The SaaS Moats Are Crumbling, but the Opportunity Is Bigger

SaaS stocks are getting hammered. The pundits are writing eulogies. And the argument fueling both goes something like this: why pay for software when AI can just build whatever you need on the fly? It’s a seductive narrative — part populist revolution, part techno-utopian fantasy. And like most seductive narratives, it’s just plausible enough to be dangerous.

Let me be clear: major SaaS platforms are under real pressure, facing disruption on multiple fronts simultaneously. But the idea that everyone will simply conjure their own bespoke software from AI prompts is the enterprise equivalent of “I’ll just cut my own hair — how hard can it be?” (Spoiler: it’s always harder than you think, and the results are rarely pretty.)

The truth is more nuanced, more interesting, and frankly more consequential for anyone building or buying marketing technology today. Software isn’t dying. But a lot of SaaS companies might be. The difference will come down to whether they can recognize that the thing they're selling isn’t what they thought it was — and reinvent accordingly.

Before we get to where SaaS is heading, it’s worth being honest about what’s not working anymore. Three forces are eroding the traditional foundations of SaaS value faster than most incumbents want to admit. A fourth force could actually save them — but it demands a kind of openness that many SaaS companies profess but few practice.

For years, SaaS companies competed on features. Ship a better dashboard, a slicker workflow, a cleverer automation, and you’d earn a spot on the shortlist. That playbook is breaking down.

When AI can generate functional software components in minutes, the half-life of a feature advantage approaches zero. Your shiny new capability is someone else’s weekend hackathon project. Features have gone from competitive differentiators to commodity expectations. They’re the ante, not the winning hand.

This doesn’t mean features don’t matter. A well-designed feature still beats a sloppy one. It means features are the cost of entry, not the competitive advantage. When AI compresses the time to build from months to days, your feature lead is a melting ice cube.

Here’s where things get genuinely disruptive. For most SaaS products, the user interface has been the product in the eyes of customers. The screens, the workflows, the clicks and drags and dropdown menus — that’s what people have been paying for and interacting with daily.

But AI assistants and agents are increasingly operating above the application layer. They don’t need your lovingly crafted UI. They need your data, your logic, and your domain models. The interface becomes optional. Or at least, one of many possible interfaces.

Think about what this means: if an AI agent can launch a campaign, adjust segments, and optimize spend by talking directly to APIs, then the carefully designed “campaign builder” screen you spent two years perfecting just became... a nice-to-have.

This is unbundling in its most profound form. Not the unbundling of features into point solutions (we did that in the 2010s), but the unbundling of interaction itself from the underlying platform.

For decades, SaaS platforms derived enormous power from being the “system of record,” the authoritative home of customer data, transaction histories, campaign performance, pipeline metrics. If you wanted your data, you had to come through them. It was a lock-in strategy disguised as a service.

But the rise of cloud data platforms — Snowflake, Databricks, BigQuery, and their kin — has fundamentally changed this equation. As these platforms increasingly serve as the shared data foundation underpinning all of a company’s business applications, the raw data is no longer a locked asset “owned” by any single SaaS product.

This doesn’t mean data is irrelevant. But data alone is a weak moat when every application in the stack is reading from and writing to the same cloud data layer. If your primary value proposition is “we hold your data,” you’re in trouble, because your customers’ data teams are already busy consolidating that data somewhere you don’t control.

Platform ecosystems remain one of the most durable competitive advantages in software. The network effects, the switching costs, the compounding value of a thriving partner and developer community. These are genuine moats that AI can’t easily replicate.

The problem? Most SaaS companies talk about ecosystems but act like product companies. Their “platform” is really just their product with some APIs and a marketplace that’s more of a marketing page than an actual economy. They think “ecosystem” means “we have integrations” rather than a genuine philosophy of openness, where data flows freely, third parties build real businesses on your platform, and customers can compose their own solutions rather than consuming yours as prescribed.

That philosophy of openness is harder than it sounds. It means ceding some control. It means enabling others to create value you didn’t envision. It means — and here’s the uncomfortable part — sometimes watching partners build things that compete with your own features and being okay with it, because a thriving ecosystem is worth more than any single feature you’re protecting.

Very few SaaS companies have the institutional courage for this, which is exactly why ecosystem strength will separate the winners from the also-rans in the next era of software.

Strip away the features, the interfaces, the proprietary data lock-in, the ecosystem lip service, and what you find underneath is something far more valuable than any of those things ever were. Something most SaaS platforms have been accumulating for years without quite realizing it.

Context.

