Category: Leadership

How decisions become motion — workforce, organizational design, execution speed.

Browse: Workforce · Org Design · Execution

  • The Return of Industrial Time

    The Return of Industrial Time

    For the last two decades, a lot of management culture has learned to think in software time.

    Build. Test. Release. Measure. Iterate.

    That operating logic changed how companies build products, how teams organize work, and how boards talk about speed. It made experimentation respectable in places that used to reward only long planning cycles.

    My read on this: that lesson is still useful, but it is no longer enough.

    A growing part of the strategic agenda is not moving on software time. Electricity grids, energy systems, ports, factories, semiconductor supply chains, defense production, railway capacity, industrial permitting, and resilient sourcing all run on a different clock.

    They require capital before certainty arrives. They depend on permits, suppliers, safety, skills, land, regulation, maintenance discipline, and long-term demand signals. They take years to build and decades to amortize.

    This is the return of industrial time.

    The interesting leadership problem is not choosing between speed and patience. It is knowing which clock a decision belongs to.

    The software clock changed executive expectations

    Software gave leaders a powerful idea: speed can reduce risk.

    If a team can release a small version quickly, observe real behavior, and adjust, it does not need to pretend that every answer is known upfront. That logic has shaped far more than product development. It influenced strategy processes, innovation portfolios, transformation programs, and investor communication.

    The software clock is visible in how companies now talk about pilots, minimum viable products, agile delivery, platform thinking, data loops, and continuous improvement.

    I think that mindset still has enormous value. Faster feedback improves capital allocation. Faster decision loops reduce internal friction. Better data can reveal what customers, suppliers, and employees are actually doing, not only what the organization hopes they are doing.

    But the software clock also creates a temptation: the belief that every important problem can be de-risked through rapid iteration.

    That belief breaks down when the strategic problem is physical.

    You cannot A/B test a power grid in the same way you test a landing page. You cannot scale a defense-industrial base with the same reversibility as a software feature. You cannot rebuild semiconductor resilience quarter by quarter. You cannot fix underinvestment in infrastructure with a sprint review.

    Industrial systems can and should become more digital, more transparent, and more adaptive. But their underlying constraints remain material. Increasingly, they come with a price tag and a lead time that no roadmap can compress.

    Industrial time is slower because reality is harder

    Power grid control room overlooking high-voltage transmission lines at sunrise
    Industrial time is slow because physical capacity, permits and infrastructure cannot be compressed into software cycles.

    Industrial time is not slow because managers are old-fashioned. It is slow because the work sits inside physical, financial, and institutional constraints.

    Three numbers make the point.

    Grids. The International Energy Agency has warned that grids risk becoming the weak link in the energy transition unless investment accelerates. Its grid report says annual grid investment needs to double to more than USD 600 billion by 2030, and new transmission lines routinely take 5 to 15 years to plan, permit, and complete. IEA Executive Director Fatih Birol put it bluntly: "We must invest in grids today or face gridlock tomorrow." In the United States, the Department of Energy's National Transmission Needs Study estimates the country must more than double regional transmission capacity by 2035. That is not a communications problem. It is a capacity problem.

    Europe's investment gap. Mario Draghi's report on European competitiveness matters because it turns a familiar policy debate into an industrial-time problem. Its headline figure – roughly EUR 750-800 billion of additional investment per year – is not just a financing number. It is a statement about the scale of energy, defense, deep tech, infrastructure, and productivity capacity Europe would have to build. The report's core message is that Europe needs a different growth trajectory, not just better language around competitiveness. That lands as a management signal as much as a policy one.

    Semiconductors. A chip ecosystem is not one factory. It is design capability, advanced tools, specialty chemicals, materials, packaging, testing, energy, talent, customers, and export-control exposure. The CHIPS Act logic itself reflects this: the United States put USD 52.7 billion behind domestic semiconductor manufacturing and research because capacity is a multi-year industrial problem. TSMC's Arizona build-out, which began as a USD 12 billion project and later expanded, is now reported as a USD 165 billion U.S. investment. In mid-2026, TSMC CEO C.C. Wei told shareholders it would be "a long time before we can meet customer demand".

    Advanced semiconductor fabrication campus with clean industrial equipment, logistics docks and power infrastructure
    Semiconductor capacity is an ecosystem of tools, materials, energy, talent and long ramp-up times.

    The same pattern appears in defense. Europe can announce higher defense ambitions quickly, but ammunition output, supplier depth, testing capacity, skilled labor, and common procurement cannot be improvised. NATO's Jens Stoltenberg described the need to "shift from the slow pace of peacetime, to the high-tempo production demanded by conflict". That is industrial time in one sentence.

    The binding constraint is no longer the speed of the interface. It is the speed at which physical capacity, capital, skills, and permits can be brought into being.

    What this looks like inside companies

    The point becomes clearer when you look at company cases.

    Ford's electric-vehicle build-out is one example. A product with heavy software content still depends on battery plants, cell production, equipment orders, supply chains, trained workers, and industrial ramp-up. Ford described BlueOval City as part of its more-than-USD-30-billion EV investment through 2025. That is not a quarterly optimization exercise. It is a multi-year industrial bet.

    Orsted is another. The company took an impairment of roughly USD 4 billion in 2023 and cancelled its Ocean Wind 1 and 2 projects in New Jersey after supply-chain inflation, higher interest rates, and permitting delays made fixed-price contracts uneconomic. CEO Mads Nipper pointed to "significant adverse developments" in the supply chain and said the company was "extremely disappointed" to cease the projects. The deeper point is that industrial-time projects front-load commitment, then absorb the variance of a multi-year supply chain.

    Boeing shows a different version of the same issue. After the January 2024 737 MAX door-plug blowout, the FAA blocked Boeing from expanding 737 MAX production until quality systems were fixed. Demand was not the bottleneck. Industrial integrity was.

    And TSMC's Arizona expansion shows why industrial capability cannot simply be copied from one geography to another. The company has had to manage cost and timeline pressure in the United States, with reporting around TSMC's Arizona build-out pointing to substantially higher U.S. construction costs than in Taiwan. A fab is not just a building. It is an ecosystem.

    These are not failures of intelligence. They are encounters with a clock that does not negotiate.

    The harder management problem: two clocks, one company

    I do not think the answer is to become slower.

    The harder task is integration.

    A company that only thinks in industrial time becomes too slow. It over-plans, protects legacy processes, and treats every decision as irreversible. It may preserve reliability, but it loses learning velocity.

    A company that only thinks in software time becomes careless. It mistakes optionality for strategy. It launches too many pilots, underestimates physical dependencies, and treats capital-intensive systems as if they can be refactored later without cost.

    The way I see it, modern leadership needs both disciplines.

    Digital speed matters where reversibility is high and learning is valuable: customer insight, forecasting, demand sensing, workflow automation, internal transparency, scenario modeling, and decision support.

    Industrial patience matters where reversibility is low and execution risk compounds: plants, grids, logistics nodes, critical suppliers, regulatory approvals, safety systems, and long-lived assets.

    The mistake is applying the wrong rhythm to the wrong problem.

    Capital allocation becomes the test

    Executive strategy room with industrial infrastructure model, digital dashboard, hourglass and analog clock
    The real management test is whether capital, skills and capacity line up before the next shock arrives.

    Industrial time turns strategy into a capital-allocation test.

    It is easy to endorse resilience in a board presentation. It is harder to fund redundant capacity, dual sourcing, inventory buffers, grid connections, cybersecurity hardening, supplier development, and workforce training before the next disruption makes the need obvious.

    The same is true at national scale. The Draghi investment gap and the IEA grid investment number describe the same uncomfortable truth: agreement does not build capacity. Capacity follows from committed capital, credible timelines, aligned incentives, and operational ownership.

    The question I would be asking myself is simple:

    Where are we pretending that a strategic dependency is only an operating cost?

    If energy availability can constrain growth, it is strategic. If a supplier bottleneck can stop production, it is strategic. If a missing skill base can delay execution for years, it is strategic. If regulatory approval, grid access, or logistics capacity determines market entry, it is strategic.

    Industrial time makes these dependencies visible.

    It also changes the meaning of efficiency. In software time, efficiency often means reducing waste, shortening cycles, and automating repetitive work. In industrial time, efficiency also means keeping enough capacity, redundancy, and competence to survive stress.

    A system optimized only for the normal case can be financially elegant and strategically fragile.

    Andreas's view

    My read on this: the next advantage is temporal discipline.

    The companies that do this well will not become nostalgic industrial planners. They will still use digital tools aggressively. They will use better forecasting, better data, better scenario models, and faster feedback loops to make long-cycle decisions less political and less blind.

    But they will also recognize that some commitments have to be made before certainty arrives.

    I don't think the next decade rewards organizations that simply move fast. It rewards organizations that know when speed is a learning tool and when early commitment is the real advantage.

    Three things I'm watching:

    • Whether Europe can turn the Draghi diagnosis into actual capacity: energy, defense, capital markets, compute, and industrial execution.
    • Whether AI infrastructure pushes grid access, power contracts, cooling, chips, and data-center permitting into the center of corporate strategy.
    • Whether companies start treating suppliers, energy, skills, and resilience as strategic assets rather than procurement line items.

