The MarTech Canvas: New Composable Architecture Standard
Scott's composable canvas is the right architectural vision for the AI age. However, one of the rings need to be strengthened
A Never-Ending Cycle
There is a recurring joke in enterprise software circles. Every decade or so, the industry pronounces the end of complexity. New architecture. New paradigm. This time, things will just work together. And then, a few years later, the same people are in the same room, arguing about why their systems still don’t talk to each other.
Scott Brinker’s March 2026 report in partnership with Databricks, “The New MarTech Stack for the AI Age”, is the most serious attempt in years to break that cycle. His central argument is compelling and largely right: the rigid, vertically layered “stack” of the last two decades is giving way to a composable canvas, where a unified data foundation replaces the integration plumbing that has consumed so much of marketing’s energy and budget.
I recommend reading it. Brinker is one of the few analysts who writes with both technical precision and strategic clarity. The 3rd Age of MarTech framing is useful. The five-ring architecture model is thoughtful. The call to treat data as the shared substrate, not an asset to be moved between systems, is long overdue.
There is, however, a deep dive required at the centre of the model. But before getting to that, it is worth pausing on something the diagrams alone will tell you.
A New Solution Architecture Pattern
Look at four architecture diagrams published in the last twelve months: Brinker’s five-ring composable canvas from chiefmartec, Salesforce’s Agentforce model, and Snowflake’s Modern Marketing Data Stack. They come from different organisations with different commercial interests and different starting points. And yet they have all arrived at the same shape.
Not metaphorically. Literally. Every single one uses a concentric semicircle - data at the centre, capabilities radiating outward, execution at the edge. The rigid rectangular stack diagram, with its vertical layers of boxes and arrows, is disappearing. In its place: rings, orbits, gravitational centres.
This is a convergence of architecture. When competing vendors and independent analysts independently reach the same visual language to describe how marketing systems should work, that visual language is telling you something true about the underlying structure.
The implication for practitioners is direct and underappreciated. Solution architects who have spent their careers drawing boxes and arrows - data flowing left to right, systems stacked top to bottom - need to retrain their spatial intuition. The question is no longer “what sits above and below” but “what sits closer to or further from the data core.” Proximity to data is the new hierarchy. Composability is the new integration. And the diagram you draw of your architecture is no longer a plumbing schematic. It is a capability map.
Brinker mentions: The old model asked, “What layer does this belong to?” The new model asks, “What capabilities do I need and how easily can I swap one for another?” That shift in question is a shift in how solution design itself is practised. Architecture has moved from physical to logical. From fixed position to dynamic composition.
Every major platform in the industry has accepted this. The disagreement now is about who occupies the centre.
Where I Align with the Report
Before the (little) gap, credit where it is due. The report’s diagnosis of the current state is spot on. Enterprise martech stacks have grown organically - Brinker’s phrase “organic in the same way one might refer to an overgrown yard” is exactly right. The accumulated integration debt is staggering. According to the report’s own survey data, integration remained a top-three challenge for the majority of respondents even in late 2025, after decades of vendors promising to solve it.
The composable canvas addresses this structurally. By treating data as the shared substrate rather than something to be extracted, transformed, and loaded between isolated systems, the model eliminates the O(n²) integration complexity that has plagued best-of-breed stacks. Every capability that plugs into the unified data foundation does not need its own bespoke connection to every other capability. Complexity approaches O(log n). That is a generational improvement.
The report’s treatment of the semantic layer as “the keeper of coherence” is particularly important. Without shared definitions, a unified data foundation is just a larger swamp. A semantic layer that standardises what “customer,” “engagement,” and “value” mean across the organisation is what makes the canvas coherent rather than chaotic. Bryce Peake from Domino’s captures this intent in the report when he says they organise their semantic layer by decisions rather than by org chart structure. That is a decisioning-first posture - and it is the right one.
The introduction of context graphs is also one of the report’s most original contributions: a living record of decision traces that captures not just what happened but why it was allowed to happen. It is a concept that decisioning practitioners have long needed a name for.
All of this is well-argued. But the report’s most important ring - the fourth one, labelled Decisioning - is also not that well developed. I hypothesise that this is because of the limitation of scope and balancing the overall narrative.
