• "several top Microsoft executives [during the period where they invested $10 billion in OpenAI] told colleagues they thought OpenAI’s business would eventually fail, even if its technology was good,
• As I discussed in last week's Friday premium column, OpenAI and Microsoft are currently negotiating over the terms under which it would allow OpenAI to convert to a for-profit entity, in which OpenAI is offering Microsoft...a reduction in its revenue share (down to 10%), less stock (33%, down from 49%
• Many AI experts see AGI as the point at which generative AI systems achieve humanlike intelligence, but OpenAI and Microsoft are at odds over the issue.
• OpenAI executives including Sam Altman believe they are close to being able to declare that their AI tools have achieved the AGI level of proficiency, according to people familiar with the matter.

• The Financial Times reported last week that Microsoft was prepared to walk away from its "high-stakes" talks with OpenAI, and closed their piece with this quote:

  A Silicon Valley veteran close to Microsoft said the software giant “knows that this is not their problem to figure this out, technically, it’s OpenAI’s problem to have the negotiation at all”.
• Walking away from these negotiations would effectively end OpenAI's ability to become a for-profit, cutting its recent $40 billion round to $20 billion, make it impossible to take the company public, and likely make it much harder — if not impossible — for OpenAI to raise further funds.
• To be clear, OpenAI plans to raise another $17 billion in 2027, and investors generally invest in a company because they believe they'll receive a return.
• OpenAI is Microsoft's biggest Azure customer, providing a projected $10 billion in revenue in 2025, just under 77% of Microsoft's $13 billion in projected annual revenue (not profit) from AI according to The Information
• Microsoft made $4.7 billion from AI in 2024, with $2.7bn (57%) of that coming from OpenAI
• Top executives at Microsoft believed OpenAI would die when they gave it $10 billion in funding in 2023.
• Microsoft is directly representing in leaks to the media that it is willing to block OpenAI's conversion

9 more annotations...

• Top contributors demand access to remote work. Half of the leading performers work at organizations requiring fewer than three days in the office, and 63% would accept a pay cut for more remote flexibility.
• “High performers thrive when they have the autonomy to choose the work environment that best suits their needs,”
• Two-thirds of employers reported increased retention rates as a result of their hybrid policies, with an average increase of 34%.
• employers with mandated set days achieved the highest retention rates, with one-day-per-week mandates showing a 41% increase, demonstrating that significant flexibility is important.
• While nearly half of employers believe they communicated return-to-office policies well, only 36% of employees agree.
• Even more concerning, 46% of employers think they consulted employees in planning, but just 21% of workers feel they were heard.
• While 93% of employers and 90% of employees recognize collaboration technology as essential, only half of employees feel equipped with consistent tools across locations. Jones advocates for “transformative technologies” that not only support hybrid work but also enhance it.
• suggests that HR leaders work to help “design workspaces that inspire creativity and connection
• As Jones points out, wellbeing is closely linked to organizational success when workspaces are designed to encourage collaboration and flexibility.
• flexible hybrid policies are strongly associated with better mental, physical, social and emotional health.
• Getting hybrid work right unlocks the full potential of our people and builds a more inclusive, productive way of working,” Jones notes.
• “What your people need today may look totally different in five years,” the study notes

10 more annotations...

• We don’t transform organisations simply by adding new tools, but rather by changing the art of the possible and what the organisation is capable of.
• But too often, capabilities are seen through the lens of enterprise IT architecture as being technical rather than strategic building blocks
• Even today in enterprise AI, companies often focus only on task automation or incremental improvements to existing ways of working.
• A capability is not just a tool or a process. It is an organisational superpower that brings together systems, data, services, software, and skills in order to create new potential.
• This is where orchestration becomes critical. AI service orchestration is not just about connecting systems, but coordinating intelligence and action across the organisation in ways that are coherent, context-aware, and responsive.
• Capabilities do not become transformative through technical implementation alone. They become strategic when they enable new forms of coordination, learning, and adaptability across the organisation.

• It is not just about reducing effort or speeding up processes. Instead, it becomes a means of:

  • Surfacing Signals: Detecting weak signals, blind spots, and emerging patterns from across the enterprise environment.

  • Coordinating Action: Aligning distributed teams, systems, and services around a shared understanding of what matters.

