I ditched my CMS, replaced it with an AI coding tool, and accidentally stumbled into an argument about the future of software.
The CMS I Didn't Need
I used to run my website on a traditional CMS platform. Drag-and-drop editor, template gallery, monthly subscription. The usual.
It lasted about six months before I quietly abandoned it. Not because the platform was bad. It was fine. The problem was everything around it: logging in to make a small text change, fighting with formatting that never quite matched what I wanted, copying blog post drafts between tools, wrestling with integrations that didn't quite integrate. The admin overhead of a "simple" website platform was eating into the time I should have been spending on actual content.
So I dropped it. Moved everything to static files and an AI coding tool. Now I write in conversation. I describe what I want. The AI has my tone of voice guide, it can compare against every previous post, and it helps me break a single idea down into a content series. A bit of back and forth, a commit, and the content goes live.
No login screen. No drag-and-drop editor. No template gallery.
And here's the bit I didn't expect. The interface layer, the thing I was paying a monthly subscription for, turned out to be the least valuable part of the whole setup. What actually matters is the context: my tone of voice, the patterns across my existing content, the reasoning about what makes a good post versus a mediocre one. The AI doesn't need a dashboard to do that work. It needs the context.
I didn't have a name for what I'd built until I read a piece by Scott Brinker about the future of SaaS. Turns out, I'd accidentally become a one-person case study for something much bigger.
What Actually Has Value?
Brinker's argument landed with me because I'd already lived it in miniature. His thesis is that the traditional things SaaS companies rely on to keep customers (features, a nice UI, holding your data hostage, "ecosystem" integrations) are all weakening at the same time.
Think about it.
Features are table stakes. AI compresses build time from months to minutes. Your shiny new capability is someone else's weekend project. Features are the ante, not the winning hand.
The UI is being bypassed. I bought my CMS for its interface. I left because the interface got in the way. The same thing is happening at scale. AI agents don't need a "Campaign Builder" screen. They need an API. Your carefully designed interface is becoming optional.
Data lock-in is losing its grip. Cloud data platforms have made portability easier. If your only value is "we hold the data," that's not much of a position anymore.
Most "platforms" are fake ecosystems. They're products with a marketing page for integrations. A real ecosystem requires giving up control, and most SaaS companies aren't willing to do that.
Strip all of that away and what's left? The accumulated wisdom of the platform. The governance rules, the domain patterns, the taxonomies. The context.
Data Isn't the Same as Meaning
This is the distinction that clicked for me.
A record says "Customer X visited the pricing page." That's universal, static. Any system can capture it.
Context says "Customer X visited the pricing page, but they're an enterprise customer in Q4, with low engagement, a history of high returns, and three open support tickets. This pattern suggests churn risk, not a buying signal." That's personal, situational. It's the meaning, not the measurement.
Brinker calls this shift Context-as-a-Service. Instead of selling software that does things, you're selling the intelligence for how things should be done. Raw data goes in. A context engine (built from governance rules, historical patterns, and domain expertise) produces smart actions. Those actions serve both human users and AI agents.
I think he's right, and I think it explains my own experience. My old CMS was Software-as-a-Service. It gave me tools. My current setup is closer to Context-as-a-Service. The AI doesn't just write; it writes in my voice, informed by my content history, following my standards. That context is the product. The interface is incidental.
AI without context is just clever autocomplete. To be useful, it needs fuel. And that fuel is the domain knowledge that SaaS platforms have been accumulating for years, often without realising it was their most valuable asset.
The Part That Should Worry SaaS Companies
My personal example is trivially small. But scale it up and the implications are serious.
If you're a SaaS company whose value is mostly "we built a nice UI for this workflow," you're exposed. AI agents don't care about your UI. They'll talk directly to your backend logic, or they'll recreate it entirely. We wrote about exactly this pattern in our piece on the real cost of running a business. The tool explosion of the last two decades created a world where businesses pay for dozens of interfaces they barely use. AI is starting to collapse that.
