The moment I knew something had shifted permanently was not the Mercedes-Benz announcement or the Series B or any of the coverage that followed. It was a Thursday afternoon when a client I had onboarded two years ago forwarded me a renewal quote from Zapier and asked, in the flat tone of someone who has already made a decision and wants confirmation, whether there was any reason not to move. The quote was for roughly the same capability they already had. The number had gone up. I told them there was no reason not to move. The migration took four days. They have not mentioned Zapier since.
That is what the end of a platform war looks like from the ground level. Not a concession speech. Just a series of individual decisions that stop going one way and start going another, until the aggregate is obvious to everyone except the people writing the investor updates for the losing side.
I want to be precise about what I mean when I say n8n won, because won does not mean Zapier is dead and Make.com is closing their offices. Both tools will continue to exist and both will continue to have users. What won means in this context is that the default assumption for anyone building serious automation infrastructure has shifted. Eighteen months ago, if you asked a mid-size operations team what they were using for workflow automation, Zapier was the expected answer and n8n required an explanation. Today that is inverted for any team with a technical person involved in the decision. n8n is the explanation-free answer and Zapier is the one that requires justification.
The technical reasons for this are real and I have written about them before so I will not repeat the full case. The short version is that the Code node closes every capability gap that would otherwise require a workaround, the self-hosting model makes the pricing conversation irrelevant at scale, and the AI integration story is better than either competitor because HTTP Request nodes do not care what API they are calling. I connected Claude to an invoice processing workflow last year using nothing but an HTTP Request node, a Code node to structure the prompt and parse the function call response, and a Set node to route the output. The documentation for that integration was incomplete in the specific way n8n documentation is often incomplete, which is that it showed the happy path clearly and omitted what happens when the API returns a 429 with a retry-after header that your error handling node needs to read and respect. I worked that out from the community forum. It took two hours. It has not broken since.
What Zapier does next is the more interesting question and I think the answer is already visible if you look at where they have been investing. The AI features they have been shipping are not aimed at the Lukmon Isiaq market. They are aimed at the executive assistant who needs to automate her email triage without involving IT and does not know what a webhook is. That is a real market. It is also a market that does not overlap much with the one n8n is winning. Zapier is retreating upmarket into the non-technical user segment and dressing it as AI transformation, which is a reasonable strategic move if you accept that you have lost the technical buyer. The pricing stays high because the non-technical buyer is less price-sensitive and more likely to stay once they have built dependencies.
Make.com is in a harder position because their competitive differentiation was always the interface, and a good interface is worth less when the tool you are being compared against has a Code node. I have had three separate clients move from Make to n8n and in every case the trigger was hitting a complexity ceiling, not a price ceiling. Make’s canvas is genuinely better to look at than n8n’s. That advantage does not survive the moment when you need to do something Make’s modules cannot express cleanly. The feature they would need to build to address this is something close to a sandboxed JavaScript execution environment, which is exactly what n8n already has, and building it would mean accepting that they have been competing on the wrong dimension for two years.
CrewAI and the agent frameworks are trying to position themselves as the next layer above workflow automation, the layer where you do not build explicit workflows but instead give agents goals and let them figure out the steps. I have watched four CrewAI demos in the last six months. Every one of them has the same structure, which is an impressive-looking multi-agent coordination producing a correct output on a well-defined task in a controlled environment. I have not yet seen a CrewAI demo that shows what happens when one agent produces a subtly wrong intermediate output that the next agent uses as a valid input and the error propagates silently to the final result. That failure mode is not theoretical. I have seen it happen in LangChain implementations with real data. The fact that CrewAI is not showing it in demos does not mean it does not exist.
n8n’s response to the agent frameworks is sensible. They have added AI agent nodes that can call tools and loop on results, but they have kept the execution model deterministic and auditable by default. The agent runs inside the workflow. The workflow is still a workflow. When it breaks, you can see where.
The war being over does not mean the work is over. It means the baseline has moved, and the tools that do not accept that are going to spend the next three years explaining why their pricing model still makes sense.
Olaitan Oladipo holds a BSc in Sociology from Olabisi Onabanjo University. He is a self-taught automation builder who has spent years inside n8n doing the work that most tutorials skip: debugging OAuth errors at 2am, migrating client automations from Make.com mid-project, fighting reverse proxy misconfigurations on AWS EC2, and figuring out through trial and error what actually holds up in production versus what only looks clean in a demo.
He is not a developer by training and not a SaaS founder. He is the person in the Discord server who actually answers the question instead of linking to the docs.
His writing on n8n Automation Tutorial covers self-hosting, AI agent workflows, tool comparisons, and the security vulnerabilities the automation industry would rather not discuss. He has built AI-assisted invoice approval flows using OpenAI function calling, connected Claude via HTTP Request nodes, and holds considered opinions about Zapier, Make.com, LangChain, and CrewAI that their marketing teams would not appreciate.
He writes for people who are technical enough to follow a tutorial but experienced enough to want the honest version.