    The telling indicator will be whether management teams can hold both clocks in their head at the same time.

    Move fast where learning is cheap. Commit early where capacity will be scarce. Use data to shorten decision cycles, but respect the physics of assets, infrastructure, and institutions.

    The world is becoming more digital and more industrial at the same time.

    That is the leadership rhythm I think matters now.

    Sources

    https://commission.europa.eu/topics/competitiveness/draghi-report_en

    https://www.iea.org/reports/electricity-grids-and-secure-energy-transitions

    https://www.iea.org/news/lack-of-ambition-and-attention-risks-making-electricity-grids-the-weak-link-in-clean-energy-transitions

    https://www.energy.gov/oe/national-transmission-needs-study

    https://www.semiconductors.org/chips/

    https://pr.tsmc.com/english/news/3210

    https://www.cnbc.com/2025/03/03/tsmc-to-announce-100-billion-investment-in-us-chip-plants.html

    https://www.tomshardware.com/tech-industry/semiconductors/tsmc-ceo-c-c-wei-says-it-will-be-a-long-time-before-we-can-meet-customer-demand-tells-shareholders-that-he-will-keep-prices-stable-refrain-from-implementing-price-hikes

    https://9to5mac.com/2023/08/04/us-made-tsmc-chips/

    https://corporate.ford.com/articles/electrification/blue-oval-city/www/

    https://www.cnbc.com/2023/11/01/orsted-axes-two-new-jersey-wind-projects-takes-4-billion-writedown.html

    https://www.faa.gov/newsroom/faa-halts-boeing-max-production-expansion-improve-quality-control-also-lays-out-extensive

    https://www.nato.int/en/news-and-events/events/transcripts/2024/02/15/press-conference

    • European Commission: The Draghi report on the future of European competitiveness
    • International Energy Agency: Electricity Grids and Secure Energy Transitions
    • International Energy Agency: "Invest in grids today or face gridlock tomorrow"
    • US Department of Energy: National Transmission Needs Study
    • Semiconductor Industry Association: CHIPS Act overview
    • TSMC: U.S. investment expanded to USD 165 billion
    • CNBC: TSMC total U.S. investment reported at USD 165 billion
    • Tom's Hardware: TSMC CEO C.C. Wei on customer demand
    • 9to5Mac / NYT summary: TSMC Arizona construction-cost premium
    • Ford: BlueOval City and EV investment
    • CNBC: Orsted offshore wind impairment and cancellations
    • FAA: Boeing 737 MAX production expansion halted
    • NATO: Defense industrial production remarks
  • Agentic AI Is About to Leave the Screen

    Agentic AI Is About to Leave the Screen

    Why robotics may become the next operating layer for AI, and what changes when the physical world becomes programmable.

    Andreas's view

    My read: robotics is still being framed as hardware, when the more important shift is that AI is becoming an operating layer for physical work.

    The market tends to split into two shallow stories. One treats robots as factory equipment. The other treats humanoids as spectacle: impressive demos, big forecasts, uncertain timelines.

    I don't think either framing is enough. The more interesting story is that agentic AI gives robotics a new operating layer. Robots are not only getting better bodies. They are starting to get better ways to interpret context, plan actions and coordinate with digital systems.

    That changes the question. It is no longer only: what can AI answer? It becomes: what can AI do when it can perceive, decide and move?

    For leaders, the implication is practical: start mapping where physical work could become programmable. The strategic question is not "Should we buy robots?" It is where sensing, decision-making, workflow automation and safe machine execution could change cost, throughput, resilience or customer outcomes.

    From chatbot to operator

    The first wave of generative AI lived in a text box. It wrote, summarized, translated, coded and made knowledge work faster.

    The second wave is more ambitious. Agentic AI plans, checks, books, routes, escalates and triggers workflows. It turns AI from an interface into an operator.

    Robotics is where the operator model starts to touch the real world.

    If chatbots made AI visible, and agents make AI operational, robotics makes AI physical.

    This is a much bigger jump than the interface suggests. A chatbot operates in language. A software agent operates in digital systems. A robot operates in environments where physics, safety, maintenance, regulation and human trust all matter at the same time.

    That is why I would not start this discussion with humanoids. Humanoids are one form factor. The bigger story is physical AI: models, sensors, actuators, chips, batteries, simulation, edge computing, fleet software and enterprise workflows coming together.

    Robotics is not one market

    Robotics market stack showing industrial, service, medical, defense, consumer, and humanoid robot segments
    Robotics is not one market. It is a connected stack of industrial, service, medical, defense, consumer, and general-purpose systems.

    Robotics is already a real market, and it is much broader than the humanoid headlines.

    Industrial robots remain the established core: welding, assembly, painting, material handling, electronics, automotive and semiconductor manufacturing. Professional service robots cover logistics, warehouse automation, inspection, cleaning, hospitality, agriculture, construction and security. Medical and care robots include surgical systems, rehabilitation devices and hospital logistics.

    Defense and security robotics adds unmanned aerial, ground, surface and underwater systems, counter-drone capabilities, explosive ordnance disposal, reconnaissance, logistics and infrastructure protection. Consumer robots cover domestic devices such as vacuums and lawn robots. Humanoid and general-purpose robots sit on top of this stack as an early-stage category for environments built around human bodies.

    The data matters because it grounds the story. The International Federation of Robotics reported 542,000 industrial robot installations in 2024 and a global operational stock of 4.664 million units. Asia accounted for 74% of new deployments. China alone represented 54%.

    Service robotics is smaller and more fragmented, but it is moving. IFR's World Robotics 2025 service robot summary reported that worldwide sales of professional service robots grew 9% in 2024 to more than 199,000 units. Medical robots grew 91% to nearly 16,700 units.

    Market forecasts point in the same direction, even if the exact numbers should be treated carefully. Goldman Sachs sees the humanoid robot market reaching $38 billion by 2035. Morgan Stanley outlines a much larger long-term scenario: a potential $5 trillion humanoid market by 2050, including supply chains, repair, maintenance and support.

    Defense is one of the clearest signals that robotics is becoming a strategic technology segment, not only an automation category. Fortune Business Insights estimates the military robots market at $19.82 billion in 2025 and projects it to reach $42.90 billion by 2034. The exact number matters less than the direction: militaries are shifting from isolated unmanned platforms toward fleets, autonomy, sensing, secure communications and human-machine teaming.

    The point is not to believe every forecast. The point is that robotics is starting to look less like a hardware niche and more like a debate about who controls the operating layer of physical work.

    Why the cycle feels different now

    Robotics has had false dawns before. What makes this cycle worth watching is that several constraints are shifting at once.

    AI models are becoming more useful for perception, planning and adaptation. Google DeepMind describes Gemini Robotics as bringing AI agents into the physical world; Google's later Gemini Robotics-ER 1.6 work focuses on spatial logic, multi-view understanding, task planning and success detection.

    Simulation is improving too. Robots need data, but the physical world is expensive and slow. Synthetic environments, world models and simulation frameworks can compress training cycles. That is why NVIDIA's physical AI announcement matters: Jensen Huang called this a "ChatGPT moment for robotics" and framed physical AI as models that understand the real world, reason and plan actions.

    Enterprise demand is also clearer than before. Labor scarcity, warehouse complexity, aging populations, healthcare capacity, nearshoring and infrastructure build-out all create demand for automation that can work beyond perfectly structured factory cells.

    The market is not waiting for household humanoids. It is starting with work.

    It is also starting with security. The U.S. Department of Defense's Replicator initiative is built around all-domain attritable autonomous systems: lower-cost systems that can be fielded, updated and replaced faster than traditional platforms. NATO's DIANA Rapid Adoption Service recently awarded an R&D contract for undersea robotics and describes its role as helping Allies "move faster from identified capability need to real-world solutions." That is the defense version of the same physical AI thesis.

    Operator overseeing autonomous drone, ground, and undersea robotics systems for defense and security missions
    Defense robotics is shifting from isolated platforms toward autonomous fleets, sensing, secure communications, and human-machine teaming.

    Humanoids are the headline, not the whole story

    Humanoids matter because the world is built for people. Door handles, stairs, shelves, tools, kitchens, hospital rooms and factory aisles assume a human body.

    If robots can operate in those environments, the cost of automation changes. Companies may not need to redesign every workflow around a fixed machine. The machine could adapt to the workflow.

    That is the promise. It is also where the hype gets dangerous.

    Most useful robotics deployments will start where the economics are precise: structured tasks, high labor scarcity, safety risk, repetitive physical work, expensive downtime or environments where human work is hard to scale.

    Amazon is a useful case because it shows the less cinematic version of the future. The company says it has deployed its one millionth robot and introduced DeepFleet, a generative AI foundation model designed to coordinate robot movement across fulfillment centers. The stated goal is a 10% improvement in robot fleet travel efficiency.

    That is how physical AI will often arrive: not as a robot that looks like a person, but as a system-level improvement in throughput, cost, safety or resilience.

    The recent signal: capital is moving toward physical AI

    The last few weeks made the theme harder to dismiss.

    Germany's NEURA Robotics announced a Series C round of up to $1.4 billion in June 2026, backed by investors including NVIDIA, Amazon, Qualcomm, Bosch, Schaeffler, the European Investment Bank and Tether. NEURA founder David Reger put the strategic point plainly: "The future of AI will not only live on screens."

