• la différence cruciale ne tient pas à la nature publique ou privée de l’organisation, mais à sa soumission ou non à la concurrence.
• Le reporting a envahi le quotidien professionnel. Ce qui était hebdomadaire devient quotidien. Ce qui était mensuel devient hebdomadaire
• La numérisation, censée libérer du temps, a démultiplié les occasions de contrôle
• La transparence devient surveillance, la coordination devient empilement de validations. Les outils de productivité produisent plus de bureaucratie.
• Le temps consacré à justifier son travail excède souvent le temps de travail effectif.
• toutes les organisations, publiques ou privées, sont attaquées par le virus bureaucratique. Aucune n’est naturellement immunisée. La bureaucratie n’est pas l’apanage de l’État, elle guette toute structure dès qu’elle atteint une certaine taille
• . Pour les entreprises soumises à la concurrence, les alertes sont automatiques et compréhensibles par tous.
• Les réformes se mettent généralement en place pour éviter la disparition de l’entreprise.
• e problème devient beaucoup plus sérieux pour les organismes en situation de monopole.
• ucune pression concurrentielle ne venait perturber l’équilibre bureaucratique.
• ces organisations ne reçoivent aucun signal clair de dysfonctionnement. Les usagers peuvent se plaindre, les agents peuvent souffrir, mais le système continue
• L’ironie historique mérite d’être soulignée. Les entreprises privées dénonçaient la lourdeur administrative publique. Elles se présentaient comme l’antithèse du fonctionnement étatique. Pourtant, en grossissant, elles ont importé et parfois dépassé le secteur public en absurdité bureaucratique.
• Car même le privé concurrentiel n’est pas totalement immunisé. Quand une entreprise atteint une taille critique, elle crée des poches de monopole interne
• Plus de 70% des décisions importantes sont prises à des niveaux éloignés de l’opérationnel.
• Entre les deux se déploie une hiérarchie de relais. Chaque niveau transmet, traduit, déforme l’information.
• Le sur-encadrement accompagne cette fragmentation. Des managers qui managent d’autres managers qui supervisent des coordinateurs. La dilution des responsabilités atteint des sommets.
• L’exemple contemporain le plus frappant est celui de la numérisation. Elle devait simplifier, rationaliser, accélérer. Elle a démultiplié la bureaucratie en la rendant invisible.
• On évite la confrontation directe en ajoutant une règle. Personne ne discute du problème réel, on le transforme en process. Le système se blinde contre l’imprévu en standardisant jusqu’à l’absurde.
• La peur du face-à-face structure ces évitements.
• Au-delà des 4% du PIB mesurables, la bureaucratie génère des coûts invisibles plus profonds.
• La démobilisation d’abord.
• La résignation s’installe chez ceux qui ont tenté de changer les choses.
• L’énergie détournée du travail réel vers sa justification représente un gaspillage considérable
• On ne compte pas seulement les heures perdues mais aussi les initiatives non prises,
• Le sentiment d’impuissance qui en résulte affecte la santé mentale au travail
• L’Oréal a créé des « ambassadeurs digitaux » qui relient les départements en maîtrisant à la fois les données clients et les outils techniques. Ces figures transverses ont réduit de 30% le temps de coordination interservices entre 2021 et 2023.
• Berkshire Hathaway, le conglomérat dirigé par Warren Buffett, maintient son équipe de direction réduite à sa plus simple expression
• Haier, le géant chinois de l’électroménager, a développé une approche entrepreneuriale interne. Chaque unité fonctionne avec une grande autonomie basée sur des objectifs précis en matière de croissance rentable
• Le point commun de ces contre-exemples ? Ils réduisent les niveaux hiérarchiques, rapprochent la décision du terrain, créent des espaces d’autonomie réelle avec des responsabilités clairement identifiées.
• Ces organisations acceptent aussi une dose d’incertitude et d’imprévisibilité. Elles ne cherchent pas à tout standardiser, à tout prévoir, à éliminer tout risque par la règle
• Cela ne signifie pas l’absence de règles. Mais des règles en nombre limité, clairement justifiées, régulièrement questionnées.
• La bureaucratie au travail n’est pas une fatalité technique. Ce n’est pas le prix inévitable de la taille des organisations ou de la complexité des activités. C’est un équilibre culturel qui se maintient parce qu’il protège chacun de la « dépendance personnelle » – dépendre du jugement ou du bon vouloir d’une personne spécifique,
• ces solutions touchent au culturel, pas au technique. Elles nécessitent d’accepter l’inconfort de la « dépendance personnelle », de renoncer à la protection que procure la règle impersonnelle.
• Commencer petit, dans son périmètre : réduire une strate de validation, supprimer une réunion récurrente, remplacer un processus par une conversation directe. Rendre visible le coût : mesurer le temps passé à justifier versus faire. Créer des espaces protégés d’expérimentation où on suspend le reporting pendant quelques mois pour voir ce qui se passe.
• Changer de culture ne se décrète pas. Les tentatives individuelles se heurtent au système. Le changement culturel prend une génération, pas un trimestre
• Lucidité sur la difficulté réelle de transformer des équilibres culturels profondément ancrés. Refus du fatalisme parce que des marges existent, des brèches peuvent s’ouvrir,
• a transformation se fera par petites brèches plutôt que par grande réforme.
• Entre ceux qui pensent que le système est trop verrouillé pour changer et ceux qui croient encore à des marges de manœuvre, le débat continue.
• accepter passivement le statu quo garantit sa perpétuation.

