The automation layer is being eaten

I have stopped building automations in n8n and Make. Not as a protest, and not because anything went wrong with them. I just stopped needing to.

Two things happened at once. The foundation labs all shipped scheduling and cron, plus MCP and tool calling, so the orchestration primitive now lives inside the assistant you were already paying for. And separately, most serious companies have now written APIs, tools and MCP servers so that agents can talk to them directly.

The second one is the interesting half. These platforms were not out-competed by a better product. They were obsoleted by their own customers' compliance. Every company that shipped an MCP server removed a small piece of exactly the integration friction those platforms existed to absorb. That friction was the strong suit. It is going.

So here is the prediction, and I am dating it deliberately. Writing in August 2026, on a three-year horizon: without pivoting into something else, the standalone automation platforms become zombie companies or cease to exist. Some of the open-source ones persist, because open source does not need a growth rate to survive.

I want to be careful about the register here, because writing doom about other people's companies is easy and cheap. I am not predicting mismanagement. I do not think anyone there is doing a bad job. I am making a claim about a market: a category's core value proposition evaporated underneath the companies in it, and when that happens they either pivot or they wind down. Plenty of good operators have been on the wrong side of that.

And because a prediction without a falsifier is just a mood, here is mine. I am wrong if, in three years, those platforms are growing revenue on their integration business rather than on something they pivoted into. Not total revenue. Integration revenue. That distinction is the whole test, and it is checkable.

“Every company that built an MCP server removed the integration friction those platforms existed to absorb.”