    OpenAI is also leaning into the theme. Sam Altman's 2026 roadmap says 2027 may bring robots that can do tasks in the real world. Separate reporting on OpenAI Robotics hiring is best read as a secondary signal, not the core proof point.

    This is more than a robotics startup cycle. It is a convergence of AI labs, cloud-scale compute, semiconductor platforms, industrial companies and capital markets.

    That matters for Europe. If physical AI becomes an industrial operating layer, Europe is not limited to being a regulator of someone else's platform. Its manufacturing base, robotics suppliers, automotive sector, industrial software, safety know-how and Mittelstand process expertise could become part of the stack, provided capital, compute, talent and adoption speed match the ambition.

    The operating model question

    The real question is not whether to buy robots. That is too narrow.

    The better question is: which parts of the operating model become programmable when AI can act in both digital and physical environments?

    In logistics, software agents may forecast demand, rebalance inventory and dispatch autonomous mobile robots. In healthcare, AI may coordinate patient logistics while robots move supplies or support clinical workflows. In manufacturing, physical AI may help factories adapt faster to product variation, quality issues or labor constraints.

    In defense, the question is even sharper. Autonomous systems can extend sensing, logistics, surveillance, electronic warfare and force protection into environments where human presence is dangerous or too slow. This does not remove the need for human judgment. It raises the standard for command, control, accountability, cyber resilience and rules of engagement.

    The value is not the robot in isolation. The value is the loop: sense the environment, interpret the situation, decide what should happen next, act safely, learn from the outcome.

    That loop is what makes robotics strategically interesting.

    It also makes it risky.

    Governance moves into the physical world

    Executives reviewing governance controls for supervised robotics and physical AI in an automated operations environment
    Physical AI will require governance models that cover permissions, audit trails, human override, safety, and accountable operations.

    Enterprises are still learning how to govern text-generating AI. Physical AI raises the bar.

    A weak chatbot answer can mislead a user. A poorly governed software agent can execute the wrong digital workflow. A poorly governed robot can damage equipment, block a line or create a safety incident in a regulated environment.

    That means physical AI needs a governance model before it scales.

    Who owns the robot's actions? What permissions does it have? What tasks require human approval? How are decisions logged? How is an incident reconstructed? Who updates the model? Who certifies safety after the model changes?

    These are not IT questions only. They are operating model questions.

    My expectation is that the companies that do this well will not describe the work as a robot deployment. They will describe it as a redesign of work: human judgment where ambiguity is high, machine execution where repetition and safety allow, and clear escalation when the system reaches its boundary.

    What I'm watching

    Four things will tell me whether this thesis is right.

    First, whether robotics deployments move from isolated machines to fleet-level operating systems.

    Second, whether AI labs and industrial companies build repeatable safety and governance patterns, not only better demos.

    Third, whether customers buy measurable outcomes rather than robots: lower downtime, faster fulfillment, safer operations, more resilient logistics or higher asset utilization.

    Fourth, whether the market starts valuing robotics companies as platform ecosystems rather than hardware manufacturers.

    AI is moving from language to action. From action to coordination. From coordination to physical work.

    The first wave lived on screens. The next one will increasingly show up in the world those screens were designed to manage. Physical AI is not just a device transition. It is an operating model transition.

    Sources and further reading

  • Model Dependency Is the New AI Business Continuity Risk

    Model Dependency Is the New AI Business Continuity Risk

    Claude Fable 5, model dependency risk, and why AI sovereignty is no longer only about where data lives.

    Andreas's view

    I would have liked more time with Claude Fable 5.

    Not because a new benchmark table matters by itself. It does not. But because frontier models are now becoming operating infrastructure. When access disappears, the issue is no longer product disappointment. It is continuity risk.

    The early description sounded like a model that pushed several practical boundaries at once: longer autonomous work, stronger coding, better vision, better long-context memory, more capable scientific reasoning. That is exactly the kind of model you want to test yourself. Not in a demo. In messy work. In a real harness. In the kind of workflow where you can feel whether the model changes what is possible.

    Unfortunately, that window closed almost immediately.

    Anthropic announced Claude Fable 5 and Claude Mythos 5 on June 9, 2026. The launch framed Fable 5 as a generally available Mythos-class model with safeguards, and Mythos 5 as the same underlying model with some safeguards lifted for trusted cyber and biology use cases.

    Three days later, Anthropic added an update: access to Fable 5 and Mythos 5 was unavailable.

    The follow-up statement is the more important business story. Anthropic said the US government had issued an export-control directive requiring it to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States. To comply, Anthropic said it had to abruptly disable both models for all customers. Other Anthropic models were not affected.

    Based on Anthropic's public account, the directive was triggered by national-security concerns around model misuse; as of this writing, the government's full rationale has not been publicly detailed.

    This is where the story stops being about one model release.

    It becomes a preview of a much larger question: what happens when an enterprise builds critical workflows around a model that can disappear, degrade, fall back, become restricted, change policy, or become unavailable for reasons outside the enterprise's control?

    The model is becoming part of the process

    Enterprise workflows converging into a central AI model dependency
    As AI moves into real workflows, the model becomes part of the operating process rather than a standalone tool.

    Most companies still talk about AI models as if they were tools. You pick one, connect it to a use case, monitor cost and quality, and move on.

    That framing is becoming too simple.

    In real deployments, the model is increasingly part of the operating process. It sits inside support flows, developer environments, research workflows, compliance review, sales operations, procurement analysis, risk triage, security workflows, and internal knowledge systems.

    The more capable the model, the more tempting it becomes to build around its specific behavior.

    That is where the dependency starts.

    A company does not only depend on the model name. It depends on latency, context length, tool use, refusal behavior, reasoning style, pricing, data-retention policy, regional availability, safety fallbacks, API contracts, rate limits, and the model's ability to work inside a specific harness.

    The harness matters. A workflow may depend on prompt structure, tool calls, memory files, evaluation thresholds, orchestration logic, fallback assumptions, and human review steps. If the underlying model changes, the process can change with it.

    Sometimes that is manageable. Sometimes it breaks the economics. Sometimes it changes the risk profile. Sometimes it simply means the workflow no longer works.

    Fable 5 is a case study in availability risk

    The Fable 5 launch itself was ambitious. Anthropic described strong performance in software engineering, knowledge work, vision, scientific research, long-context memory, and life sciences. It also described safeguards that would route some sensitive requests to Claude Opus 4.8 instead of allowing Fable 5 to answer directly.

    That already shows the new shape of frontier AI products. The "model" is no longer a single stable object. It is a capability layer plus policy logic, classifiers, routing, monitoring, data-retention rules, trusted-access programs, and usage conditions.

    Then came the suspension.

    Anthropic said the government directive was based on national-security authorities and that it had to remove access for all users. It also said it disagreed with the action and believed that applying this standard across the industry could halt new frontier model deployments.

    Whether Anthropic or the government is right is not the central issue for enterprise leaders.

    The customer was not in control. Even a short-lived availability window is enough to expose the broader risk: pilots, evaluations, procurement decisions, and roadmap assumptions can all form around capabilities that may not remain available.

    The sharper lesson is that access to frontier capability has become a business continuity variable.

    The hidden risk: model concentration

    Model concentration is becoming a real operational risk.

    The risk is not that one provider has an outage. Enterprises already understand cloud outages. The risk is that AI capability is becoming more specific and less interchangeable.

    If two models both answer emails, switching is easy.

    If one model can run a long-horizon code migration, interpret screenshots, manage tool calls, maintain working memory, and reason through edge cases in a particular way, switching becomes harder. The replacement may be available, but the workflow may need to be redesigned.

    That is a different kind of lock-in.

    It is not only commercial lock-in. It is cognitive and procedural lock-in. Over time, the organization's workflows, prompts, review habits, and escalation paths start to conform to the model's strengths and weaknesses.

    This will matter most in high-value use cases: engineering modernization, cyber defense, regulated research, legal work, finance analysis, industrial design, life sciences, and autonomous operations.

    The more strategic the use case, the less acceptable it is to rely on a single model path with no tested fallback.

    What enterprise architecture needs to change

    AI continuity architecture with a primary model unavailable and fallback paths active
    A model continuity plan needs tested fallback paths, evaluations and escalation logic, not just a procurement preference.

    The practical answer is not to avoid frontier models. That would be the wrong lesson.

    The answer is to treat model dependency as architecture, not procurement.

    Companies need a model continuity plan.

    For every important AI workflow, leaders should know which assumptions are model-specific. What breaks if the model is removed? What happens if it falls back to a weaker model? What if data-retention rules change? What if the API remains available in the US but not in Europe? What if a safety classifier suddenly routes part of the work elsewhere?

    This is not theoretical governance paperwork. It is operational design.

    The risks are different, and they need to be named differently. An outage is not the same as a policy withdrawal. A safety-routing change is not the same as capability degradation. A regional access restriction is not the same as a pricing change. But all of them can alter a workflow that the business has started to rely on.

    The better pattern is a portfolio:

    • a primary frontier model for maximum capability
    • a tested fallback model for continuity
    • evaluations that measure task quality across model options
    • abstraction layers that keep prompts, tools, memory and workflow logic portable where possible
    • logs that show when fallbacks happen and why
    • contracts that address availability, regional access, data policy and notice periods
    • human escalation when model behavior changes materially

    The important word is tested. A fallback that has never been used under realistic workload is not a fallback. It is a slide in an architecture deck.