  • Enabling Responsiveness: Reacting to changes in context by dynamically adapting workflows, reallocating resources, or triggering the right interventions at the right time.
• Strategic orchestration allows organisations to operate less like rigid hierarchies and more like adaptive systems. It replaces the need for tightly managed processes with flexible, context-aware collaboration between humans and machines.
• The real shift is not about adding more AI. It is about making the organisation more coherent and capable.
• Cross-system co-ordination without process debt: most organisations rely on brittle, linear workflows that are difficult to adapt once they are embedded. Orchestration enables more flexible and context aware task execution by connecting services, systems, and human decision points dynamically.
• Real-time context-aware escalation: not all events require escalation, and not all alerts should be treated equally. Orchestration can filter, prioritise and route issues based on live context and business impact, reducing noise and surfacing signals that matter.
• Multi-agent collaboration for complex tasks: some tasks are too complex for a single AI service or human team to handle alone. Strategic orchestration enables multiple agents and humans to work in sequence or in parallel, contributing different capabilities to a shared outcome
• Adaptive service routing across teams and time zones: as work becomes increasingly distributed, orchestration can allocate tasks based on expertise, availability and business priority rather than fixed schedules or team structures.
• These examples show that orchestration isn’t about replacing any one system, bit instead is about unlocking human and system potential through better co-ordination, dynamic alignment and the ability to act with context and intent.
• Capability mapping: strategic orchestration brings together systems, services, datasets, software and skills. Mapping these components allows organisations to understand their current maturity, expose duplication or disconnects and design for modular growth.
• Loops and layers: rather than a big bang deployment, orchestration should be implemented as a set of nested loops that deliver incremental value. Initial loops might involve simple automation and visibility across workflows. Later loops can introduce agentic co-ordination, adaptive routing and dynamic escalation.

• A structured capability model can clarify what is required to orchestrate effectively:

  • Core Systems: Event-driven architecture, orchestration engines, observability platforms

  • Data Sets: Metadata, service logs, workflow telemetry, enterprise knowledge graphs

  • Software: Agent frameworks, reasoning engines, integration middleware

  • Services & Processes: API-exposed capabilities, modular workflows, adaptive policies

  • Skills: Service design, orchestration engineering, prompt development, human-in-the-loop design, AI governance

15 more annotations...

• Les critiques les plus courantes – "le modèle fabrique des fausses information", "on ne peut pas lui faire confiance" – supposent implicitement qu’<!----><!---->il devrait se comporter comme un oracle infaillible<!----><!---->.
• Dans un raisonnement déductif, l’énoncé "A implique B" établit une certitude universelle, tandis que pour un LLM, constater que "phrase A" est suivie de "phrase B" n'implique qu’une chose : c’est la poursuite statistiquement la plus vraisemblable
• a valeur d’une idée ou d’une proposition ne tient pas à sa conformité à la Vérité , mais aux conséquences pratiques qu’elle entraîne :
• une croyance ne demeure vraie que "<!----><!---->tant que les événements continuent de la rendre vraie<!----> ".
• l’opinion destinée à être finalement admise par tous les chercheurs<!---->
• est vrai ce qui "<!----><!---->fonctionne<!----><!---->",
• La "vérité" des LLM réside donc dans leur capacité à nous aider à résoudre un problème spécifique<!----><!---->, même si leur processus est probabiliste et non déductif
• une idée n’est pas vraie parce qu’elle séduit ou convainc à l’instant, mais parce qu’elle résiste aux chocs de l’expérience répétée, et une croyance reste confortable "jusqu’à ce que nous soyons<!----> <!----><!----><!---->réveillés de notre rêve flatteur par des faits rugueux (
• À vouloir enfermer les modèles de langage dans le carcan de la Vérité, nous les condamnons à l'échec, tout du moins selon nos critères de validation habituels
• guider leur évolution à partir de l’observation concrète des usages : ce qui fonctionne, ce qui échoue, ce qui apporte de la valeur ou amène des risques.
• un LLM bien intégré n’est ni un oracle infaillible ni un usurpateur de raison ;<!----> il est un "copilote statistique" qui, associé à l’esprit critique humain, augmente la portée de l’action
• Comment puis-je tirer parti de ses forces tout en maîtrisant ses faiblesses ?<!----><!---->".

11 more annotations...

• This radical overhaul of HR will not happen by simply swapping out infrastructure: Peoplesoft for Workday or Workday for Rippling—it comes from running HR like a business and thinking of HR as a product.
• oving radically away from staid bureaucracy to agile operations
• “I run HR like a business and think of HR as a product.”
• “You always have to ask yourself: Is your customer going to buy it? HR loves to roll things out that people don’t necessarily need. So how do you roll things out that they want to ‘buy?’  Are these products that will help their business?”
• mandates and policy only go so far in shaping culture, creating engagement in the workforce, excitement around learning and capability building in the company.
• For that, a product mindset—and a product manager function— is required.
• Making that trip—from a policy-led organization to a product one- requires six essential moves.

• #1. Lay a foundation to work from.

  #2. Build a product-first culture.

  #3. Establish a product management framework.

  #4. Nurture product management talent.

  #5. Curate artifacts.

  #6. Run around data.

• #1. Lay a foundation to work from.