37 more annotations...

• That AI will replace all of our jobs and we will live in this wonderful utopia. All by 2030, packaged and tied up neatly with a bow. It’s all rather lovely. It’s also entirely fantasy.
• , in this one I want to dive into the meaning of work, what it actually is and why AI isn’t going to replace everyone’s jobs
• these AI leaders went off madly in all directions with great proclamations and definitive statements.
• While these AI inventors are clever in many ways, they sort of fall short of a nickel in other ways. Like understanding how culture actually works, or how people might react when you tells them their life is essentially pointless
• Work is part of how we as humans make meaning. Work is part of our identity too. Social hierarchies and organization. If we take work away from us, it cuts into our identity. Work is also how we organize our sense of time…anybody here working for the weekend? For holidays and vacations, marking and organizing our years.
• Work too, is a social exchange. It’s part of how we feel we give back to society and as a result
• We also like to be good at things. A chef, a lawyer, an engineer, boat builder, carpenter. It is the felt application of experience
• Coders, AI giants and their leadership are smart people. Very smart. But that doesn’t mean wise
• Their entire identity (and net worth) is wrapped up in being invested in technology that they believe is world changing and historical. Their mindset is that everything and anything humans do can be reduced to code.
• Humans aren’t labour units. That’s just Taylorism on steroids. But every culture in human history, without exception, has understood work as embedded in webs of meaning, obligation, status, and identity that economics alone cannot capture.
• we cannot even eliminate jobs that every agrees are pointless. The compliance officer reviewing the policy written to comply with the policy reviewed by a different compliance officer. The meeting to plan the meeting. The consultant hired to recommend hiring more consultants.
• . Bureaucracies are an effective form of political system and we love us our political systems.
• . Every technology that was predicted to smooth out and flatten bureaucracy, from emails to ERP systems, the internet, well, ended up making them more complex
• So from all that above we have to also consider that structurally, such a wiping out of jobs is just not in the cards. We’d have to change the entire global tax base. In just a couple of years. Consumer demand would collapse, so not taxes. All the legal and regulatory frameworks around the world, would have to be redesigned. In just three years. And entire financial systems would have to be redesigned too. In just three years. This is at best suicidal. Eliminating all employment would send the world into a tailspin. To say the least.
• . The assumption of one dimensional thinking that we get from these AI tech giants is that you can simply remove “labour” and nothing else will be impacted. That’s like removing the bees and thinking the flowers will be fine.
• oday, massive civil unrest, the collapse of governments, all within less than 4 years. Not going to happen. Elected politicians and even dictators, prefer to stay in power. Bureaucracies are everywhere. The power of citizens is immense.
• Such structural changes at this scale take time. We will rethink the way we domesticate AI, as we have done with all technologies.