    The board-level questions are practical:

    • Which AI workflows would stop if the primary model disappeared?
    • Which fallbacks have been tested under real workload?
    • Which model-specific assumptions are embedded in prompts, tools, policies and contracts?
    • Who owns model continuity: procurement, architecture, risk or the business unit?

    Why this becomes a sovereignty issue

    European enterprise connected to AI capability nodes with one external path interrupted
    AI sovereignty increasingly means asking who can interrupt a capability, not only where data is stored.

    For Europe, this is where the story becomes uncomfortable.

    AI sovereignty is often discussed as data residency, cloud location, compliance, or whether a model is hosted in Europe. Those questions matter. But they are no longer enough.

    The Fable 5 episode shows another layer: model availability can be shaped by decisions outside the customer's jurisdiction.

    A European company may comply with European law, host data in Europe, and still depend on an AI capability that can be changed or removed because of a US policy decision, provider safety decision, capacity constraint, licensing change, or export-control interpretation.

    That does not mean every company needs to train its own frontier model. That would be unrealistic for most. It is also not an argument for isolationism or for purely national AI stacks.

    It does mean Europe needs to think about sovereignty at the level of operating capability.

    Can critical public-sector, industrial, defense, healthcare, financial and infrastructure workflows continue if a non-European model is restricted? Are there European or allied alternatives? Are there open-weight or locally deployable fallbacks for lower-risk parts of the workflow? Are procurement teams asking for exit paths? Are regulators looking at operational resilience, not only privacy?

    The sovereignty question is shifting from "Where is the data?" to "Who can interrupt the capability?"

    That is a much harder question.

    The strategic lesson

    I still would have liked to play with Fable 5.

    That is partly curiosity. Frontier models are easiest to understand when you test them against real work. You learn more from one serious workflow than from a benchmark chart.

    But the more important lesson is the one created by not being able to use it.

    The future of enterprise AI will not be decided only by who has the strongest model. It will also be decided by who can build resilient operating models around unstable capability layers.

    Models will improve. Policies will change. Access rules will shift. Providers will make safety decisions. Governments will intervene. Capacity will be constrained. Prices will move.

    For CIOs, CTOs and risk leaders, the question is no longer whether to use frontier models. It is whether every critical AI workflow has a tested continuity path.

    The companies that win will not be the ones that pretend this volatility does not exist. They will be the ones that treat model volatility as a design constraint from the beginning.

    Sources and further reading

  • AI’s next bottleneck may not be intelligence. It may be Earth.

    AI’s next bottleneck may not be intelligence. It may be Earth.

    For the last two years, the AI debate has been mostly about intelligence.

    Which model is ahead? How fast are capabilities improving? Will agents replace tasks, jobs, or whole workflows? Can Europe regulate the technology fast enough?

    All valid questions.

    But the next constraint may be less abstract. It may be physical.

    Power. Grid capacity. Land. Cooling. Permits. Transmission lines. Water. Construction time. Capital allocation.

    The AI race is turning into a gigawatt race. And if the space-data-center discussion is any signal, the next frontier may not just be cloud regions. It may be orbit.

    My read: the executive conversation has to move from "Which AI model should we use?" to "What physical infrastructure does our AI strategy depend on?"

    The scale shift

    Chart showing typical data center power use from 5-10 MW to 100 MW and 1 GW
    The scale jump matters: 10 MW is a facility, 100 MW is industrial infrastructure, and 1 GW becomes a regional energy strategy.

    A modern hyperscale data center is not a large office building with servers. It is an industrial energy asset.

    The International Energy Agency says average data centers draw around 5-10 megawatts. Large hyperscale facilities increasingly require 100 megawatts or more. That number sounds technical, so translate it.

    One megawatt running continuously for a year equals 8.76 gigawatt-hours. A 100 MW data center therefore consumes 876 GWh per year, or 0.876 TWh. At 90% utilization, still roughly 0.8 TWh per year. The IEA compares this to the annual electricity demand of about 350,000 to 400,000 electric cars.

    A 1 GW AI campus is ten 100 MW hyperscale data centers. Running continuously, it consumes 8.76 TWh per year.

    For comparison, Germany's annual electricity consumption is roughly 500 TWh. The EU is around 2,700 TWh. The US is around 4,000 TWh. So one 1 GW AI campus would be small at continental scale – about 0.3% of EU electricity consumption or 0.2% of US consumption – but huge at local grid scale.

    That local point matters.

    Put a 1 GW load in the wrong county, with weak transmission and slow permitting, and it is not "0.2% of America." It is a grid emergency, a political fight, and a capital allocation problem.

    Now consider the language around terawatts. Elon Musk's recent "Terafab" discussion was about chip manufacturing, not a conventional data center, but the vocabulary matters. AI infrastructure ambition is moving from mega to giga to tera. A theoretical 1 TW compute or manufacturing footprint running continuously would consume 8,760 TWh per year. That is more electricity than the US and EU combined.

    That does not mean a 1 TW data center is around the corner. It means the ambition curve is now colliding with the energy system.

    The current footprint

    The IEA estimates global data center electricity consumption at 240-340 TWh in 2022, excluding crypto mining. That was around 1-1.3% of global final electricity demand.

    In large economies such as the United States, China and the European Union, data centers already account for around 2-4% of total electricity consumption. That is the average.

    The local reality is more extreme.

    The IEA notes that data centers have already surpassed 10% of electricity consumption in at least five US states. In Ireland, data centers account for more than 20% of electricity consumption. Denmark projects data center electricity use could rise sixfold by 2030 and approach 15% of national electricity consumption.

    This is the important distinction: globally, data centers are still a manageable share of electricity. Locally, they can become one of the dominant loads on the system.

    Goldman Sachs Research estimates data center power demand could grow 160% by 2030, with global data centers rising from roughly 1-2% of power consumption today to 3-4% by the end of the decade. It also estimates AI could add around 200 TWh per year of data center power demand between 2023 and 2030.

    Two hundred TWh is not abstract. It is close to the annual electricity consumption of a mid-sized industrial country. And it is only the AI-related increment in one forecast.

    The backlash is already here

    Chart comparing global data center electricity share with US, EU, Ireland and local grid impacts
    Global averages hide local pressure: data centers can reach double-digit shares of electricity demand in specific regions.

    This is no longer theoretical.

    In May, several local flashpoints showed the political side of the bottleneck. Seattle was weighing a pause on large data centers. Durham, North Carolina passed a 60-day moratorium on data-center development. A Texas county paused data-center construction in rural areas for a year. Utah approved a data-center project described as twice the size of Manhattan, triggering backlash. Tennessee was considering legislation that would let data centers self-power with limited regulation.

    Different places, same pattern.

    AI infrastructure is colliding with local politics. Communities are asking who gets the jobs, who pays for grid upgrades, who carries water risk, who absorbs noise and land-use impact, and who benefits from the compute.

    This is the part of the AI story many executives still underestimate. It is not enough to have GPUs. You need permission. You need interconnection. You need credible energy sourcing. You need community acceptance.

    The future of AI may be decided as much in planning boards and utility queues as in model labs.

    Why energy is now part of AI leadership

    Executive checklist for AI energy strategy and infrastructure planning
    AI energy strategy is now an executive checklist: economics, thresholds, model allocation, partnerships, and efficiency.

    For a long time, digital leaders could assume infrastructure would scale behind the scenes. Cloud abstracted away servers. SaaS abstracted away operations. Developers increasingly acted as if compute was infinite, elastic, and mostly someone else's problem.

    AI breaks that illusion.

    Training frontier models is energy-intensive. Inference at scale may matter even more because successful AI products are used continuously. Agents add another multiplier: they do not just answer one prompt. They plan, call tools, retry, search, generate, check, and act. A single user request can become dozens or hundreds of model calls behind the scenes.

    That makes energy not just an engineering issue but a leadership issue.

    If AI becomes a core production layer, power becomes part of product economics. Latency becomes part of geography. Energy procurement becomes part of risk management. Infrastructure partnerships become part of market entry. Sustainability claims become harder to defend if absolute consumption rises faster than efficiency improves.

    The better question is not whether AI uses "too much" energy.

    The better question is: are we using scarce energy for high-value intelligence, or are we wasting it on low-value automation theatre?

    The opportunity

    The upside is enormous.

    AI can help design better grids, forecast demand, optimize industrial processes, improve cooling, accelerate materials science, reduce waste, and make energy systems more flexible. The same technology that increases electricity demand can also improve how electricity is produced, routed, stored, and consumed.

    There is also a market opportunity.

    Companies that solve the infrastructure layer will not just be suppliers to AI. They will become strategic gatekeepers. Power developers, grid operators, data-center builders, cooling specialists, chip designers, construction firms, nuclear developers, storage providers, and energy software companies are moving closer to the center of the AI economy.

    This is especially relevant for Europe.

    Europe often frames AI competitiveness around regulation, foundation models, sovereignty, and talent. All matter. But infrastructure sovereignty may become just as important. If compute depends on power availability, grid speed, and data-center capacity, then AI sovereignty is partly electricity sovereignty.

