Slack has just shipped something called Surfaces, and if you build software for a living you should pay attention to it even if you never touch Slack yourself. The Verge covers the launch here — the short version is you describe what you want inside a channel, and AI builds a working mini-app on the spot. A form, a tracker, a little workflow tool. No dev ticket, no waiting on IT, no separate SaaS subscription.
I've spent forty years watching platforms absorb the tools that used to be separate products. This is that pattern again, just faster and more visible.
The pattern isn't new, the speed is
"Describe it, get a tool" is exactly the bet I made with the Masher suite years ago. RSSMasher takes a feed description and gives you content. BookMasher takes a brief and gives you a manuscript. The whole idea is that most people don't want to learn an interface — they want to describe an outcome and get the output. Slack building that natural-language-to-working-tool loop directly into chat is the same instinct, just applied to internal business apps instead of content.
What's different is where it sits. When I build a tool, it's a destination — you have to decide to go and use it. Slack's version lives inside the workflow you're already in. You're mid-conversation about a shipping delay and someone spins up a tracker without leaving the thread. That's not a small distinction. Friction is the thing that kills adoption of standalone software, and Slack has just removed a huge chunk of it by refusing to make you leave the app you're already living in eight hours a day.
What this means if you sell SaaS
If your product is a single-purpose tool that solves a problem people currently solve with a spreadsheet, a Slack channel and some duct tape, you need to be honest about how much of your value is the tool itself versus the fact that nobody had built the thing yet. Because now, plenty of people can build the thing themselves, badly, in thirty seconds, inside the app they were already using.
That's not the end of indie SaaS. It's the end of indie SaaS that competes purely on "I made the thing you'd have to build yourself." The tools that survive this wave are the ones with genuine depth behind the simple request — proper data handling, integrations that actually work at scale, output quality that a quick in-chat AI build can't match, and judgement baked into the workflow rather than a generic prompt response.
That's the bet behind the Masher tools too. Anyone can ask an AI to "write me some content." Getting content that's actually publishable, on-brand, and doesn't need three hours of editing afterwards is a different problem, and that's where the real engineering sits. The describe-it-get-it moment is the cheap part. The valuable part is everything that happens after the first draft.
The platforms have an unfair advantage, and that's fine
Slack, Microsoft, Google — they all have the same play available: they own the workflow, so they can bolt generative building directly into it. An indie builder can't compete with that distribution, and shouldn't try to. What you can do is go deeper than a platform ever will, because a platform's incentive is breadth, not depth. Slack wants Surfaces to work for everyone's vague idea of a tracker. It has no incentive to be excellent at your specific niche.
The practical move for anyone running SaaS right now is to stop treating "AI builds it for you" as a threat and start treating it as a filter. It weeds out the products whose only pitch was convenience. It leaves standing the products with actual expertise, actual data, actual taste built into the output. If your tool can be replaced by someone typing a sentence into Slack, it probably should be.
Where I think this goes next
Give it eighteen months and every major workflow platform will have some version of this — describe it, get it, inside the tool you already use all day. The interesting fight won't be "will AI build simple internal tools." It already can. The fight will be over who builds the depth layer underneath: the data pipelines, the domain-specific quality control, the judgement calls a generic prompt can't make. That's where I'm putting my effort with Masher, and it's where I'd point anyone else building SaaS right now too.
Build the depth. Let the platforms have the surface.
— Wayne