15 more annotations...

• Même si ça va vite, même si ça va “tanguer fort”, personne n'a envie de se faire disrupter et personne - au sens des entreprises ou des individus - n'est capable à grande échelle d'encaisser et de maîtriser les changements comme cela du jour au lendemain. Les forces de rappel et les freins à la diffusion dans les organisations sont trop puissants.
• Ne vous méprenez pas ! Ça va vite, et même très vite. Néanmoins, plus nous sommes proches d'une technologie ou d'une innovation, et c'est le cas chez OCTO, plus nous avons tendance à surestimer sa facilité de diffusion dans les organisations.
• a société dans son ensemble ne maîtrise ni les ressorts ni l’ampleur des bouleversements actuels. Nous avons parfois tendance à projeter nos propres inquiétudes sur le reste du monde, comme si chacun suivait ces évolutions avec le même niveau d’information que nous.
• Ensuite, on ne remplace pas du jour au lendemain des personnes par de l’IA au sein de processus anciens — parfois même archaïques
• L’automatisation à grande échelle est d’autant plus efficace que les processus sont explicites, répétables et modélisables
• nous avons besoin des humains pour donner “des coups de clés 12” ici et là.
• , la menace la plus immédiate ne réside peut-être pas dans le remplacement direct des emplois par l’IA, mais dans son instrumentalisation.
• L’IA sert alors de paravent technologique pour masquer des erreurs de gestion ou des choix stratégiques discutables.
• Pour intégrer efficacement l’IA dans ses processus — et en tirer de réels gains de productivité — encore faut-il accepter de les repenser en profondeur.
• Autrement dit, il faut revoir les processus et leurs organisations avant de chercher à les automatiser. Or transformer des processus au cœur des organisations prend du temps — parfois trop au goût du métier qui espère des gains immédiats.
• Nous pouvons nous accorder sur un point : à plus long terme, la transformation sera durable et profonde.
• Les cadres ont une assiette fiscale plus importante (ndlr : que les cols bleus) par personne (cotisations, charges, impôts). Chaque chômeur supplémentaire représente des cotisations en moins, des prestations en plus, et un système de retraite encore plus sous tension. (Enfin) … politiquement, c'est une masse qui vote plus.”
• : si les IA peuvent remplacer des cols blancs, elles ne consomment pas. Or ce sont bien les ménages qui soutiennent, directement ou indirectement nos économies.
• e politique pourrait être amené à intervenir. Sous quelle forme — régulation, fiscalité, mécanismes redistributifs, soutien à l’emploi, interdiction — la question reste ouverte
• un mécanisme de redistribution des richesses ou à un éventuel partage des gains de productivité lié à l'IA pourrait être réfléchi.
• Ça va vite, trop même à mon goût ;-)
• aucune entreprise n’a envie de subir une disruption brutale et aucune organisation n’est réellement capable d’absorber des transformations aussi profondes du jour au lendemain.
• Pour que l’IA ait de l’impact durable dans les processus métier, cela suppose de repenser ces processus et ça dans les organisations ça prend du temps.
• Si trop de cols blancs se retrouvent au chômage, il est possible que le politique entre en jeu pour limiter la casse sociales et économique.

17 more annotations...

• We find no correlation between the resulting measure of gen AI automation potential per sector and the development of sector-level profit margins since the launch of ChatGPT.
• We also found that this relationship is not mediated by differences in gen AI adoption levels across sectors
• In sum, even as companies pour resources into gen AI, the resulting productivity gains appear to be competed away, passed on to customers or suppliers rather than retained by firms–an early sign of commoditization.
• When technologies remove a bottleneck in one part of the value chain, they don’t eliminate scarcity, they relocate it.
• New technologies don’t just shift value pools along existing chains, they also create new ones. By removing constraints, they enable business models that were previously impossible.

• The Optimizer: Enhance existing capabilities.