  “It’s difficult to run the trains on time if the tracks need a lot of work.” HR as a product needs a modern HR infrastructure to run on.
• Modern HRIS systems, if well architected, can, according to Toohey, “elevate the employee experience and add value where it matters most, with our people—our customers.”
• #2. Build a product-first culture.
• A product-first culture comes from a customer-first mindset. It comes from understanding who your customers are, what they’re in business to do, and what the business is driving at.”
• Building a product-first culture isn’t just about changing processes—it’s about reprogramming how the organization thinks about value and innovation.
• working through a continuum from policy-driven, and developing habits around consistent processes, a focus on employees, and the development and use of data.
• #3. Establish a product management framework.
• Good product management takes lessons from project management, agile development, and design thinking. Whichever framework you use, it should cover two things: How you organize your work for smooth execution and how it connects with your customer to meet their needs and add value.
• Toohey talks about a flywheel: “It’s three parts, how we build, how we sell, and how we service.
• #4. Nurture product management talent.
• They have to be able to see beyond the usual constraints and ask, ‘What could we create if we weren’t limited by how things have always been done?’”
• #5. Curate artifacts.
• “Think of artifacts like building blocks for HR as a product. They allow you to standardize the way you solve problems but also provide flexibility for innovation.”
• #6. Run around data.
• When you treat HR like a product, every decision should be backed by data. Whether it's employee engagement metrics, adoption rates for new tools, or the impact of development programs, data should guide every step. You can’t improve what you don’t measure.”

22 more annotations...

• while EX has become a buzzword in the boardroom, many practitioners still struggle to articulate its business value. We're great at describing programs, initiatives, and feel-good stories. We're not great at connecting the dots to revenue, retention, productivity, or other strategic objectives.
• This gap isn't just academic—it's a threat to the credibility of EX and culture practitioners.
• Similarly, my time at McKinsey changed how I think about EX, primarily due to the ​Organizational Health Index​ (OHI).
• McKinsey's ​approach​ starts with business strategy and works backward to people practices, not the other way around.
• Instead of asking "How can we make employees happier?" the question becomes "What organizational capabilities do we need to win, and how do our people practices enable them?"
• y I think the way I do about driving EX as a product (see ​Chief of Work​) with desired business outcomes.
• I was seeking. But, to be fair, connecting EX to business outcomes is hard in any situation, much less on the spot.
• PepsiCo's real estate team enhances the workplace experience for frontline employees. The majority of their 300,000+ workforce, who work in non-office sites and directly interact with products and customers every day.
• Feedback loops include visible "You Said It → We Did It" boards so "deskless" workers know their voices matter.
• When frontline workers feel valued and equipped to perform their jobs effectively, it is reflected in safety metrics, quality indicators, and customer relationships.
• For Campari, culture work isn't about employee satisfaction—it's about deal success. With roughly 40 acquisitions completed since 1995, they've learned that preserving organizational identity while enabling agility is essential to M&A performance.
• Their approach of "distilling, codifying, narrating, and activating" organizational culture creates both continuity and adaptability.
• post-acquisition retention rates, time-to-productivity for integrated teams, and revenue performance of newly acquired brands. When you're completing a major deal every nine months for three decades, culture integration becomes a core business capability, not an HR nice-to-have.
• Now a one-month intensive program is followed by blended learning—two-thirds client engagement, one-third continuous development—with remarkable results. New hires now reach 60% productivity in month two and 80% by month three, while reducing per-employee development costs.

• The path forward isn't complicated, but it requires discipline:

  1. Design backward from business strategy. Start with the capability you need (see ​“North Star” vision​), then work toward the EX that enables it.

  • PepsiCo's frontline focus improves customer experience.
  • Campari's culture work enables M&A success.
  • Wix's learning programs deliver measurable ROI.

  2. Establish your business hypothesis upfront. Before launching any EX initiative, complete this sentence (and don’t start the program until you can):

  If we improve [specific employee experience], we expect to see [specific business outcome] because [logical connection].

  3. Measure business impact, not just employee satisfaction. Engagement scores and NPS are inputs, not outcomes. Track metrics that matter to your execs: retention costs, time-to-productivity, safety incidents, customer satisfaction, revenue per employee.

14 more annotations...

• ut “product-first mindset” isn’t just a buzzword; it’s a necessary shift.

  • From constantly responding to requests to proactively designing solutions that meet the ongoing needs of employees and the business
  • From an endless cycle of one-off fixes to building solutions that people want to use and that scale value over time
  • From HR’s current firefighting state to a future-proofing one

• HR responds to needs as they come — “Help me resolve this cross-team conflict that’s blowing up my project timeline!” — provides a quick fix, is the hero of the moment, then moves on.