    A European AI strategy without an energy strategy is incomplete.

    The space question

    Conceptual space-based AI data center with solar arrays orbiting above Earth
    Space-based data centers are not a near-term replacement for terrestrial infrastructure. They are a signal that the AI compute curve is pushing beyond the grid.

    The more provocative version of this debate is space.

    A few years ago, data centers in orbit sounded like science fiction. Now Bloomberg is writing about how to build them. McKinsey has made the case for space-based data centers. University researchers are exploring the idea because AI energy demand is rising. Google and SpaceX have been linked in recent coverage to the broader possibility of AI data centers in space.

    The attraction is obvious: continuous solar power, less terrestrial land pressure, potentially easier cooling through radiative systems, and the strategic appeal of moving part of the compute layer off Earth.

    The problems are just as obvious: launch cost, maintenance, radiation, latency, orbital debris, security, regulation, and basic economics.

    But the fact that serious people are asking the question matters. Space data centers are not a near-term replacement for terrestrial infrastructure. They are a signal. The AI compute curve is steep enough that people are looking beyond the grid.

    When a technology forces executives to ask whether the data center belongs in orbit, something fundamental has changed.

    What leaders should do now

    The call to action is practical.

    First: put energy into the AI business case. Every serious AI initiative should have a compute and energy view, not just a model and vendor view. If the project scales 10x or 100x, what happens to cost, latency, emissions, and capacity?

    Second: use real thresholds. A 10 MW workload is a large facility. A 100 MW workload is industrial infrastructure. A 1 GW workload is a regional energy strategy. Treat them differently.

    Third: separate high-value intelligence from low-value automation. Not every workflow deserves heavy AI. Use frontier models where judgment, ambiguity, and leverage justify the cost. Use smaller models, retrieval, caching, rules, and process redesign where they are enough.

    Fourth: make infrastructure a board-level topic. If AI is strategic, then power supply, data-center capacity, cloud concentration, and sustainability are strategic. CIOs, CTOs, CFOs, COOs, and sustainability leaders need one shared view.

    Fifth: build partnerships beyond software. The AI stack now reaches into energy markets, utilities, real estate, cooling, semiconductors, construction, public policy, and eventually maybe space.

    The leadership shift

    The first AI leadership question was: "What can this technology do?"

    The second was: "How does it change work?"

    The third is now emerging: "What does it require from the physical world?"

    This is where the debate becomes more serious.

    AI is not just a software wave. It is a capital investment wave, an energy demand wave, and an infrastructure coordination problem. The limiting factor may not be imagination. It may be megawatts.

    Executives should not panic about that. But they should stop treating it as somebody else's problem.

    Models matter.

    But electricity decides where the models can run. And if the curve continues, the strategic question may become even stranger:

    How much intelligence can Earth afford to host?

    Sources and further reading

  • AI: creating or destroying jobs?

    AI: creating or destroying jobs?

    The AI-jobs argument has split into two camps that aren’t actually arguing about the same thing.

    Jensen Huang told CEOs at GTC that firing people for AI shows “no imagination” — radiologists, he points out, are more numerous now than before AI entered radiology. Marc Andreessen calls the displacement narrative “completely fabricated” and points to Jevons Paradox: cheaper labor produces more demand, not less. The WEF Future of Jobs Report still projects net +78 million jobs globally by 2030. Challenger’s Hiring Plans index was up 157% year-over-year in March.

    A week later, Block laid off 40% of its workforce. Jack Dorsey said engineering work that needed weeks now happens in a fraction of the time. Block is still hiring AI engineers.

    So which is it?

    My read: both sides are right. They’re answering different questions about different decades. Most of the public argument is two conversations pretending to be one.

    The optimist case

    Three pieces hold it up.

    The historical record is strong. Keynes wrote in 1930 that his grandchildren would work fifteen-hour weeks. Reality 2025: OECD average is thirty-seven hours, Americans clock 1,976 hours a year. Mechanization, electrification, the computer, the internet — every general-purpose technology was forecast to end work, and every one produced more jobs than it eliminated. In 1900, 41% of Americans worked in agriculture; today it’s 2%. The jobs went somewhere.

    Jevons Paradox is real. When something useful gets cheaper, demand rises. If AI makes cognitive work twenty times cheaper, you don’t end up with one-twentieth the cognitive work. You end up with twenty times the cognitive work, deployed against far more problems. Andreessen’s “Super-PhD in every field” captures it.

    A big chunk of the labor market is hard to displace. Licensed jobs (medicine, law, accounting), unionized jobs (skilled trades, transit, public safety), and public-sector roles add up to a large fraction of US employment. Not protected because they’re irreplaceable in some technical sense — protected by institutions that move slowly.

    Each piece is correct. The question is whether they’re enough.

    Where the optimist case breaks

    Radar chart of AI capability versus observed usage across eight occupations from the Anthropic Economic Index, showing the deployment gap.
    The deployment gap: theoretical AI capability dwarfs observed usage by occupation. Source: Anthropic Economic Index.

    The Anthropic Economic Index plots theoretical AI capability against observed AI usage by occupation. The two lines look almost nothing alike — capability is broad and high; usage is narrow and concentrated. There’s a gap between what AI can do and what it’s actually doing.

    Read that gap two ways. The optimist reading: deployment is slow, friction is real, the labor market reabsorbs shocks like it always has. The harder reading: the gap is the queue — it’s where displacement comes from over the next five to ten years, not from new capability but from deployment catching up to capability that already exists.

    94% of cognitive job tasks are theoretically automatable today; 33% actually are. The space between is the transition zone. It’s not science fiction. It’s not contested. Most of it will close. Block’s layoffs sit on the second reading.

    The historical-record argument also has a footnote that doesn’t get enough weight. AI is the first general-purpose technology to automate cognitive labor at scale. Every prior wave automated muscle, then narrow categories of cognitive work — but never the universal category of “thinking and writing and analyzing and deciding.” The tractor displaced farm hands; they moved into office work. The PC displaced typewriters and clerks; they moved into knowledge work. AI doesn’t have an obvious “moved into” destination, because the destination of every prior wave is the category AI now automates.

    The TIME / Contextual AI benchmark chart makes the universality vivid. AI surpassed human-level performance on handwriting recognition around 2015, then speech, then images, reading, language, common sense, math, code generation. The rate at which new tasks fall is increasing.

    The trades-and-physical-work counterargument is weaker than it looks. Yes, 57% of jobs depend on physical presence or craft work AI can’t currently replicate. But 70% of positions inside blue-collar companies — the dispatcher, the accountant, the customer-service rep — are white-collar-adjacent and fully exposed. And if displaced knowledge workers all migrate into trades, wages collapse from saturation. Bank of America projects billions of humanoid robots by mid-century with hardware costs falling from $35,000 to under $15,000; one analyst projects robot-hours at four to six euros. Even physical work has an expiration date.

    So the optimist case is strong for a long-run answer. It’s much weaker for the next ten years.

    The displacement case

    Not “AI replaces all jobs.” That’s the optimists’ caricature, and once you reach for it the displacement case looks weak. The serious version is more specific.

    Three vertical bars on dark navy: high-skill rising, middle-skill shrinking with downward arrow, low-skill stable — the AI barbell economy.
    The barbell economy: high-skill productivity rises, low-skill stable, the middle hollows out.

    It’s structural: the middle is being squeezed. The labor market is shifting from a K-shape into a barbell. High-skill technical roles are more productive — the same Anthropic data shows code, analysis, and research at the top of the productivity-gain distribution, with usage approaching 60% of theoretical capacity. Low-skill physical roles in care, hospitality, manual handling, and trades are stable for now. The middle is shrinking: bookkeeping and paralegal work, content writing and copywriting, junior finance and analyst roles, customer service, entry-level coding, marketing copy, translation, project coordination, junior tax preparation.

    Germany has already seen roughly 90,000 AI-related job losses in the first months of 2026. The risk is not mass unemployment in aggregate. Aggregate unemployment can stay low for years while the middle hollows out. The risk is a split labor market — and a split society — in which the people who staffed the middle no longer have a clear path up or sideways.

    The Anthropic Economic Index BLS panel makes this concrete: hiring of younger workers in AI-exposed occupations has slowed, even as overall employment numbers haven’t moved much. That’s what early-stage hollowing looks like — the entry-level rung disappears first, before the established middle does.

    Five years of that compounds into something the historical record didn’t have to absorb.

    My read — the three-phase shape

    Horizontal timeline 2025 to 2040+ split into three colored zones: red displacement, amber strain, cyan abundance — AI jobs transition phases.
    Three phases of the AI jobs transition: displacement (2025-2030), strain (2030-2035), abundance (2035+).

    The clearest three-phase framing is German — chronological, not parallel.

    Phase one — displacement (~2025-2030). AI displaces knowledge work faster than the labor market rebuilds. The middle hollows. Aggregate unemployment may not move much; entry-level paths in white-collar roles narrow sharply. The optimists are right that the technology eventually creates new categories. They’re wrong about the timing.

    Phase two — strain (~2030-2035). Strain shows up in places that aren’t unemployment: tax-base erosion, weakened consumer demand, capital returns rising while labor’s share of national income falls to historic lows. Public-sector and licensed-job cushions hold initially but come under fiscal pressure. The political consequences sharpen.