   

  Often, companies are deploying gen AI technology in the hope of enhancing existing capabilities—becoming faster, more productive, or more responsive. Examples include digital publishers, such as BuzzFeed, embracing LLMs to develop content, or Bloomberg integrating gen AI into its terminal service to quickly summarize earnings calls or financial reports. Such use cases can unlock new growth potential or greater efficiency, fostering a competitive advantage for the firms that are quickest to effectively deploy new technologies.
• However, these advantages are likely to be only temporary because they rely on activities that are increasingly commoditized by gen AI technology.
• But in the longer term, as gen AI proliferates, these activities will be commoditized

• The Curator: Solve new frictions.

   

  Durable advantage is unlikely to come from deploying gen AI to do, faster and more efficiently, that which your firm is already doing. Rather, it will stem from shifting your focus to doing what has become harder to do with the emergence of gen AI technology.
• Focusing on a strategy of judgment or curation is not viable for every firm. Moreover, the underlying frictions may not persist indefinitely. Perhaps machine intelligence will, in the future, become very skilled at validating or selecting content

• The Reinventor: Build new business models.

   

  Entirely new offerings are made possible because gen AI makes it possible to remove constraints around costs or operational complexity.
• Thus, a useful way to identify potentially valuable deployments of gen AI is to evaluate how the technology might redouble existing advantages.
• since gen AI is particularly powerful at processing knowledge and making it more accessible, it may unlock new revenue streams associated with existing unique skills and capabilities that were previously locked inside the organization.
• s the technology evolves, new capabilities will emerge and new business models will become possible and valuable, just as existing ones may again become commoditized. One implication for executives is that they should not seek differentiation by looking for activities that cannot be accomplished by gen AI yet because the value pools associated with them will shrink as gen AI evolves.

• Repositioning for a Gen AI-powered Ecosystem

   

  The above strategies focus on deploying gen AI internally to enhance products or operations, with the goal of achieving margin expansion, greater speed, or scale. But there is another path, which involves a more external perspective: recognizing the changes gen AI will drive in the broader market, leaders can reposition their firm as a valuable node in this emerging ecosystem.
• it means adjusting existing or developing new value propositions for a reality in which gen AI becomes part of all market activities.

• The Re-packager: Rethink offerings for gen AI-powered choices.

   

  As choices are increasingly informed by —and may in the near future increasingly be made directly by—AI agents, firms must repackage their offerings for machine consumption
• Repositioning for agent-mediated choice is an important move and, as the above example illustrates, early movers are benefitting. However, in the long run, this adaptation will become table stakes, and may even drive commoditization

• The Supervisor: Validate gen AI’s accuracy.

   

  As gen AI models and autonomous agents proliferate and start to act as economic actors in their own right, there is an increasing need for appropriate supervision to enhance credibility and ensure reliability
• This “compliance-as-a-service” approach converts hard-won verification and regulatory expertise into a new income stream
• More broadly, any companies with proprietary, verified data, such as clinical trial results, financial analysis, or engineering manuals, could transform that knowledge into a validation service for LLMs,

• The Supplier: Provide input to gen AI.

   

  Finally, gen AI models also need input for training—an increasing need, as publicly available human-created text data is predicted to run out by 2028—as well as access to data that could be helpful in answering queries in real time.

• Leaders can identify the most valuable strategies for their company in the gen AI age by asking themselves the following set of questions:

   
   
  • Which of our current differentiating assets or capabilities are at risk of becoming commoditized by gen AI?
   
  • Which new frictions is gen AI technology causing in our markets or for our customers? Is solving them valuable? And, if so, how can we do it?
   
  • Where can we deploy gen AI to redouble existing advantages, for instance, by extracting new value from differentiated capabilities or assets?
   
  • What entirely new business models become possible if we remove historical cost or complexity constraints using gen AI?
   
  • Which of our assets could become inputs or infrastructure for the gen AI-driven business ecosystem?

21 more annotations...