SaaS platforms have long been called “systems of record,” and some have upgraded their branding to “systems of truth” or “single source of truth.” But both framings undervalue what actually matters. Records are just data. And truth implies something universal and static — the same for everyone, in every situation.

Context is different. Context is personal and situational. It’s knowing not just what the data says, but what it means for this marketer making this decision, or for this customer in this moment.

Context is where domain expertise meets operational reality.

The best SaaS platforms are steeped in it, often without realizing it. Domain expertise is embedded in every workflow, every governance rule, every taxonomy that organizes messy real-world complexity into something coherent. The patterns that drive consistency across a business, the coordination required across dozens of moving parts, the decisioning logic that turns raw data into smart action. It’s all there, accumulated over years of serving thousands of companies in a specific domain.

That is the real treasure. Not the records. Not the interface. Not any particular feature. The contextual intelligence — the deep, structured, domain-specific knowledge embedded in the platform.

This contextual intelligence operates on two dimensions.

First, it powers the people running the operation, helping marketers work with greater consistency, better coordination, and sharper decisioning across an increasingly complex landscape.

Second, and perhaps more transformatively, it shapes how a business engages with its customers. Not just in any single interaction, but across the full arc of the customer journey. Every touchpoint shaped by deep domain understanding, every communication informed by the customer’s specific situation and history, every experience built coherently on the ones before it.

Marketers are already using generative AI to move beyond simple substitution and if/then branching into far more dynamic, adaptive engagement. But generative AI is only as good as the context it’s given. Without rich, structured, domain-aware context, you get content that’s fluent but generic — technically impressive and strategically empty.

Great context is what makes AI-driven engagement actually reliable and effective.

This is where a concept in the AI world becomes critical: context engineering. The discipline of assembling the right instructions, the right curated data, and the right tools to execute actions, all tailored to a specific situation. Context engineering is what transforms a general-purpose AI from a clever autocomplete into a capable, trustworthy operator.

It’s also hard. For most companies, it is true engineering — complex, resource-intensive, and requiring deep domain knowledge that can’t be faked.

And this is precisely the opportunity for SaaS platforms. The best ones have spent years accumulating exactly the domain expertise, governance logic, and operational patterns that context engineering demands. They’re uniquely positioned to do the context engineering for their customers, packaging all of that accumulated intelligence into a service that AI agents and human users alike can tap into.

This is where SaaS platforms have an opportunity to evolve into something more powerful than what they are today: context-as-a-service (CaaS) platforms.

A CaaS platform doesn’t just store data or present interfaces. It provides the contextual substrate — domain intelligence, governance guardrails, operational scaffolding — that makes any interaction, human or agentic, informed and reliable. In this model, the platform becomes less of a destination and more of a foundation.

Its value isn’t in controlling the experience; it’s in making every experience smarter.

This evolution won’t happen automatically, and it won’t happen for platforms that try to have it both ways — claiming to be open while funneling everything through their own interfaces and their own agents.

The CaaS opportunity demands genuine openness. Not “open” as a marketing buzzword, but open as an architectural and philosophical commitment:

Let people build what they want. If a customer wants to build a custom interface that sits on top of your platform’s context layer, let them. If a partner wants to create an agent that orchestrates actions across your platform and three competitors, enable it. Your job is to be the best source of domain context, not to control every interaction.

Let people use the tools they want. The agent ecosystem is exploding. Customers will use AI tools you’ve never heard of, built on frameworks that didn’t exist last quarter. Your platform needs to be accessible to all of them through robust, well-documented, standards-based APIs and protocols — not just through your own proprietary agent framework.

Invest in the ecosystem more than the product. This is the hardest shift for most SaaS companies. It means allocating engineering resources to developer experience, partner enablement, and API quality with the same intensity you allocate to your own product features. It means measuring ecosystem health — partners, integrations, third-party innovations — as a top-line success metric.

The platforms that embrace this will discover something counterintuitive: by giving up control of the interface layer, they can become more essential, not less. When you’re the authoritative source of domain context that every agent in the ecosystem depends on, you’ve built a position that’s far more durable than any UI ever was.

The old pillars of SaaS value — features, interfaces, data lock-in — are crumbling. What’s left standing is something more durable and more valuable: the domain expertise, the governance logic, the operational intelligence that these platforms have accumulated over years of serving real businesses.

The work ahead is productizing context engineering. Structuring that intelligence, opening it up, and making it the foundation that an entire ecosystem of humans and agents can build on.

The platforms that commit to this work won’t just survive as context-as-a-service providers. They’ll become more essential than they ever were as traditional SaaS — because in an agentic world, context isn’t just another feature. It’s the whole game.

Never a dull moment in martech,

Scott