Lessons from Composability: How We Avoided Lock-In Once, and Why We Need to Again
From MACH to AGNTCY, the case for openness in the age of AI agents is more urgent than ever.
We’ve seen this movie before.
A wave of digital innovation, followed by a stampede toward the biggest platforms, then a costly realization: closed systems cost more than they seem.
The last time enterprises faced real platform risk at scale, it centered on commerce. Leaders watched innovation grind to a halt under the weight of tightly coupled suites. Roadmaps were vendor-bound. Customization became brittle. Every pivot came with a price. The response was composability: a design philosophy and operating model that gave organizations back their leverage. With the MACH Alliance serving as a neutral coalition for education and alignment, composable commerce became more than a buzzword. It became a blueprint.
Now, as AI agents explode in popularity, it’s time to ask: are we building for flexibility having finally learned our lesson, or are we walking back into the same theatre, watching the same movie, hurtling towards the same costly realization?
The Rise of Agents, the Risk of Lock-In, the Road to Take
We’re now seeing the early signs of a similar inflection point. AI agents and automation platforms are proliferating, and with them, a wave of consolidation. The field is moving fast, and the draw of full-stack solutions is strong. But speed, as we’ve learned time and again, can hide risk.
Shortcuts today can hardwire tomorrow’s regrets and if digital leaders aren’t careful, the agent layer could become the next lock-in layer: a dependency wrapped in convenience.
There is, of course, an alternative. There is a path for agents that mirrors the logic of composable commerce. By decoupling intelligence, execution, interfaces, and data layers, it’s possible to evolve AI capabilities incrementally. This isn’t a high-minded play for architectural purity. It’s strategic optionality that lets organizations mix best-in-class tools, own their automation roadmap, and reduce the surface area of vendor risk.
This is where AGNTCY enters the conversation. It’s not a vendor, platform, or product. It’s an open ecosystem initiative designed to bring standards, tooling, and community support to the emerging world of composable agents. Its mission echoes what made MACH successful: clear principles, open participation, and a commitment to keeping enterprise buyers in control. As agents become embedded in how companies operate, that control becomes non-negotiable.
Lessons from Composability in Commerce and What Composability Looks Like in AI Systems
To understand why this matters, it helps to recall what composability solved in commerce. It wasn’t just about headless architecture or API-first infrastructure. It was about reclaiming the ability to change course without tearing everything down. When a new channel emerged, you could plug it in. When customer expectations shifted, you could adapt quickly. And that adaptability translated directly into competitive advantage.
Now, as AI becomes embedded in workflows, the stakes are just as high. Agents aren’t simply chatbots or productivity hacks. They’re autonomous actors that execute tasks, handle logic, interface with systems, and increasingly, make decisions. Locking that capability into a single vendor or tightly integrated stack limits future flexibility. Worse, it creates new dependencies on opaque models and proprietary orchestration engines. Composability in this context means designing agents whose parts can be understood, swapped, upgraded, or retired independently.
What does that look like in practice?
It means separating agent memory from agent goals. It means decoupling UI interfaces from the logic layer. It means using open protocols for inter-agent communication and shared context. It means storing business-critical knowledge and instructions in transparent, portable formats. Each of these design choices makes it easier to scale up without losing control.
Leading the Ecosystem
This isn’t theoretical. AGNTCY is already working on specifications, reference implementations, and shared tools to support this kind of modular agent design. It's an open-source initiative, but more importantly, it's an open-participation model. Like MACH, it welcomes input from vendors, users, and independent contributors alike. The goal is to create a common foundation that any team can build on, without dictating business models or limiting choice.
That neutrality is key. In the early days of composable commerce, it wasn’t always clear which patterns would win. The MACH Alliance helped reduce that ambiguity. It provided vocabulary, benchmarks, and a trusted forum for discussion. AGNTCY aims to do the same for agents. It’s not a certification body, and it doesn’t seek to define winners. But it does offer guidance for those trying to navigate a fast-changing space without getting boxed in.
What Digital Leaders Should Do Now
For digital executives, the takeaway is straightforward: ask now what will be hard to change later.
Look at the agent platforms you're evaluating and ask whether they let you own your data, your decision logic, your interfaces. Ask how easy it will be to move an agent’s capabilities to another runtime. Ask whether the orchestration layer is open or opaque. Ask whether your team could fork and extend if needed. The answers to these questions are the early signals of whether you're buying optionality or deferring a reckoning.
We don't need to repeat the mistakes of the past decade. We don’t need another wave of costly replatforming. Composability taught us that modularity is not a constraint but a competitive advantage. It gave companies the ability to change without breaking. If we apply that same lens to the AI agent layer, we can build systems that scale with us, not against us.
AGNTCY isn't the solution. It's the scaffolding for solutions. And just like with MACH, its real value will come from how the community uses it—not just to build tools, but to build confidence that there’s a better, more flexible path forward. Lock-in is not inevitable. But avoiding it takes intention. We have the playbook. We just have to use 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.