    Phase three — abundance (after ~2035). The deflation the optimists describe arrives. Costs collapse across categories. What costs $100 today costs a few cents. The median 2040 lifestyle, on a flow-of-services basis, looks something like today’s high-net-worth lifestyle on every dimension except positional goods. Both Andreessen and Huang are right about the destination.

    That’s the timeframe trap. Both sides are correct on their respective horizons. The honest version of the optimist case includes the transition pain. The honest version of the displacement case includes the recovery.

    What this means for how leaders think about the next ten years: the question isn’t “do we believe in AI displacement, yes or no.” That question is roughly answered. The task is to assume real displacement in the middle, plan for it, and carry the organization through to the recovery in a way that keeps the institution and its people whole.

    Three things I’m watching

    1. Whether the entry-level signal becomes a leading indicator. The slowing of hiring for younger workers in AI-exposed occupations is, in my read, the most important early signal. Aggregate employment numbers lag; entry-level absorption leads. If the slowdown becomes a structural break, phase one stops being a forecast and becomes a measurement.
    2. Whether the licensed and public-sector cushion holds when fiscal space tightens. The structural-protection argument is strong only as long as the institutions that protect those jobs don’t themselves come under fiscal pressure. Phase two erodes the tax base. The question is whether legislatures and regulators are protecting genuinely-essential public-sector employment or post-hoc subsidizing the share of the workforce the private sector can no longer place.
    3. Whether the recovery looks like restored employment or restored income. Phase three is consistent with both. Jobs come back in new categories — the historical track record. Or they don’t come back at scale and the recovery is income-shaped: UBI-like distribution of the deflation surplus rather than wage-based participation. These look very different politically. The shape of phase two is what determines which one we get.

    No one has confidence on these three questions yet. I’m watching them because the answers will tell us, in roughly the next five years, what the transition phase actually costs.

    The destination is not in serious doubt. The road is.

  • Every knowledge worker is a manager now

    Every knowledge worker is a manager now

    Every knowledge worker is a manager now. Agentic AI has turned individual contributors into managers of AI agents, and first-line managers into leaders of managers of agents. The job descriptions have not caught up yet. The operating models have not caught up yet. The reskilling plans have not caught up yet. All of that is lagging the capability frontier by twelve to eighteen months — and the organizations that close that gap first will operate at a structurally different throughput than the ones still writing job descriptions for the jobs that existed in 2023.

    The shift: agentic AI crosses the line from tool to colleague

    For the first year and a half after ChatGPT, the thing called “AI” in most organizations was a better search box. A more patient editor. A faster rough-draft generator. Useful, but still a single-interaction tool. You asked, it answered, you moved on. The job of the knowledge worker did not fundamentally change — they just had a slightly sharper pencil.

    What changed in the eighteen months leading into 2026 is the arrival of agentic models. The word “agent” in that context is not marketing. An agent is a system that can do a sequence of things, hold state across those steps, make decisions about what to do next, use tools, and come back with a completed multi-step task. That is a categorically different interaction than “ask question, get answer.” It is closer to “give a junior colleague an outcome to produce and trust them to produce it.” The commercial consequence of that shift is the subject of this post.

    Knowledge-worker image candidate K02-HC-pipeline: HC2 — INPUT-AGENT-OUTPUT-JUDGE-SHIP pipeline with human at JUDGE
    Input → agent → output → judge → ship. The human stays at the judgment node.

    The role change: ICs become managers of agents

    The individual contributor job has silently changed. Writing short summaries of long content — once a junior-to-mid task — is now an agent task. The human role is to specify the outcome, check the output, and decide what to do with it. Meeting preparation — the pre-meeting brief of background, context, attendees, prior touchpoints — is now an agent task. The human role is to feed the context, review the brief, and adjust the framing. Drafting a first pass of almost any structured document — a proposal, a plan, an analysis — is now an agent task. The human role is the editor, not the author of the first draft.

    The common thread is that the IC’s job has shifted from doing to specifying outcomes and judging output. Those are management skills. Not in the metaphorical sense — in the literal sense. Framing a task clearly enough that someone (or something) else can execute it. Evaluating whether the execution meets the specification. Deciding when to iterate and when to ship. These are exactly the skills that used to distinguish a first-line manager from a senior IC, and they have become baseline requirements for an IC working with agents.

    Knowledge-worker image candidate K03-HC-editor: HC3 — colleagues editing agent outputs + overlay text
    The new role for the IC: editor of agent output.

    The org change: first-line managers become leaders of managers of agents

    If every IC is now a manager of agents, then every first-line manager is now a leader of managers of agents. Their job is no longer to supervise execution — the agent is doing the execution. Their job is to coach the humans on their team in how to specify outcomes, how to judge output, how to know when an agent is producing garbage, and how to scale their orchestration over time. That is a completely different job than the first-line management job of three years ago, and it requires a different skill set.

    Two structural consequences follow. First, the middle management layer compresses because a first-line manager leading managers-of-agents can reach further than one managing direct executors — the coordination overhead per report drops when the reports are themselves operating on a multiplier. Second, the definition of “span of control” stretches, but not infinitely: the Dunbar layers still govern the number of humans a manager can hold relationships with, even if each of those humans is now operating agents underneath them. The org chart can get flatter. It cannot get unbounded.

    Knowledge-worker image candidate K05-WILD-conductor: WILD — human conductor directs an orchestra of AI agents
    One human, many agents — the conductor metaphor for first-line management at scale.

    The strategic consequence: orchestration is now a baseline skill, not an advanced one

    The skill that used to distinguish senior managers from junior ones — the ability to frame work so someone else can execute it and judge whether their execution is good — is now a baseline IC capability. Orchestration is the new baseline. Writing is the new baseline. Judgment about output quality is the new baseline. The organizations that will operate at structurally higher throughput over the next five years are the ones that reskill their IC population around these baseline orchestration skills, rather than hiring more specialists who each do one thing well.

    Talent leverage, not headcount, becomes the scoreboard. A commercial organization that operates at 300 humans with strong orchestration capability can outproduce a commercial organization that operates at 600 humans with legacy IC job descriptions. The difference is not about working harder. It is about operating model. The 300-human organization has fewer Dunbar breakpoints, shorter decision loops, less cross-functional friction, and a higher per-seat agent-multiplier. All of that is the consequence of a single structural decision made at the job-description layer.

    So what boards should do

    Three actions sit on the CEO agenda over the next two quarters. First, rewrite the IC job descriptions for every knowledge-worker role in the organization so that orchestration and output judgment are explicit baseline capabilities, not bonus ones. Second, rewrite the first-line management job description so that coaching for orchestration is the core of the role, not supervision of execution. Third, audit the reskilling plan against the assumption that every knowledge worker in the organization is now a manager and needs to be trained as one — because the capability frontier has already shipped and the only question is whether the organization catches up in quarters or in years.

    Boards that do not require a reskilling plan at this scope are budgeting against an operating model that does not exist anymore. The plan does not need to be perfect. It needs to exist. The gap between organizations that have this plan and organizations that do not is the structural competitive advantage of the next five years, and it is already being measured — in throughput, in decision velocity, in the quiet retention of the top performers who can see the gap coming.

  • Model deprecation is the new continuity risk

    Model deprecation is the new continuity risk

    Four rectangles in a row with the leftmost ghosted, simple connecting arrows
    A — model lifecycle row.

    OpenAI announced the discontinuation of the Sora web and app experiences on April 26, with the Sora API following on September 24. The first deprecation triggers in two weeks. Enterprises that built workflows on Sora since launch are not facing a model upgrade — they are facing a workflow rebuild on a four-month timeline. This is the first prominent enterprise-facing AI deprecation event of the cycle, and the precedent it sets matters more than the specific product involved.

    Model deprecation is no longer a developer-tier concern. It is an enterprise governance question that deserves a place on the risk committee agenda. The real shift is happening here: AI dependency without continuity is becoming a board-level risk in 2026.

    The shift: dependency without continuity guarantees

    The pattern of the past two years has been to build agent workflows on whichever foundation model was demonstrably best at the time, with little contractual commitment from the model provider about how long that model would remain available. Provider terms have improved — Azure OpenAI’s twelve-plus-six-month commitment for generally available models is the strongest standard in market — but most enterprises have not negotiated equivalent terms with their chosen providers. They built on capability, not on continuity.

    When the provider sunsets the model, the enterprise’s options are bad. Migrate to a successor model that may behave differently in subtle ways — requiring re-validation of every governed use case. Renegotiate at the eleventh hour for extended access at unfavorable terms. Or absorb the operational disruption of the workflow simply not working until rebuilt.

    The Sora event is small in dollar terms but large in precedent. The next deprecation will involve a more enterprise-critical model, and the enterprises that did not see this one coming are not going to see that one coming either.

    A single thread connecting a workflow box to a model box, the thread visibly fraying near the model with a clock above
    Built on capability. Not on continuity.

    The role change is the addition of an AI continuity discipline

    Inside enterprises that take this seriously, a discipline is emerging that did not exist in 2024 — AI continuity management. The work overlaps with vendor management, with disaster recovery, with model risk management, and with regulatory compliance, but it is structurally distinct from all of them. The discipline involves maintaining an inventory of model dependencies by workflow, negotiating continuity commitments at procurement, running successor-model regression tests on a regular cadence, and ensuring that the documentation chain meets the rebuild-readiness standard.