• manipulation d’un modèle mental (ou « modèle du monde ») à petite échelle de la réalité. Pour Craik, le cerveau fonctionne comme une machine capable de traduire les événements externes en symboles, de les transformer par un processus d’inférence, puis de retraduire ces résultats en actions ou en prédictions.
• permettant à un organisme d’anticiper l’avenir sans s’exposer aux dangers immédiats du monde physique.
• Cette perspective définit la cognition humaine comme une économie d’effort et de risque. En possédant ce double interne du monde, l’être humain devient capable de tester des hypothèses sans avoir besoin d’agir.
• Il a montré que nos erreurs de logique ne sont pas des pannes de l’intelligence, mais des limites de notre mémoire de travail : nous échouons lorsque nous ne parvenons pas à simuler tous les modèles alternatifs d’une situation donnée
• cette absence de modèle interne est ce qui limite les IA actuelles basées sur l’apprentissage via des modèles de langage (LLM). Elles manquent de « sens commun ». En substance, le langage seul ne suffit pas pour comprendre le monde.
• ’évolution vers ces modèles permet également de surmonter le fossé entre la donnée brute et la compréhension réelle.
• Comment créer ces modèles sachant qu’il faudrait que ce soit automatique pour être efficace? Un modèle est par ailleurs presque par définition subjectif. S’il est facile de coder la loi de gravité de façon causale, les modèles économiques, par exemple, ne seront pas aussi si
• Une machine peut-elle acquérir une intuition causale véritable sans posséder un corps pour éprouver la réalité physique ? Damasio répondrait par la négative car pour lui, pas de conscience sans corps. Heureusement, envisager un progrès notable grâce aux modèles ne nécessite pas une réponse préalable à cette question.
• Pendant longtemps, le paradigme était le système expert, qui supposait que la connaissance était formalisable sous forme de règles explicites:
• Ce n’est pourtant pas exact et ce paradigme a conduit à un échec presque total de l’IA à la fin des années 90
• Passer à l’utilisation de modèles mentaux représenterait un 3e paradigme, une étape supplémentaire majeure

9 more annotations...

• But do you remember what never got shared, ever? The end-of-year performance reviews. Nor the salaries. Nor the bonuses. And there was definitely no agile team in sight when colleagues competed for a promotion.
• But guess who was the only one suffering when the product didn’t sell? When the only ones picking it up were three friends and two former colleagues? You, again. All other “team members” got decently paid and moved on to the next job.
• Teams Are More Productive -
• Teams Are More Innovative -
• Teams Are More Daring
• Teams Are More Effective
• Teams Are More Intelligent
• All these pluses of teamwork are benefits for the organization, not for the team member.
• People Don’t Contribute Equally
• People Point at Each Other
• People Get into Fights
• Some People Drain Everyone
• Alignment Eats the Work
• it’s the team members who usually suffer more from these drawbacks than the organization.
• The benefits of teamwork accrue mostly to the organization, while the drawbacks of teamwork are carried by the team members.

• A group is a collection of people. A team is a group with a shared goal.
• The goals of individual group members can be similar, but they work independently
• A team, however, shares a goal that requires interdependence.
• There’s coordination, shared accountability, and usually some degree of collective identity. If one person disappears, the whole thing starts to wobble and might even collapse
• The Group Feels Safe -
• The Group Feels Supportive
• The Group Feels Empowering
• Teamwork doesn’t exist among monkey species. In the animal kingdom, unrelated monkeys don’t work toward a shared goal unless there is some form of individual payoff.

• Humans are monkeys. We’re wired for social groups. We’re not wired for selfless teamwork.
• My happiest days are those when I have an empty calendar and my only conversations are with my spouse, friends, customers, and the local barista.

• Teamwork comes with a tax, and someone should pay the tax.
• In the case of teamwork, the argument “you will enjoy the social bonding of the team” isn’t a valid sales pitch. It’s a really bad value proposition. A professional doesn’t need the stress of teamwork to enjoy the benefits of being a social creature. They might as well go to the local bar. That’s what monkeys do and what social groups are for: all the benefits without the hassle of teamwork.

