If #AI agents can instantly compare, switch, and optimize across services in real-time, then brand, network effects, and even embedded integrations lose power.
The classic strategies for retaining users—whether through a superior product, first-mover advantage, or viral loops—could become temporary at best, requiring companies to constantly re-earn their relevance.
A handful of questions sit at the center of what I’m trying to figure out right now with Through The Boundary:
- On which layers does AI actually replace coordination, and on which does it make structure *more* necessary?
- What is the economic foundation of coordination languages, and why do some ontologies survive while others rot?
- How does semantic context engineering become an organizational capability, not just a technical one?
- What can create sustainable advantages in this age, and what should be shared openly in the Commons?
- When structure replaces hierarchy, what holds identity together?
- What breaks first when we fail to find shared meaning with each other, and what possibilities do we miss by doing so?
To make any real progress on these, I have to keep reading, listening, and pattern-matching across very different conversations. I’ve been doing this kind of curation for almost 15 years; for me, it isn’t just about collecting links, it’s a form of applied research. So I might as well share what I found most relevant.
In the carousel below, I’ve pulled 6 essential perspectives from the last few weeks of research.
Sources and references:
SaaS commoditization means standard software is no longer a differentiator. Its value can only be captured when it is absorbed into ultra-personalized business contexts.
The rise of "vibe coding" takes this to its logical conclusion, driving the emergence of organizational microcells. Small teams and individual operators can now author their own bespoke operational infrastructure on demand, rather than force-fitting their workflows into enterprise platforms. And that’s actually *needed* to anchor the value that vibe coding brings about.
When AI removes the traditional software development superstructure, the organizational overhead needed to instantiate a business capability drops to near zero, but you need:
- enough autonomy and modularity
- good enabling systems, scaffolding, and constraints
to let teams capture the value that these tools bring to your organization.
If you have these, the strategic bottleneck shifts entirely. When technical execution is heavily commoditized, problem definition and context design become the ultimate differentiators. The critical capability is no longer knowing how to build, but articulating boundaries, rules, and user experiences clearly enough for teams to interact with agentic software to materialize them.
PS: interesting data below on vibe coding spend. Replit, Framer, and Lovable… we went from a weekend hackathon to a startup prototyping standard to an enterprise workflow in 18 months.
During the last few months of @Boundaryless_ client work with enterprise orgs, I’m witnessing the dawn of the new organizational dysfunction that @zetalyrae warns about: a few overwhelmed humans are now cognitive bottlenecks for the whole org.
This is what’s happening: when organizations use AI to eliminate coordination superstructure (the layers of middle management, process coordinators, liaison roles, etc.), they assume that -amongst other benefits- the cognitive work of those roles just... disappears.
No! It concentrates.
The AI handles tasks, but not the judgment, context-holding, and integration work that coordination roles quietly performed.
Good news: as Borretti himself points out, the returns to deep judgment and real knowledge are now higher than ever. The humans who can hold context across the org become its highest-leverage nodes.
At the same time, this is exactly the pressure behind the need for Context Engineering: i.e., making knowledge and intent architectures explicit so that this activity of stewarding, understanding, and directing doesn’t concentrate in a few people. You can have a bunch of context gardeners, but they shouldn’t make every decision or set every direction.
P.S.: turns out the highest-leverage cognitive node of an org is allowed to close the laptop and look up every once in a while. Madrid sky as evidence. “Touch grass” is a very important part of the job description now :)
A few weeks ago Jack Dorsey named the two skills he thinks will stay durable:
> judgement - knowing what actually makes sense to do
> accountability - how we deliver real value to a customer or user
I've been sitting with it for a few weeks, and every client conversation on the people-side keeps landing in the same place: which skills are actually durable in the Autonomous Age?
Dorsey's frame is a very good starting point, but it leaves the operator with nothing super actionable.
How we frame it with our clients at @Boundaryless_: as the Autonomous Age progresses, an organization collapses into pure entrepreneurship (i.e., judgement and accountability at their highest density).
For this to happen, the skill of “running a business” has to be distributed across the org. This means separate capabilities with separate P&Ls, stated value propositions, and service level agreements.
Durable skills are the ones that let you exist as a separate node, and let the org around you be made of them.
A new issue of Through the Boundary is out: in the last few weeks I sat down separately with @nielspflaeging author of Organize for Complexity, and @alexeykri , co-author of Org Topologies and 10XOrgs. They emerged from different problem spaces, use different vocabularies, and draw different diagrams.
And yet when I put their ideas side by side next to Team Topologies, DDD, Wardley's ILC, and Haier's Rendanheyi (and our 3EO Toolkit), the same three-way distinction appeared in every single one.