The Fourth Ring Is an Invitation
In Brinker’s five-ring model, decisioning sits between context-as-a-service platforms and the outer ring of apps and agents. It is described, correctly, as where AI decisioning engines and reinforcement learning models optimise next-best-actions, and where orchestration controls resolve contention when multiple agents want to reach the same customer simultaneously.
That description is accurate. But across 33 pages, decisioning receives perhaps two paragraphs of substantive treatment. A report of this scope and ambition cannot do everything - and Brinker is explicit that he aims to set a north star for martech architecture, not to map every discipline within it. That is a reasonable scope decision.
What it creates, though, is a conspicuous opening. The canvas is beautifully specified. The decisioning layer that determines what gets painted on it is not. And that gap matters because the canvas metaphor, for all its elegance, is neutral about outcomes. A composable architecture does not determine which next-best-action is actually best. A unified data foundation does not resolve the question of which signals should drive which interventions for which customers in which context.
Those are decisioning questions. And they are not answered by better data plumbing. The history of MarTech is littered with organisations that had excellent infrastructure and poor outcomes that could segment customers in real time, but had no coherent logic for what to do with those segments.
There is a version 2 of this work waiting to be written - one that takes the composable canvas as its foundation and asks: what does decisioning discipline look like when it operates on this substrate? How is decisioning governance distinct from data governance? What does it mean to treat decision tracing as a first-class architectural capability rather than an audit afterthought? When should decisioning run as a standalone service rather than being embedded inside individual apps and agents?
These are not minor extensions. They are a discipline in their own right. And the composable canvas is precisely the right foundation from which to develop them.
Three Things the Canvas Still Needs to Decide
For practitioners building on Brinker's framework today, three decisioning gaps need filling - regardless of which platform occupies the data core.
Decisioning governance, not just data governance. Data governance answers: is this data accurate and accessible? Decisioning governance answers a different question: who is authorised to make which decisions, by what logic, and with what accountability when outcomes diverge from intent? An organisation can have impeccable data governance and still have twelve different teams applying twelve different decisioning logics to the same customer.
Decision tracing as a first-class capability. Brinker’s context graph - a living record of why decisions were made, not just what happened - is one of the report’s most original ideas. Organisations that systematically capture decision rationale will compound their decisioning intelligence faster than those that do not. That deserves to be a headline architectural principle, not a footnote.
Decisioning as a standalone service, not an embedded afterthought. Brinker rightly notes that decisioning may be embedded in apps and agents or run standalone. That flexibility is architecturally sound but strategically dangerous if organisations default to embedding. When decisioning logic lives inside individual channels and agents, it fragments - replacing integration debt with decisioning debt.
The MarTech Canvas Is Ready. The Decisioning Discipline Is Next.
The convergence across Brinker, Salesforce, and Snowflake is significant. When organisations with entirely different commercial interests all arrive independently at the same architectural shape, that shape is no longer a vendor preference. It is a structural fact about how modern marketing systems need to work.
Brinker has mapped the terrain with precision and generosity. The composable canvas is the right north star. Data at the centre. Capabilities are composed around it. Agents executing at the edge. This is where the industry is heading, and the report makes a compelling case for why organisations should orient their technology decisions around it now rather than waiting for the destination to become obvious.
What the canvas needs next is a practitioner-owned decisioning discipline to claim its fourth ring properly - one that is as rigorously specified as the data core, as independently governed as the semantic layer, and as portable across platforms as the composable architecture itself. That work is a natural extension of Brinker’s framework.
The New MarTech Stack for the AI Age is available at chiefmartec.com. This essay represents the independent analysis and perspective of MarTech Square. No vendor has reviewed or influenced this commentary.
Upcoming Webinar
Next week, I am joining Jonathan Moran from SAS to explore how real-time, AI-assisted decisioning closes the gap between customer data and meaningful customer experiences - and what organisations need to do differently to get there.
Transform Customer Data Into Real-Time Marketing Decisions Tuesday, 31 March · 1:00 PM ET · Hosted by SAS and ADWEEK
If decisioning is on your agenda this year, this is worth your time. Register here
let's manifest a Martech Square + ChiefMartec collab, covering the decisioning piece more in-depth!