From apps to infrastructure: martech's most important migration

For most of the SaaS era, martech platforms won by shipping applications.

That sounds obvious, almost tautological. Software companies sold software. But more specifically, they sold opinionated software: a particular interface, a particular workflow, a particular set of abstractions about how marketing should work.

Here is how campaigns should be built. Here is where segmentation happens. Here is how journeys are orchestrated. Here is the cockpit. Climb in.

That model created enormous value. It helped standardize a still-young discipline. It gave marketers and marketing ops teams not just tools, but operating models. For years, that was exactly what the market needed.

Until it wasn’t.

Over the past few months, I’ve had a number of conversations with senior marketing ops leaders that all pointed in the same direction.

People who once lived inside marketing automation platforms such as Braze, HubSpot, Marketo, and Salesforce now talk about them differently: as back-end services. Useful. Important. Still very much in the stack. But increasingly not where they spend their day.

Instead, they’re building their own apps and agents on top of them.

At a recent dinner I hosted, one head of marketing ops described orchestrating highly personalized campaigns through custom agents layered on top of Hightouch, with flexibility no commercial app had given her. Another mentioned several SaaS tools he’d simply stopped renewing — from lead routing to revenue intelligence. Not because they were bad, but because what his team could now build was more precisely tuned to their data, their workflows, their business logic.

These aren’t rogue vibe-coders recklessly rolling their own. They’re experienced marketing ops professionals, serious about data quality, reliability, and compliance.

The same pattern surfaced in the customer interviews I did for The New Martech Stack for the AI Age. Marketing leaders described custom apps and agents that outperformed their commercial tools, often by a wide margin. The reason was the same every time: the apps were theirs. Tuned to their specific data, context, and competitive reality in ways that horizontal packaged software couldn’t deliver.

This isn’t a story about martech platforms dying though.

Jason Lemkin of SaaStr recently wrote that a year ago, Salesforce was essentially shelfware for his company. Still paying for it, barely using it. Then his team went all-in on AI agents, and Salesforce became, in his words, “the most important piece of software we run.”

Not because Salesforce had suddenly transformed, but because Jason’s growing fleet of AI agents needed somewhere to live. A central hub where data flows in, context gets shared, and autonomous systems don’t step on each other. In that framing, Salesforce isn’t acting like an app anymore. It’s acting like infrastructure.

Jason’s advice to executives weighing which CRM to buy: “The CRM decision is no longer a CRM decision. It’s an AI infrastructure decision.”

For most of the SaaS era, the value of most martech platforms was in the application layer: the vendor’s point of view, embodied in windows and workflows, about how marketing should work. That was the product customers were buying.

But as teams increasingly spin up their own apps and agents — expressing their own opinions — the packaged application becomes just one way to deliver utility. Sometimes still a very good one. But no longer the unquestioned center of gravity.

What matters more is the shared foundation underneath: the infrastructure that lets all those apps and agents work together safely, coherently, and at scale. Data. Identity. Orchestration. Governance. Observability. Semantics. Context.

In other words, the value of martech platforms is shifting from presentation to plumbing.

That may sound like a demotion, but it isn’t. Plumbing is what makes the whole building livable. (Trust us on this one.)

Aaron Levie, CEO of Box, recently toured a group of enterprise CIOs and came back with a pointed observation: “Headless software dominated my conversations. Enterprises need to be able to ensure all of their software works across any set of agents they choose. They will kick out vendors that don’t make this technically or economically easy.”

Kick out. Not migrate away from. Not evaluate alternatives to. Kick out.

For martech platforms, those are the stakes. And it reframes the question marketing and marketing ops leaders need to be asking. Not “which application should we buy?” but “which infrastructure can we build on?”

For years, build vs. buy was framed as a binary. Either your team cobbled something together from scratch — expensive, risky, slow — or you bought a packaged application and adapted your workflows to it. Most bought.

It isn’t build or buy anymore. It’s build and buy — at different layers of the stack. Buy infrastructure. Build tailored apps and agents that run on top of it.

Think of it this way: do you want to be a painter or a canvas manufacturer?

Most companies want to paint. They want to create workflows, automations, agents, and experiences that reflect their own strategy, their own data, their own customers. They don’t want to weave their own canvas or grind their own pigments. They want good materials from the art supply store — and then make something distinctly theirs.

You’ve probably seen hot takes on this, that building your own apps and agents is like churning your own butter. Pithy. Catchy. Wrong.

AI builders aren’t butter churners. They’re chefs. They buy their butter — and flour, and a professional oven. What they’re no longer buying is the frozen dinner: someone else’s opinion about what the meal should taste like.

Companies don’t want to build their own infrastructure.

Infrastructure is less sexy than apps.
Infrastructure offers less differentiation.
Infrastructure is riskier to build and slower to prove out.
Infrastructure requires deeper engineering expertise to get right.

Great infrastructure makes everything that companies build on top of it better.

Your apps and agents are more secure and robust.
They’re more reliable and stable under load.
They’re more performant at scale.
They’re more integrated across your organization — because they’re all drawing from the same foundation.

This is why there’s enormous commercial opportunity in providing it — possibly more opportunity than there used to be in packaging horizontal apps.

Because as that infrastructure becomes more accessible and easier to leverage, more teams will build more things on top of it — making that infrastructure more valuable.

It’s a kind of Jevons Paradox applied to platforms: openness compounds. In a recent conversation with Ravi Dodda, CEO of MoEngage, he reported their usage significantly increased when they opened up their platform to AI agents via MCP. It enabled more people to do more things with their software, unlocking more value. That’s the model.

For marketing ops leaders, this shift is mostly opportunity. For incumbent martech vendors, it’s considerably more complicated.

Their product identity, pricing, packaging, and economics were largely built around the application layer. The interface was the product. Reorienting around infrastructure isn’t a roadmap update. It requires changing the fundamental story a company tells about its own value.

That’s a worldview transplant. And not everyone will complete it.

Dharmesh Shah, co-founder and CTO of HubSpot, recently framed the direction plainly: “Being agentic is not just about agents running on our platform, it’s about agents running our platform — being able to operate it.” And: “Every B2B software company is (or should be) building an agentic version of their product. One that can be used programmatically by agents, not just interactively by humans.”

That’s infrastructure thinking. The screen becomes optional. The API becomes essential.

But this also requires a different kind of platform worldview. Not the old ecosystem thinking, where success meant other applications sort of playing nicely with yours. Something more expansive.

If your platform is genuine infrastructure, you’re not curating a partner gallery. You’re enabling a Cambrian explosion of apps and agents you’ll never see coming, built by customers and partners for use cases you’d never have thought to package.

That’s what context-as-a-service (CaaS) is really about — not a feature, but a posture. The platform’s job is to make context available, reliable, and composable for whatever gets built on top of it. (See my earlier articles here and here.)

It also opens up new horizons of opportunity for agencies and service providers who pair creative and strategic imagination with the technical craft to render that vision into hypertailored, competitively differentiated apps and agents for their clients. For those willing to develop that craft, this is springtime.

Trust me on the plumbing,

Scott

P.S. Speaking of what’s being built, Frans Riemersma and I are releasing the brand new State of Martech 2026 report and the updated 2026 Marketing Technology Landscape at #MartechDay on May 5. Free to attend, live or on-demand. If you want to be first to get them, register here.