Well… after years of not posting anything, I think this is a good topic to come back with.
Maybe AI lowers the barrier to creating software. But for many business workflows, code was never the main challenge. If anything, this makes clearer what no-code was solving in the first place.
There is a lot of fuss now about AI coding agents, vibe coding, and how software is becoming easier and easier to create. And to be honest, I agree with that part. AI is already useful in many aspects. It helps people move faster, test ideas faster, explore alternatives faster, and yes, write code faster too. It would make no sense to deny that. I’m using that myself for several internal tools I have right now.
But from my POV, that still leaves a very important question: was code ever the main challenge for the kind of apps many of us build with AppSheet? I don’t think so.
The real challenge
When people think about AI disrupting programming, I think many are looking at the wrong layer, at least for now.
Yes, AI can help create software. Yes, AI can write code and it can even help with software engineering tasks beyond just code writing. But for many business workflows, code was never the main issue anyways, we know that from experience.
The real challenge was translating real-world challenges into business logic.
That is where things usually get messy. Not in the code itself, but in the process behind it: who should do what, what should happen first, what happens if data is missing, what if approval depends on some condition, what if one user should see one thing and another user should see something else, what if the rule changes next month, what if the same workflow behaves differently depending on the business case. That is the part that tends to be hard, that’s what we have been doing for years.
So when I look at this topic, I do not think the key question is whether AI can generate code. I think the better question is whether generating code solves the actual difficulty behind business software. In many cases, I don’t think it does (and that’s how I’ve made a living so you better believe me).
What no-code was solving
And this is where no-code starts making even more sense to me in the current AI era, or at least things start to click for me to another degre.
Most people building with AppSheet are not coders at all. “Citizen developers” are called that for a reason. Usually, they are the people who know exactly what they need. They know the workflow, the pain points, the bottlenecks, the exceptions, the people involved, the business need… they just don’t know how to translate that into software (less so how to code your way there).
That is why I don’t think the value of AppSheet was ever just “you do not need to type code”. That sounds too superficial.
The real value was that a business person (or a consultant/developer like us) could take something they already understood very well and turn it into something working: an app, an automation, a report, a process with actual logic behind it. IMO, that is what no-code was solving in the first place.
What AI improves, and what it does not remove
Again, I am not saying AI is useless. Not at all. I would be stupid if I say so and most probably this post wouldn’t “age well”.
AI is great for reducing friction. It helps with prototyping, explaining, researching, debugging, generating code, restructuring code, and many other tasks. In some cases it is genuinely impressive. Just look at how much it has changed in a relatively short time, and most probably in a couple of years this new category of software will be the new UI: voice agents, on the go interfaces, etc.
But even when using AI agents for software engineering, not coding per se, there is still a big gap today.
Because someone still needs to oversee what is being built. Someone still needs to review effectively. Someone still needs to understand if the result is actually correct, maintainable, secure, and aligned with the real need.
And this is where I think many people underestimate the problem.
To use AI well in a software engineering context, you still need to speak technical language to some degree (do you know the difference between web frameworks such as Svelte, Angular or Vue? Do you know the advantages of Rust over Zig or using Biome instead of ESLint/Prettier? Do you know if Electron is the right choice? What about Tauri? -I’m sorry, I think I already made the point-). Maybe you don’t need to be an expert and you don’t have to understand technicalities perfectly, sure, but enough to inspect architecture, permissions, integrations, data flows, assumptions, edge cases, security implications, and all those “boring” details that become very important later.
Otherwise, what happens? Usually one of two things: either you 1) ignore the details and trust that things are OK (aka “vibe coding”), or 2) you try to learn enough to review properly (good! Keep going!). The first path is where security holes and bad assumptions usually appear. The second path means the problem didn’t disappear: it just moved.
So yes, AI can generate a lot. But it does not automatically give someone the ability to review what was generated. And I think that difference matters more than people admit.
Why AppSheet still makes sense to me
This is where AppSheet still feels very relevant (who would have thought).
Not because AI failed (there is such thing?). Not because AppSheet does everything. Not because no-code is somehow “better” than other approaches.
Simply because AppSheet solves a different part of the problem.
It already gives you much of the surrounding layer that AI-generated custom software still leaves on your shoulders: infrastructure, platform, authentication, integrations, automation, security context, deployment shape, and a constrained environment with a lot less invisible surface to blindly trust.
And yes, despite my philosophical stance, I think the Google side matters here too.
AppSheet is under Google, integrates very naturally with Google Workspace, and benefits from Google infrastructure and its larger platform context. That does not make it perfect, of course (just ask @zito), no tool is, but it does mean something important: a business builder does not need to own and review an entire custom stack from scratch just to solve a workflow problem.
To me, that is a very big deal, and I’m sure it is for you too.
My reflection after all this time
So when I see people wondering if AI may disrupt software to a core degree, I think the better question is not, “Can AI create software now?” Of course it can help with that, just look at youtube and search “vibe coding” (hate the term TBH).
The better question is, “What was AppSheet really solving in the first place?”
And from my POV, the answer is not just “no-code” in the superficial sense. It was giving business people a way to turn real operational needs into working software without requiring them to become programmers, infrastructure owners, or effective security reviewers first.
That is why I still think AppSheet (and no-code in general) is relevant today. Maybe even easier to understand today than before. Because when code becomes cheaper, you start seeing more clearly that code was not the whole point.
What is your take? Have you built projects using AI Coding agents?
Now, what about AI Agents and AppSheet? I’m excited to show you a surprise soon, stay tuned