1. Core units that face the market and carry the strategic risk.
2. Supporting units that serve them with something distinctive.
3. Generic commodity units that could be outsourced but aren't.
Basically: anybody who has been putting out genuinely important ideas on how we organise work in complex organisations in the last fifteen years has reached this skeleton.
I feel like this is not a coincidence.
The skeleton is there because these three economic relationships (with the market, with internal distinctiveness, and with commodity overhead) are structural features of any organisation above small-startup scale.
Their vocabulary diverged because each school grew inside a different community, but the convergence was always latent.
What's been missing across all six is the substrate beneath the skeleton: a contract grammar that specifies what units actually exchange with each other, and not just how they communicate. We need to know who pays whom, who shares upside, and who carries the investment risk. That's what Boundaryless has been formalising, and it's the second half of this essay.
If you find yourself reaching for Team Topologies' stream-aligned team in one conversation and 3EO's micro-enterprise in the next, read it as a confirmation that, yes, there is a convergence! :)
P.S. the long-form version of this argument will appear in a forthcoming whitepaper. This piece is a practitioner-level note; the whitepaper will do the long-form work and gives fuller context.
→ https://t.co/Wty1fPVqau
The reading list:
1. BCG Henderson Institute – Beyond Tomorrow: Four Scenarios for the World of 2050 https://t.co/hWgpj80fie
2. McKinsey & Company – State of Organizations 2026 https://t.co/Kx5tgVo7GB
3. Andy Matuschak – Apps and programming: two accidental tyrannies https://t.co/942ix9zEpr
4. Venkatesh Rao – AI in World Machine Theory https://t.co/gb8UQBvTqC
5. What is Artificial Experience (AX) https://t.co/E2WiBNnRMm)
6. The Culture of AI Engineering (Pace Layers) https://t.co/E1kRai7OPc
Mainstream strategy is finally realizing that architecture is destiny. Recent reports from BCG and McKinsey clearly point to modularity and deep flexibility as the invariant strategies for an uncertain future - yet they consistently stop short of providing the execution blueprint.
To bridge this gap, we have to look across disciplines.
This curated reading digest from my latest newsletter connects macro-global scenarios with the deep mechanics of software architecture, pace layers, and AI context engineering. It’s a synthesis of how we must design organizations and product portfolios to remain coherent in a chaotic, AI-infused age.
All links are available in the first comment.
The Bhagavad Gita you see on my desk has become a permanent read lately. I got back to it after a recent strategic session, when an Indian client dropped an idea from it that sent me straight back to its pages.
On rereading it, a specific idea struck me with incredible force:
"It is better to live your own destiny (dharma) imperfectly than to live an imitation of somebody else's life perfectly."
For decades, the corporate world has done the exact opposite. Companies mechanically installed (or at least tried) external templates: the Spotify model, generic Agile frameworks, Rendanheyi… all copy-pasted operating systems from other successful companies.
AI is destroying the copy-paste playbook.
Artificial Intelligence is extraordinarily good at reproducing the statistically legible. Consequently, organizational conformity is losing all strategic value. If your company is built primarily as a clean implementation of an external playbook, you are transforming into interchangeable infrastructure.
In a world of commoditized intelligence, strategic value shifts entirely toward something far more difficult to reproduce: your unique organizational metabolism (your local constraints, cultural frictions, and historical context). In a healthy ecosystem, every node must contribute something non-redundant. It follows that a copy pasting doesn't add much value.
Good news: real transformation never behaved like a software installation anyway (even pre-AI); organizations have to evolve organically through internal tensions, friction, and localized negotiation.
Stop trying to master someone else’s dharma perfectly.
Focus on building your own.
I’ve now published the first three issues of Through The Boundary, the newsletter where I work through my research questions as I try to understand how to align strategy and organizational design in the age of AI.
Most writing about AI and organization falls into two -disappointing- camps: either it extrapolates efficiency gains without questioning how the value frame is changing, or it offers a cheap critique built on superficial foundations.
Through The Boundary is different.
So far, I’ve explored:
(link to the individual essays below)
[1] What is an organization today (good news: I have a working definition!)
[2] Why "we want to be the platform" is the right but incomplete answer in markets that are opening up.
[3] How context engineering is becoming the discipline of authoring the concepts, boundaries, and constraints that let agentic work happen.
The essays strive to have a high signal/noise ratio. If these are themes you are actively wrestling with inside your organization and are drowning in superficial “content” around these themes, consider subscribing.
https://t.co/ZY61P3Az03