If your value is "we hold the data and you can't leave," that's weakening by the month.
But if your value is "we understand your domain so deeply that our context engine makes every tool smarter," you become the foundation everything else is built on. You stop being a destination (a place users visit) and start being the infrastructure (the thing everything connects to).
Brinker's phrase for the new moat is ecosystem density. Stop measuring logins. Measure how many apps, agents, partners, and APIs connect to your context layer. The denser that ecosystem, the harder you are to replace.
There's a catch, though. You have to be radically open. You can't be a context platform if you insist on controlling every interaction. Let people build. Let people choose. That's a hard pill for companies who've spent years designing walled gardens.
I've Watched This Play Out at Enterprise Scale
My CMS example is deliberately small. But I didn't come to this thinking cold.
I spent twenty years at two of the most successful marketing services companies in the industry, building software that connected the business with its clients. The domain was content production and marketing at scale. Thousands, sometimes millions of assets a year. Every client different: different systems, different brand guidelines, different approval workflows, different degrees of involvement, different definitions of what "done" looks like.
The systems we built always centred on the client's needs in that domain. Configuration. Asset type definitions. Metadata schemas. Hooks into PIM systems. Workflow rules that encoded the specific way this client wanted this type of asset handled at this stage of production. And underneath all of that, a deep understanding of how it actually plays out when you're running at volume. That configuration, those rules, that accumulated domain intelligence? That was the true gold dust. Not the software itself.
But here's the thing. For the last ten years, the industry put enormous emphasis on UI. And for good reason. To drive a process and capture meaningful KPIs, you need the squidgy thing between the keyboard and the chair to: a) understand what they're doing, b) be motivated to do it (which meant investing in polished, engaging interfaces), and c) do it quickly. The entire UX discipline existed because getting data into the system was the hard part. Beautiful interfaces were the bribe.
Now look at what's happening. ChatGPT and its competitors have shown that the interface layer is converging on something we had over thirty years ago. A single text input. A conversation. The same pattern as AIM, MSN Messenger, or IRC before that. The interface we spent a decade perfecting is being replaced by the interface that already existed before most of us had broadband.
There will always need to be a UI to interact with a system. But the UI is no longer the differentiator. It's no longer the moat. The moat is the context underneath it: the client configuration, the domain rules, the metadata structures, the operational knowledge of how it all connects at scale. That's the bit that can't be replicated by spinning up a new chat interface over a weekend.
I lived this before I had a name for it. Brinker just gave me the vocabulary.
Where I'm Less Certain
I buy the direction of this argument, but I'm not sure how fast it plays out.
Most SaaS companies aren't going to wake up tomorrow and reinvent themselves as context platforms. They have revenue models built on seats and features. They have customers who still want dashboards and buttons. The shift Brinker describes is real, but it's measured in years, not quarters.
And there's a question about who actually has good context. Plenty of SaaS platforms have accumulated data without turning it into genuine domain intelligence. Having ten years of customer records isn't the same as understanding what those records mean. The context engine doesn't build itself.
For small businesses like mine, the practical reality is simpler. I found that the context I'd built up (my tone, my content patterns, my standards) was more valuable than any interface. That might be true for others too, or it might just be true for someone who writes a lot and happens to enjoy working in a terminal. I'm honestly not sure how universal the pattern is.
The Shift Is Real, Even If It's Slow
Features? Commoditised. UI? Increasingly optional. Data lock-in? Dissolving.
Context? That's where the value is moving.
I don't think SaaS is dying. But I do think the companies that survive the next decade will be the ones who figured out that their product was never really the software. It was the intelligence that made the software worth using.
This thinking applies if:
- You're a SaaS company wondering why feature releases aren't moving the needle like they used to
- AI agents and integrations are bypassing your carefully designed workflows
- Your "competitive moat" is really just a nice interface that someone could rebuild in a weekend
- You've been accumulating domain knowledge for years and haven't figured out how to turn it into a product
Wondering what your context layer actually looks like? Let's talk about it.