Architecture is the AI Strategy
What the MACH Alliance's new playbook gets right, and what it means for the decisions you're making right now.
If your enterprise AI pilots are underdelivering, the instinct is to look at the obvious suspects: the model you chose, the vendor you bought from, the use case you started with. The MACH Alliance's new Build to Move playbook suggests you're looking in the wrong place.
The real variable isn't the AI, it's the architecture underneath it.
That might not be a comfortable argument for organizations that have spent the last two years running experiments and building internal AI functions, but the data backs it up. According to the MACH Alliance's 2026 Enterprise Technology Report—drawn from 600 decision-makers across seven global markets—organizations with composable architecture are six times more likely to achieve clear AI ROI than those without.
That gap doesn't close by picking better tools. It closes by building differently.
The two-layer problem
The playbook draws a useful distinction between two things that are often conflated: foundations and the agent ecosystem.
Foundations are the open, composable, connected systems that power modern digital experiences. The agent ecosystem is what comes next: many agents, from many vendors, coordinating across the enterprise to do useful work. The playbook is clear that these aren't sequential phases, they're interdependent. Without solid foundations, agents have nothing reliable to plug into. Without an eye on the agent ecosystem, you're building foundations that stop short of the value they could deliver.
This is where a lot of enterprises are quietly stuck. They've made meaningful progress on composable architecture in parts of their stack and they've also started running agent pilots. But the two tracks haven't connected, and the pilots are hitting the same wall that composable work was supposed to remove.
Jason Cottrell, CEO of Orium and President of the MACH Alliance, puts it plainly saying, "We keep hearing that AI is the strategy. It's not. The architecture is the strategy. AI is just what runs on top of it, and right now, most stacks can't support what people are trying to build."
The implication is worth sitting with. If you ran a composable commerce initiative two or three years ago, you weren't just modernizing your digital experience layer. You were—whether you framed it this way at the time or not—building the foundation your AI strategy would eventually depend on. The organizations that understood that connection early are the ones showing up in the seventy-eight percent.
What open, composable, and connected actually mean for agents
The playbook's three principles—open, composable, connected—are familiar territory for anyone in the MACH ecosystem. But it's worth examining what each one means specifically in an agentic context, because the stakes are different.
Open isn't just about avoiding vendor lock-in anymore. When a non-human identity needs to authenticate, act, and be held accountable, "open" means documented behavior, observable outputs, and explicit data portability. An agent operating in a black box isn't open; it's a liability.
Composable means agents can rearrange capabilities to fit a task without requiring a replatform to do it. Modularity and independent deployability aren't architectural preferences in an agentic world. They're operational requirements. An agent that can only work within a tightly coupled system isn’t an agent at all. It’s an expensive automation.
Connected is where the real-time requirement becomes non-negotiable. Agents coordinating across five systems simultaneously can't be working off batch data. They need real-time event interfaces, standardized cross-system protocols, and the ability to act across vendor boundaries without custom integration work for every handoff.
Together, these three principles describe not just a technology posture but an accountability posture. The enterprises that will deploy agents reliably are the ones that can see what their systems are doing, change them without breaking everything else, and trust that they're operating on current information.
Where most enterprises actually are
The playbook offers a three-tier maturity model: Curious, Active, and Pro. It's a useful frame, but it's worth being honest about what these tiers look like in practice, because there's a meaningful gap between where organizations place themselves and where they actually are.
Most enterprises with any composable momentum will self-identify as Active: modular in motion, APIs in place, some legacy dependencies being phased out. And in a narrow technical sense, many of them are right. But Active, as the playbook defines it, also means at least one AI or agent-driven feature in production, and a shared agent enablement layer with scoped credentials being stood up. By that measure, a lot of organizations that feel Active are still mid-Curious.
That's a crucial calibration point. The maturity gap matters because it determines which first moves are actually available to you. Cottrell, characteristically direct about this dynamic, says, "Most organizations think they're further along than they are. They've done some composable work in one part of the stack and called it done. But agents don't care about your roadmap. They need the whole foundation."
The practical implication: before you invest further in agent capabilities, it's worth pressure-testing your foundation honestly. The playbook suggests a concrete starting sequence: encapsulate legacy systems with an API and event layer first, replace capabilities behind that layer one at a time, and use the next contract renewal to reset vendor expectations even if you're not switching. It's unglamorous work. It's also what separates the organizations that will scale agents from the ones that will keep explaining why pilots didn't land.
The posture shift
Build to Move has an important piece of framing to it that deserves more attention than a single playbook can offer: the idea that "build to move" is an operating posture, not a project outcome.
Most enterprise technology programs are still designed around a fixed end state. You scope the transformation, execute it, stabilize, and maintain. That model doesn't hold in an environment where the agent landscape is shifting monthly, where new protocols are being standardized in real time, and where the competitive implications of architectural decisions are compressing from years to quarters.
Unlike in the past, leading in the agentic era won't mean having the most sophisticated platforms. Instead, it’ll depend on being able to manage—and excel—in an environment of continuous change. Organizations with governance models that support decentralized decision-making, vendor relationships that preserve optionality, and teams structured to absorb and act on new capability as it arrives will thrive.
Cottrell puts a fine point on it: "This isn't a project you finish. The companies that are going to win in the agentic era are the ones that accepted that early, and stopped designing for a fixed end state."
The Build to Move playbook won't tell you which platform to buy or which agent framework to adopt. What it does, more usefully, is name the conditions under which any of those choices will actually work. Architecture first, posture over project, and build for the change you can't yet see.
The 6x gap is real, but so is the path to closing it.
Leigh Bryant
Editorial Director, Composable.com
Leigh Bryant is a seasoned content and brand strategist with over a decade of experience in digital storytelling. Starting in retail before shifting to the technology space, she has spent the past ten years crafting compelling narratives as a writer, editor, and strategist.