse.
• il reste difficile de cerner précisément ses usages monétisables.

19 more annotations...

• At the same time, only 11% of organizations say they can currently enable effective co-learning, according to the report.
• “Closing the readiness gap isn’t just about adopting new technologies. It’s about building organizations where people and AI can thrive together,”

• Redesign roles and workflows for human-AI collaboration

   

  Close explains that companies need to rethink jobs, feedback loops and processes so people and AI can work seamlessly together.

• Embed learning into everyday work

   

  According to Close, future-ready organizations make skill development continuous, not an afterthought. More than 80% of execs, she says, believe learning is most effective when it’s built into employees’ everyday activities

• Design AI tools around people

   

  AI tools should fit naturally into employees’ workflows, Close says, adding that this strategy involves making AI work the way people work,

• Hardwire trust and accountability

   

  According to Close, clear governance and transparency are essential for adoption. She points out that nearly three-quarters of workers surveyed say they are more willing to embrace AI when governance is clear and transparent

• Lead with curiosity and creativity

   

  Leaders must model experimentation and innovation, encouraging teams to question assumptions and explore new ideas. Along those lines, she adds, more than 80% of executives told Accenture they are focused on building a culture of curiosity

5 more annotations...

• Agentic AI systems can autonomously manage complex tasks, optimize processes, and proactively identify opportunities or risks, reducing the need for constant human oversight.

• According to Pitchbook data there was a 265% increase in venture capital investment in agentic AI between Q4 2024 and Q1 2025.

   

• Gartner predicts that by the end of 2027, more than 40% of agentic AI projects will be canceled.

   
• The main reasons include escalating costs, unclear business value, and inadequate risk controls.
• Gartner estimates that the market is still immature, with companies deploying either AI assistants or not having enough opportunities to use the technology to begin with.
• Business leaders must resist the temptation to deploy agentic AI indiscriminately and instead focus on use cases where agentic AI’s unique capabilities create measurable business value.

• Why Most Agentic AI Projects Fail

   

  The primary reason for failure is misalignment between the technology’s capabilities and the business problem at hand.
• Many deployments are little more than advanced chatbots or robot-process automation (RPA) tools with a conversational interface.

• Identifying the Right Use Cases for Agentic AI

   

  To avoid these pitfalls, Gartner recommends that business leaders first evaluate where agentic AI is truly the best fit. Not every workflow or customer pain point requires the autonomy and complexity of agentic AI.
• • Complex, dynamic environments: Agentic AI excels where tasks require ongoing adaptation, decision-making under uncertainty, and the ability to handle exceptions. Examples include supply-chain optimization, cybersecurity threat response, and dynamic credit approvals.
   
  • Multi-step, multi-agent collaboration: Workflows that span multiple systems, teams, or business functions, such as orchestrating end-to-end procurement or managing cross-functional incident response, are well-suited to agentic AI’s strengths.
   
  • High-value, low-frequency decisions: In scenarios where the stakes are high and the cost of human error is significant, agentic AI can deliver measurable improvements in quality, speed, and scale. Examples include warranty claims and hospital discharge.
   
   
• Conversely, use cases that are highly structured, repetitive, or well-served by existing automation should not be forced into the agentic AI mold.

• Making the Business Case for Agentic AI

   

  Leaders can evaluate the business value of an agentic AI implementation for a sustainable competitive advantage by considering the cost savings, quality, deployment speed, and scale.

• 1. Develop a strategic technology roadmap.

   

  Prioritize outcomes that enhance measurable efficiencies—speed, cost savings, scale, and quality—rather than chasing the latest AI trend.

• 2. Cut through the hype with composite AI.

   

  Rather than relying solely on agentic AI, explore composite solutions that combine multiple AI techniques such as machine learning, symbolic reasoning, and traditional automation to address customer pain points.

• 3. Conduct a rigorous cost-benefit analysis.

   

  Before greenlighting any agentic AI project, estimate potential returns by considering both direct and indirect benefits, such as cost savings, revenue growth, and customer satisfaction

• 4. Evaluate feasibility and readiness.

   

  Assess the technical and organizational readiness for agentic AI. Do you have AI-ready data? Are workflows sufficiently mature and documented? Are risk controls in place?

• 5. Focus on enterprise productivity, not just user productivity.

   

  The greatest value from agentic AI comes from orchestrating actions across siloed applications and business units, not simply improving individual user experiences.

• The market will not reward those who pursue agentic AI for its own sake or succumb to vendor hype. Instead, sustainable competitive advantage will accrue to leaders who do the following:

   
   
  • Select use cases where agentic AI’s unique capabilities deliver outsized business value.
   
  • Integrate agentic AI thoughtfully into broader technology portfolios.
   
  • Continuously measure and communicate ROI to stakeholders.

16 more annotations...