  It feels productive. It feels like adding value. And on the surface, it is.
• • HR stays reactive; it’s constantly addressing symptoms rather than root causes
  • HR subteams develop their own fixes in silos, each tailored to a specific need but rarely designed to work together
  • HR spins up new processes, policies, and programs all the time — each one slightly different, none truly reusable or scalable
  • Feedback loops are weak, as HR rolls out solutions without a clear way to measure whether they’re solving the core problem
• HR saw employees as customers with persistent, recurring needs who deserved well-designed and tested solutions?
• What if HR didn’t assume people will use what’s provided simply because it exists — but instead made solutions so effective, intuitive, and desirable that employees wanted to engage with them?
• What if HR didn’t define success by how many problems got fixed in the moment but by how many problems got solved at scale — and for good?
• It incentivizes reactivity, not proactivity.
• It de-emphasizes user-validated solutions.
• It doesn’t scale.
• It’s not data-driven.
• When I was an HR person, I rarely asked my clients for specific feedback. When I did, it was service-oriented: ‘Am I a good partner to you? Do you trust me? Do you want me around?’ Not ‘Did we solve the most compelling problems your business had this year?’ Because the answer would have been no.

• It was also a flop.

  Why? Because the intended users — front-line managers — were presented with a solution they didn’t request. Their specific skill and knowledge gaps were unknown, because no one asked them; the feasibility of traveling for six off-sites was unexamined; and there was no insight into which materials would best support their learning.
• What if HR didn’t define success by how many problems got fixed in the moment but by how many problems got solved at scale — and for good
• Build for scale, not one-offs.
• Adopting a product mindset demands time and resources for planning, experimentation, and iteration, which can feel like a luxury when the reactivity loop leaves teams stretched.
• Shift from “people-people” to “product-people.”
• There hasn’t been a push to develop skills related to data collection, analytics, and user-centered design. As a result, many HR pros lack them.
• Separate identity from expertise
• Shifting to a product-first approach means suspending assumptions about what’s best for employees and instead asking them what they need,
• Taking a product-focused approach to everything your organization does can create some stark realities to deal with — because if we’re staying close to the customer and listening to what their challenges are, we’re forced to look at the fact that the thing you spend a lot of money and time on is maybe not creating the value you thought.”

18 more annotations...

• Rather than thinking narrowly and simply demanding budget cuts from his direct reports, Soren convened a meeting with his department to discuss the broader question of its purpose and sustainability.
• the recognition that there were overlooked assets in the archive that might be used for a new program. This idea eventually led to a top-line contribution of tens of millions, and reductions in workforce were avoided
• we have been researching two modes of attention that people use at work: doing mode, in which people pay narrow attention to a specific task in order to control, predict and get it done efficiently; and spacious mode, in which people pay attention more expansively, without hurry, making them more receptive to relationships, interdependencies, and possibilities
• Unfortunately, we’ve found that spacious thinking is regularly suppressed in favor of doing mode in organizations.
• employees looking to excel worry that shifting into spacious mode can be interpreted as a lack of efficiency or urgency.
• all employees have some agency to shift to spacious mode, leaders play a crucial role in making it more accessible at work because they can legitimize it as a good way to spend precious time (or not).
• Leaders and managers are the key to encouraging spacious thinking because their behaviors signal what is acceptable.
• managers are narrowly focused on achieving the next short-term deliverable which means their teams are stuck in doing mode too.
• less-confident employees kept quiet around her and declined to raise complex, important challenges that would require inquiry and exploration.
• But leaders tend to overestimate how spacious their own behaviors are.
• Leaders also tend to mistakenly believe that it’s easy for their direct reports to move into spacious mode without permission—or that it would be easy to ask for that permission.
• Focus on ideas instead of tasks.
• Megan suggested that he begin some meetings by asking, “What hasn’t gone well over the last few weeks, and what have you learned from that?” or “How have you developed your team recently?”
• Bring in novelty.
• It’s almost heretical to suggest that a meeting doesn’t have to end with actions to add to our already over-stuffed to-do lists. However, other outcomes might be far more important
• the agenda had been specifically designed to be an inquiring conversation rather than their usual action-focused drills.
• For example, different venues can inspire different sorts of conversations:
• Value and reward spacious mode
• He noted that in the high-pressured transformation project they were leading, Elaine had, on several occasions, challenged the team’s viewpoint in a meeting and forced them to consider alternative courses of action
• the team agreed that the decisions they ended up making in the spacious mode that Elaine had brought into the conversation were wiser than those they had been about to make while in doing mode
• The same behavior that in one employee may be labeled as annoying, wasteful and lazy, when practiced by someone else (especially someone who is already more powerful) can be seen as strategic and evidence of their readiness to continue moving up the managerial hierarchy
• There needs to be a rebalancing of the relationship between these modes if organizations are to thrive and perform.

20 more annotations...