    Most enterprises have not staffed this discipline. The accountabilities are scattered across teams that do not coordinate. The procurement team negotiated the model contract a year ago without a continuity clause. The deployment team is building production dependencies on the model without thinking about migration cost. The risk team has not flagged model deprecation as a category. When the deprecation announcement lands, the company finds out it has no plan.

    The fix is straightforward in concept and slow in practice. Add continuity commitments to the procurement template. Build a model-dependency inventory. Designate an owner for AI continuity at the executive level. Run quarterly successor-model tests. None of this is hard. It is just unglamorous work that does not get done unless someone owns it.

    The strategic consequence is renewed buy-versus-build math

    Continuity risk changes the calculus of where to deploy AI capability. For workflows where the cost of unplanned migration is high — regulated workflows, mission-critical operations, customer-facing experiences with high switching costs — the case for either fine-tuning a frontier model into a controlled deployment, partnering with a vendor offering enterprise-grade continuity commitments, or building on open-weight models the enterprise can host indefinitely is stronger than it was in 2024. The case for relying on whichever model is best on a benchmark this quarter is weaker.

    The math is not simple. Open-weight models lag the frontier, sometimes meaningfully. Self-hosting carries operational cost that the proprietary providers absorb. The vendor lock-in to a single proprietary provider, even with the best continuity terms, is a different kind of risk than open-weight self-hosting carries. Each enterprise has to make this trade-off based on the workflow’s tolerance for capability lag versus its tolerance for continuity disruption.

    What is no longer defensible in 2026 is treating model continuity as someone else’s problem. The Sora sunset is small. The next one will not be.

    So what boards should do this quarter

    Add model deprecation to the risk committee agenda. The first deprecation event lands in two weeks. The board should at minimum understand which workflows are exposed and what the migration plans are.

    Demand a model-dependency inventory. Which workflows depend on which models from which providers, with which contractual continuity commitments. If this inventory does not exist, building it is the priority.

    Reconsider the buy-versus-build posture for mission-critical AI workflows. The 2024 default — use whichever proprietary model is best — was rational at the time. In 2026, with the deprecation precedent now visible, that default deserves an explicit reconsideration. Continuity is becoming a form of resilience. The boards that price it in this quarter will not be the ones rebuilding workflows under deadline.

    References and links

  • Team sizes are not design choices. They’re cognitive limits.

    Team sizes are not design choices. They’re cognitive limits.

    Team sizes are not design choices. They are cognitive limits. The recurring numbers that show up in military units, religious communities, hunter-gatherer bands, and commercial organizations are not management philosophy. They are a property of the animal doing the work, and any organizational structure that pretends otherwise pays a measurable tax in friction, communication overhead, quiet attrition, and decisions that arrive three weeks late.

    Two. Four to six. Eight to twelve. Twenty to twenty-five. Fifty. One hundred and fifty. The specific numbers recur across centuries and industries. In the Roman legion and the US Marines. In religious communities and hunter-gatherer bands. In tech companies, sales organizations, and the advice experienced managers give each other about when to split a growing team. It is not a coincidence. It is cognitive architecture. The constraint is no longer technology. The constraint has always been the brain doing the coordinating.

    Dunbar’s layers

    The research most commercial leaders eventually bump into is Robin Dunbar’s. Dunbar is a British anthropologist who, in the early 1990s, proposed that the size of a primate’s social group is constrained by the size of its neocortex. Extrapolating from primate data, he estimated the human number at around 150 — the number of people with whom any one of us can maintain a stable, recognisable, mutually-active relationship. He published it in the Journal of Human Evolution in 1992, and the number has been running through management literature ever since.

    The part that gets talked about less, but matters more, is that Dunbar’s 150 is not a single flat layer. It is the outer ring of a nested set, each layer roughly three times larger than the one inside it:

    • ~5 — your closest support group. The people you would call in a real emergency.
    • ~15 — your sympathy group. People whose loss would significantly affect you.
    • ~50 — your band or clan. People you know well enough to share deep context with.
    • ~150 — your active community. Stable, recognisable, mutually reciprocal relationships.
    • ~500 — acquaintances.
    • ~1500 — faces you can still recognise.

    These layers show up in the research almost regardless of whether the subject is a tribal society, an office workforce, or a social-network friend graph. And they map astonishingly well onto the team sizes that commercial organizations stumble toward by trial and error — not because anyone read Dunbar, but because the alternatives don’t work.

    Round-G candidate G01-HC-editorial-figure: HC1 — central figure + concentric silhouette tiers (matches #4511 aesthetic)
    A central figure surrounded by expanding tiers — 5, 15, 50, 150.

    The military got there first

    Armies have been experimenting with how to organize humans under extreme stress for two thousand years, and they arrived at exactly these numbers through pure selection pressure. Smaller was too fragile. Larger fell apart under fire. The numbers that survived are the numbers that work.

    A Roman legion’s smallest unit was the contubernium — eight soldiers who shared a tent, a mule, a mess, and most of their waking life. Eight. Right at the boundary between the 5-person inner layer and the 15-person sympathy group. The Romans knew nothing about neocortex ratios. They noticed that a group of eight held together in a way that a group of four or a group of sixteen did not.

    The modern US Marine Corps fireteam is four. The squad is roughly 13. The platoon is 30 to 40. The company is 100 to 150. The same ratios, twenty-one centuries later. The cognitive limits haven’t moved, because the brain they are about hasn’t.

    The tech industry rediscovered the same numbers

    The technology industry discovered the same structure and gave it different names.

    Jeff Bezos’s two-pizza rule — a team should be small enough to be fed by two pizzas — is a practical restatement of the 5-to-8 cognitive sub-layer. Amazon did not get there via anthropology. They got there by watching their own product teams stall every time they grew past the point where the whole group could fit around one table.

    Scrum teams are officially 7 ± 2 — the current Scrum Guide recommends 3 to 9 members — which echoes George Miller’s 1956 paper on the working-memory limit of around seven chunks. Miller was not writing about teams. The cognitive limit he found on how many things we can juggle at once maps cleanly onto how many people we can coordinate without losing track of where everyone is.

    Fred Brooks, in his 1975 book The Mythical Man-Month, observed that adding people to a late software project makes it later, because every new person increases the number of pairwise communication channels by roughly n(n–1)/2. Seven people means 21 channels. Ten means 45. Fifteen means 105. The coordination tax is quadratic, and it surfaces as “mysterious” slowdowns at exactly the team sizes where the math stops being manageable.

    W. L. Gore & Associates, the Gore-Tex company, built Dunbar’s number directly into its real-estate strategy. Founder Bill Gore had a rule: every time a building exceeded 150 employees, they built another building. He was running Dunbar’s ceiling inside his facility planning decades before Dunbar had published the paper.

    The Ringelmann effect, documented in 1913 and one of the oldest findings in social psychology, is the same story in a different register: as group size grows, the effort each individual contributes goes down. People pull harder on a rope when there are fewer of them holding it. Max Ringelmann measured it with actual rope-pulling experiments, and the finding has been replicated many times since in workplace and sports settings.

    Nano Banana round-2 variant R07-c09-overlay-A: C09-A — two-pizza with overlay text
    The two-pizza team — Bezos’s practical statement of the cognitive sub-layer.

    The role change: the first-line manager span is a cognitive limit, not a cost line

    A first-line manager’s direct-report span is not a matter of preference for most cognitive work. It sits around 5 to 7. Push it to 10 and managers stop coaching and start triaging. Push it to 15 and the role has reverted to being an individual contributor with a different title. Organizations that scale cleanly keep that first layer tight even when the spreadsheet says it is expensive — because the spreadsheet is not pricing the coordination tax that a wider span produces downstream.

    Minimalist line graph showing communication-channel count rising quadratically as team size grows from 2 to 15
    Coordination overhead grows quadratically with team size.

    The org change: 50 and 150 are hard boundaries

    The sub-team that actually owns a piece of work should be closer to 5 than to 10. Not because small teams are faster in principle, but because the communication-overhead curve gets steep fast after 7. Bezos was right about this, and almost every high-performing team of any reasonable size runs its real work through an informal group of four or five — regardless of what the reporting structure says on the org chart.

    When a function crosses 50 people, it needs an operational substructure. Tribes, chapters, pods, whatever the label — or the Dunbar sympathy layer breaks. When the people in a team stop knowing each other well enough that a death in someone’s family would visibly register with everyone, culture starts dying quietly. By the time anyone notices, six months have usually been lost.

    When an organization crosses 150, it runs two cultures whether the leadership admits it or not. The question is only whether the split is designed deliberately or happens by default. Organizations that handle the ceiling well accept it and build deliberate boundaries. Organizations that handle it poorly spend years pretending 400 people are “all one team.”

    Minimalist org-chart diagram with a horizontal dashed line labeled 150 separating a large unified structure above from subdivided smaller groups below
    Cross 150 and you either build deliberate substructure or get default fragmentation.