• Managers need a better pitch for teamwork, or else it’s just a scam.
• Preferring to do things alone and refusing to pay the team tax was by many in the community almost seen as heresy.
• And so we put up with all the endless meetings. We wrote books full of practices to make teamwork manageable.
• However, forgive me if I’m mistaken, but “individuals and interactions” doesn’t sound to me as being the same as “team.”
• I prefer going solo, but I fully understand that many products require teamwork.
• hey are the CXO of Themselves. And in the age of AI, with digital assistants replacing many human teammates, there will be more Solo Chiefs than ever.

• The Solo Chief doesn’t carry coffins. The Solo Chief writes the eulogies.

32 more annotations...

• I've been building a custom Lean Hospitals AI Chat — an AI assistant grounded in the principles from my book and from Toyota's management system. It has features you won't get from a generic chatbot: responses tailored to your role (frontline staff, manager, director, C-suite, consultant), suggested follow-up questions, guided prompt packs, and the ability to switch between coaching and fact-based modes.
• So I started testing. I asked the same question to base ChatGPT (the free version anyone can use) and to my custom tool, and compared the results.

• The difference was striking — and it reveals something important about the risks of using generic AI tools for Lean guidance.

      
• Lean thinking flips this. Instead of cutting people to reduce cost, you engage people in improving Safety, Quality, and Delivery. When those improve, Cost follows
• This is also why the most successful Lean organizations pair their improvement work with an explicit “no layoffs due to Lean” commitment.
• So: the question “How do I use Lean to reduce headcount?” isn't just strategically misguided. It's the kind of question that, if answered badly, can set a hospital's improvement culture back years.

• It opened well:

      

   

  “Using Lean in a hospital isn't about cutting people first — it's about removing waste so staff can focus on patient care.”
• If you approach it as ‘reduce headcount,' you'll likely damage morale, quality, and safety. If you approach it as ‘eliminate waste and redeploy talent,' you can often lower labor costs naturally through attrition, productivity gains, and smarter staffing.”
• To be fair, ChatGPT did say “don't announce Lean as a cost-cutting layoff program.” That's correct. But it said it while building a detailed plan to use Lean as a cost-cutting program that ends in headcount reduction through “attrition capture.”
• The book is very clear on this: Lean should not be used primarily as a headcount reduction tool. That approach goes against the principle of Respect for People. In Chapter 1, it's emphasized that the main goal of Lean is to improve flow, quality, safety, and value for patients–not to cut jobs.
• If leaders position Lean as a way to “do more with fewer people,” staff quickly learn not to participate fully, fearing job loss. This fear destroys psychological safety and undermines continuous improvement.
• So, if your real goal is cost savings, Lean can help, but only as a byproduct of better processes–not from cutting heads directly. Sustained improvement happens when staff feel safe, respected, and engaged, not when they fear being replaced by Lean initiatives.
• What often happens is that as you improve processes — making care safer, reducing waste, smoothing flow — you'll find staff spending less time on rework, searching, and firefighting. The question then becomes, “How do we use this freed-up capacity?”
• If your hospital has a mandate to reduce labor costs, it's better to address that openly with your teams, not cloak it as Lean
• Something to try: Ask your teams where waste frustrates them most. See what happens if you direct Lean energy there and focus on making their work more rewarding.
• A: That's a question that comes up a lot, Mark, but it's really important to pause and reflect on the intent behind Lean and its impact on people.
• Lean isn't about reducing headcount or “eliminating jobs.” At its core, Lean focuses on improving patient care by engaging staff to solve problems and make work less frustrating and more effective
• • Base ChatGPT: Here's how to do the wrong thing, in detail
      
  •  Book Mode: Here's why that's the wrong thing, with evidence
      
  •  General Mode: Here's why that's the wrong thing, and what's really going on
      
  •  Coach Mode: Here's why that's the wrong thing, and let me help you think through what to do instead
• Discovering this small gap — not mentioning “no layoffs due to Lean” — allowed me to update the instructions and guardrails in my custom AI

17 more annotations...