The cure is a thirty-year reflex
The paper's remedy is to build "one shared representation and semantic framework of the whole enterprise," a business blueprint that matures, over time, into a digital twin of the organisation. Everyone then sees the same reality, and the enterprise can behave as one: perceiving, deciding, and acting together.
It helps to be precise about the phrase that sits underneath this. A single source of truth is one canonical reference held to be wholly accurate. In data management, that is often sensible, and standardising shared knowledge for common and repeated use can legitimately produce one. The architectural reflex goes further. It treats one reference, one model, one semantic centre as the place from which the whole enterprise is meant to understand itself. At that point, a single source of truth begins to mask the legitimate differences in work, responsibility, context, and knowledge that the enterprise depends on.
That instinct has defined enterprise architecture for three decades. If the parts are disconnected, build a more complete picture of the whole and coordinate everyone to it. The picture grows richer, the model grows more connected, and coherence is expected to follow.
It does not follow. The reason is hidden in the word the paper chose.
Cohesion is not coherence
Cohesion is parts held together: bonded, attached, moving in concert. Coherence is the parts that make sense together while remaining distinct. These are different conditions, not two words for one thing.
A tightly cohesive organisation can still be incoherent. Worse, the drive to act as one can deepen incoherence, because it presses heterogeneous practices toward a uniformity that suppresses the differences coordination depends on. Coherence does not require perfect consistency or total agreement. It tolerates tension, trade-offs, and local conflict, provided the whole remains intelligible and can keep working. Cohesion pulls toward sameness. Coherence holds differences together. This is not a preference for fragmentation: holding differences together is active work, keeping distinct sources grounded, related, and translatable, so that the whole stays intelligible rather than splitting into isolated versions.
The paper aims at cohesion and offers it as the cure for fragmentation. But cohesion neither guarantees coherence nor is free of the risk of making things worse.
Three ways of reaching a centre
The paper's three proposals share a common assumption: that coherence is achieved by establishing a centre from which the enterprise can be understood, coordinated, or directed. A shared model is a centre from which the enterprise can be understood. The enterprise mindset is the mindset of someone who holds the whole in view. Leadership is an authority that designs cohesion into place. The assumption is rarely stated, yet it does the explanatory work: coherence is expected to follow once the centre is in place. The Interweaving claim is the reverse. Coherence does not originate at a centre. It emerges from grounded, translatable relations among situated sources, and that holds in all three cases. Take them in turn.
The first centre is representational. In the paper, a shared model serves as the single source of truth, and coordination means conforming to it. Interweaving locates coherence differently. The alternative is not the absence of models; it is multi-source coherence: many situated sources, each of them partial, whose relations are grounded and made intelligible enough to coordinate. These sources are not rows in a repository. They are people doing work: carrying responsibilities, making decisions, serving customers, operating systems, and managing risk. Consider a single idea, the customer, as it lives in sales, in finance, and in support. Each function holds a legitimate version shaped by the work it does. The discipline is not to flatten the three into one record and call the result truth. It is to keep them related and translatable, which is why this series treats translation between domains not as a cost to be removed but as work to be done well.
The second centre is personal. In the paper, the enterprise mindset describes the individual who connects the dots, moves between the big picture and the detail, and sees how the whole fits together. Saying that everyone can hold this mindset does not remove the figure; it distributes it. The enterprise still depends on someone seeing the whole. Interweaving does not. The paper undercuts the personal centre itself when it praises self-managed organisations such as Morning Star, where coordination runs on commitments and requests rather than on a central model or a person who sees everything. Those are among its strongest examples, and they are examples of coherence holding without anyone at the centre.
The third centre is authority. In the paper, the decisive movement toward cohesion belongs to leadership: leaders define the aspiration, author the design, and drive the change. Leadership matters, and this is not an argument against it; it is an argument against the causality. In the Interweaving view, leaders can strengthen the conditions under which coherence emerges, but they do not manufacture coherence the way an architect authors a building. The orchestra metaphor the paper uses gives away the assumption: an orchestra has a fixed score and a conductor, and an enterprise has neither. Its score is contested and drifting, which is the very condition that makes coherence hard.
The question the picture cannot answer
There is a practical point at which this reflex breaks down. The paper offers a list of fifteen perspectives an enterprise might bring into its model, then concedes it has no principle for choosing among them: start with a subset, and more will be brought in over time.
That is scope inflation stated openly, with no answer to it. It is the oldest unsolved problem in enterprise architecture: which elements belong, and why. A more complete picture does not answer the question. It postpones it.
This is where architecture has a discipline to offer. Relevance is not decided by the completeness of a model. It is decided by anchoring. An element is architecturally significant when agents depend on it in their actual practice. Anchoring is a selection criterion. It tells you what belongs by looking at where the work happens, not by attempting to picture everything and hoping the important parts survive.
What AI does to the picture
The paper is right that AI raises the stakes. But the stakes move in the opposite direction to the one the paper assumes. AI accelerates the production of models and the drift of meaning between them. The remedy of building one ever more complete model, then maturing it into a digital twin, is the move AI puts under the greatest strain, because it pours more modelling onto meaning that was never grounded.
The digital twin is useful. It is not sovereign. In Interweaving terms, it sits at the far end of a line of sight that begins with agents and their practices, passes through the Lens that decides relevance, and through the forms in which architectural knowledge actually exists, explicit, embodied, and embedded, before it reaches any enterprise model at all. The twin is the concrete pole of that model. A twin that is not anchored in practice and not grounded in meaning is a faster, prettier abstraction, and AI makes it even faster.
What becomes possible
This is worth stating as a capability, not only as a critique. The value of multi-source coherence lies not in its tolerance for fragmentation. Its value is that it lets an enterprise coordinate across differences it cannot remove. Sales, finance, and support keep the versions of the customer that their work requires, and the enterprise still coordinates where coordination matters, because those versions are grounded, related, and translatable rather than collapsed into one record. Anchoring gives architecture a principled answer to a question it has long struggled to settle, which things belong and why, without waiting for a model that is never complete. And the pressure AI applies changes character: completeness was already out of reach and is now plainly impossible, yet coherence remains within reach, because it never depended on completeness.
The harder, better problem
The task is not to picture the enterprise more completely. It is to make the enterprise coherent without a single sovereign picture, a single central knower, or a single author of the whole, by anchoring architecture in practice, grounding the meanings on which strategies and decisions depend, and keeping many situated sources intelligibly related across boundaries. The objective is not one picture of the enterprise. It is an enterprise that can still act coherently when no single picture is enough.
If the enterprise can never be fully captured by a single model, language, or vantage point, what would it take to make it coherent anyway?
/Anders W. Tell
Reimagine Architecture
through Reimagining Architectural Understanding
using Interweaving
Related articles in the Coherence Series
Related articles in the Coherence Series