    The strategic consequence: org design is surrender, not construction

    Good organizational design is mostly a process of surrender. The cognitive architecture of the humans running the teams picks team sizes for you, and the only real choice is whether to build the org chart around what actually works or to fight it and pay the tax. Every commercial organization that has tried to force a bigger number — a 12-person manager span, a 30-person “small team,” a 300-person “family culture” — has either quietly subdivided itself into groups that look suspiciously like the Dunbar numbers, or lost the thing that made it work.

    AI augmentation does not move the cognitive ceiling. It moves the throughput below the ceiling. An IC managing four AI agents is still operating inside a span of four. A manager coordinating seven sub-teams of augmented ICs is still operating inside a Dunbar-5 layer. The numbers that governed organizational design before agents are the numbers that will govern it after.

    Round-G candidate G03-SEMI-fireteam: SEMI — fireteam of 4 around laptop + overlay (matches two-pizza warmth)
    Small intimate teams stay where the work actually gets done.

    So what boards should do

    Boards should design operating models around the Dunbar layers and treat AI-augmented throughput as a multiplier on what each cognitive unit can do — not as a license to stretch the unit past its ceiling. The specific actions sit at four layers: first-line spans at 5 to 7 even under headcount pressure; sub-team ownership at 5; operational substructure at 50; deliberate cultural boundaries at 150. These are not target numbers. They are discovered numbers. Every other structure is an argument with biology, and biology does not negotiate.

    The Roman legions did not know about neocortex ratios. The Marines do not design their fireteams around anthropology papers. Jeff Bezos did not cite Dunbar when he ordered the pizzas. All three converged on the same numbers because the numbers are a property of the animal doing the work, not the work itself. The job of an organizational designer is to notice this — and then get out of the way.

    References

    • Dunbar, R. I. M. (1992). “Neocortex size as a constraint on group size in primates.” Journal of Human Evolution.
    • Dunbar, R. I. M. (2010). How Many Friends Does One Person Need? Harvard University Press.
    • Miller, G. A. (1956). “The Magical Number Seven, Plus or Minus Two.” Psychological Review.
    • Brooks, F. P. (1975). The Mythical Man-Month: Essays on Software Engineering.
    • Hackman, J. R. (2002). Leading Teams: Setting the Stage for Great Performances. Harvard Business School Press.
    • Ringelmann, M. (1913). Early social-loafing experiments, Annales de l’Institut National Agronomique.
    • Gladwell, M. (2000). The Tipping Point. Popularised Gore’s rule of 150 for the management audience.
    • The Scrum Guide — current recommended team size: 3 to 9 members.
  • Q1 layoffs hit a four-year low. Tech’s share went up 40%.

    Q1 layoffs hit a four-year low. Tech’s share went up 40%.

    What was announced

    Challenger, Gray & Christmas reported in late March 2026 that U.S. employers announced 217,362 job cuts in Q1 — the lowest Q1 total since 2022. Within that aggregate, technology-sector cuts ran at 52,050, up 40% versus Q1 2025. In March specifically, AI was cited as the rationale for 15,341 cuts — 25% of the month’s total — making it the leading single reason for U.S. layoffs for the first time on the Challenger record. Major contributors to the technology figure: Dell’s annual filing-disclosed restructuring, Oracle’s March layoffs, and Meta’s Reality Labs reduction.

    What it means

    The aggregate-down, tech-up, AI-leading combination is not three separate stories. It is one story told from three angles. The aggregate number is down because the broad U.S. economy is operating with reasonable employment; sector-by-sector cuts in legacy industries are running below historical norms. The technology number is up because the sector is going through a structural reallocation — capital is shifting from headcount-led growth to compute-led growth, and the cost base of large software companies is being explicitly redesigned around that shift. AI is the leading cited reason because it is the strategic narrative that justifies the redesign to investors, customers, and remaining employees.

    The implication for the rest of 2026: technology-sector hiring patterns will continue to diverge from the broader economy. Companies will hire aggressively for ML, infrastructure, agent operations, and applied research while shrinking headcount in functions that AI is augmenting or displacing. Net headcount may decline, but the per-employee compute and capability budget rises sharply. That changes what “growth” looks like in the financial reporting of the sector.

    Andreas’s view

    My read on this: the Q1 numbers are not a downturn signal — they are a transformation signal masquerading as cost discipline. Tech companies are not in distress. They are restructuring around the assumption that a smaller, AI-augmented workforce produces equal or greater output at a different cost basis. Some of those bets will be right; some will be the Block experience at smaller scale, where the rehire follows the cut by six to twelve weeks. The Q2 and Q3 numbers will tell us how clean the underlying productivity gain actually is.

    I don’t think the AI-as-cited-reason metric stabilizes here. It rises through 2026. Once the framing carries an investor-relations multiple — which Block demonstrated — the disclosure pattern shifts in its direction across the sector. By year-end, AI-cited cuts will likely cross 30% of monthly U.S. totals, and that will look more like a permanent baseline than a peak.

    The way I see it: the Challenger headlines document neither a labor crisis nor a productivity victory. They are capturing a sector-wide capital reallocation with a coherent strategic logic and uneven execution quality. The more interesting question to me is which side of that reallocation any given business is on — and whether its cost base reflects the structure it has today or the structure it intends to have in 18 months.

    Three things I’m watching

    Three things I’m watching as this plays out:

    1. I’ll be watching whether companies are tracking the technology-sector comparison for their own organization: revenue, headcount, and per-employee compute spend versus the closest five public-market peers. That gap is where structural exposure shows up first.
    2. I’ll be watching whether organizations hold a meaningful distinction in their communications between AI-driven productivity reductions — workflow-modeled, with measurable output — and broader restructuring justified by other factors. The market may not differentiate; but the ones with rigorous operations will.
    3. I’ll be watching Q3 unit economics against any Q1 workforce action. The reduction is on the books in Q1; whether the underlying productivity thesis holds shows up in Q3 output measures, not headcount.

    References and related signals

  • AI is becoming the narrative for layoffs. It is not yet the cause.

    AI is becoming the narrative for layoffs. It is not yet the cause.

    What was announced

    Through the week of February 16–22, 2026, the AI-cited layoff story moved from edge case to mainstream framing. AI was cited as the rationale for 4,680 February job cuts in the U.S. — roughly 10% of the month’s total. Baker McKenzie announced 600–1,000 layoffs (up to 10% of global headcount) framed as a pivot to AI-augmented service delivery. Dow disclosed 4,500 cuts in January with explicit AI-strategy framing. A Harvard Business Review piece in the same window argued that companies are laying off based on AI’s potential, not its measured performance. An Oxford Economics report from January concluded that many AI-cited layoffs were the consequence of past overhiring, not present AI productivity.

    What it means

    Two things are happening at once. First, AI productivity is real for specific workflows and starting to show up in unit-cost reductions. Second, “AI” is becoming the public-facing rationale for cost actions that boards and CEOs have wanted to take for other reasons — overhiring during 2021–2022, deteriorating margins in slower-growth segments, restructuring to a target operating model that was already in motion. The two stories overlap, and the public communication does not distinguish between them.

    For employees, the framing matters because “we are restructuring” and “AI is replacing your role” carry different signals about whether the function comes back. For investors, it matters because the market is pricing AI-cited cost reductions as durable while restructuring-cited cost reductions are typically priced as one-off. CEOs who choose the AI framing get a multiple uplift. That incentive structure tells you why the framing is becoming dominant.

    Andreas’s view

    My read on this: the next 12 months will see a steady drift toward AI-as-explanation in layoff communications, regardless of whether AI is the underlying driver. The reason is not deception — it is signaling. CEOs need a forward-looking story that the cost base will stay reduced, and “AI productivity” is a cleaner story than “we hired too aggressively in 2022.” The public record will eventually reconcile this; quarterly earnings will reveal which companies actually shipped the productivity gain and which simply downsized.

    I don’t think the workforce numbers are yet the right metric to watch. The right metric is the ratio of revenue per employee in the months after the cut. If revenue per employee climbs durably, the AI framing was substantively correct. If it plateaus or reverses while operational quality declines, the framing was a positioning move and the company will be hiring back inside 18 months — at higher cost and lower morale.

    The way I see it: when a CEO presents an AI-cited workforce action, the productivity model behind it should be specific enough to name which workflows, which output measures, which time horizon, and which control group. Where those answers are vague, the action is restructuring with AI vocabulary. That is not necessarily wrong, but the distinction matters — and I think it matters most at the board level, where the conversation should reflect what is actually driving the decision.

    Three things I’m watching

    Three things I’m watching as this plays out:

    1. I’ll be watching whether companies maintain a clear internal distinction between AI-driven productivity actions (with a workflow-level model behind them) and AI-framed restructuring actions (justified by other reasons). Both can be valid; conflating them confuses execution, and the ones that keep the distinction clean are more likely to deliver what they promised.
    2. The companies that track revenue per employee monthly for the 12 months following any AI-cited workforce reduction will have the clearest view of whether the productivity gain actually materialized — and I’ll be looking at that number as the most honest signal in the public record.
    3. I’ll be watching how specific companies get in their external communication around AI-related workforce changes. Vague “AI is making us more productive” framing tends to erode credibility internally faster than a precise statement of which work has been automated and which has been redesigned — and over the next year, that credibility gap will start showing up in retention and hiring data.

    References and related signals