I’m speaking to you from right here. The place where builders from all over the planet gather to churn out “the greatest product of all time” ’cause I’m one of them! And I can promise you one thing: SaaS interfaces are literally vanishing before our eyes.
Just a few years ago, building a SaaS was all about showing off. You had to excite and dazzle people. More buttons, more features, more widgets, more dashboards than you could ever possibly use, more 400-page manuals that no one would ever read. Overkill was the norm. “My product has 87 configuration screens.” So what? “Well, yeah! it’s a real product man.”
Today? Today, a SaaS is nothing more than infrastructure. Period. And preferably, infrastructure that’s simple, fast and easy to understand. Managers don’t have weeks to spend reading documentation written by developers who hate human beings. They want to click. They want it to work. They want to go home.
Buttons everywhere? Dashboards that look like an Airbus cockpit? Drop-down menus inside sub-menus with tabs inside those? Naaaaaaah. Just a chat. A text field. You type what you want. And it gets done.
You have to wonder if, without really meaning to, we aren’t trying to make humans a little bit dumber. Seriously. We spent twenty years teaching people how to use complex software. We created training courses, certifications and hours of YouTube tutorials. And then, suddenly, we say: “No, actually! forget all that. Just write a sentence. That’s it.”
The worst part is, it works. The absolute “worsest” part is that it’s better.
Chat as a universal interface: laziness or genius?
So, let’s ask the real question: “Are we becoming dumb?”
No just kidding (not really but.. you know..), here’s the real question : “Are we witnessing a stroke of genius in simplification, or intellectual laziness disguised as innovation?”
Honestly? Both!
On one hand, it’s brilliant. Nobody ever dreamed of spending 3 hours configuring a workflow in a tool whose name they forget every 6 months. Nobody. If you can just write, “Send a follow-up email to leads who haven’t replied in 10 days except weekends and change the tone if it’s a premium client” and it happens automatically? We’ll take that. We’ll take that in a heartbeat. And you know why! Because we, as humans, ARE LAZY!
On the other hand, there’s something nagging at the back of your mind. A little voice saying: “Wait a minute. If no one understands what’s happening under the hood anymore, who fixes it when it breaks?”
Because chat is magical! right up until the moment it gives you an answer that is perfectly confident and totally wrong. They call that a “hallucination“.
And then you stare at the screen. You stare at your coffee (or chocolate milk.. I don’t drink coffee). You stare at the screen. You think: “Right. So, what do we do now?”
And of course, there will always be someone telling you: “Naaah, that’ll never happen to me. I’m an “AI leader”. I know how to control AI. I know exactly what it can and can’t read, and what it should base its answers on! Because with me, AI will NEVER hallucinate.”
If someone tells you that, you know one thing: you’re dealing with a bullshitter.
Because even RAG can hallucinate. Yes, really. RAG. The thing we’ve been sold for two years as the miracle solution, the AI safety belt, the “don’t worry, it’ll only answer based on your documents” fix. Well bad neeews! It hallucinates too.
So why? Why, even with your documents, your embeddings, your clean retrieval process, your 3000-token system prompt (your 20 000 dollars that you paid to “AI Freelancers” to create you that rag you’re so proud of), does it all go off the rails?
Because RAG isn’t about absolute truth; it’s about retrieval. And between what gets retrieved, what gets understood and what gets generated, there are 3 points where things can go wrong. Retrieval pulls up the wrong passage. The model misinterprets it. And the generation phase does what it does best: confidently making things up.
So clearly, you can’t escape hallucinations!
Another casualty of this great simplification is documentation.
In the past, if you built a SaaS product, you wrote documentation. They were inseparable. Nobody actually read the docs, but they existed. They offered reassurance. They said: “Look, we’ve thought of you, we’ve thought of everything. We even thought of the things you won’t understand. We’re users-first!”
Today? Today, the documentation is the chatbot itself. You ask a question and it answers.
No “Getting Started” page. No FAQ. No “v2.3 to v3.0 migration guide.” Just: “Hi, how can I help you?”
And honestly that’s hugely better. But it implies one thing: we’ve stopped explaining how our products work. We’ve stopped structuring knowledge. We simply put a language model in front and said “You think! I don’t wanna..”
It’s bold. It’s also a bit unsettling.
The real issue: who controls what?
Here’s the point I’m getting at. Behind the disappearance of interfaces lies a question far more serious than “button vs chatbot.”
When your SaaS is nothing more than infrastructure called upon by agents, who decides what? Who decides which email gets sent? Who decides which customer takes priority? Who decides when an alert should go out?
It used to be you. You clicked. You configured. You knew what was happening. You were in control. You saw a button, you pressed it and something happened. It was simple, sometimes it was less simple, but it was clear.
Now, you write a sentence. An agent interprets it. A model makes a decision. An API executes it. And you? You just watch. You hope. You trust.
That isn’t necessarily a bad thing. It’s just different. And that difference isn’t talked about enough. We talk about productivity, speed and user experience. We don’t talk about the loss of control.
