Automation tool comparisons are usually feature tables where whichever product has the most green rows wins. That isn't how I decide. I decide with four questions, and the answer is usually settled before any tool is opened.
First question: which licence does the company already pay for?
If the company has Microsoft 365 — and most do — it is already paying for Power Automate, Power Apps and SharePoint. Building the workflow there adds no cost, adds no new vendor to the contract, and doesn't require IT to approve anything.
This question settles roughly seven cases out of ten. Not because Power Automate is the best tool in the abstract, but because an automation that forces a purchase has to justify that purchase, and often it doesn't.
Second question: where does the data live?
If the process lives inside SharePoint, Outlook, Teams and Excel, leaving that world is unpaid work: connectors, authentication, duplicated permissions and one more place to look when something fails.
If the process lives outside — an ERP with its own API, a storefront, a service nobody connects out of the box — Power Automate's advantage narrows, and that's where n8n comes in.
Third question: how long does the task run?
This is where Power Automate falls short, and it's worth saying plainly. For long-running tasks, for orchestrating several steps that take minutes, for calling models or slow services that need retrying with judgement, it hurts: execution limits, timeouts, and awkward debugging.
For that I use self-hosted n8n. It runs on its own infrastructure, doesn't bill per execution, and can be versioned like code.
The price is that someone has to maintain that server — and that someone is whoever set it up.
Fourth question: who will maintain it?
This is the heaviest question and the least asked. A Power Automate workflow can be opened tomorrow by anyone in IT with permissions, because it lives where they already look. A self-hosted n8n needs someone who knows it exists, where it runs, and how to restart it.
When the client won't have that person, building the technically more elegant solution is handing them a debt with a due date on it.
Where Make fits
I use Make (formerly Integromat) rarely, for one specific reason: it's convenient for wiring external services together, but it bills per operation, and in an internal process with volume the invoice climbs without warning. For a marketing workflow with few steps and third-party services it fits well; for the process that holds up a company's operations, I'd rather not have a meter running.
The honest summary
- Internal process, data in Microsoft 365, in-house people: Power Automate, no hesitation.
- Integrations outside the ecosystem, long tasks, orchestration: self-hosted n8n.
- Wiring third-party services at low volume: Make does the job.
- The company won't have anyone to maintain a server: don't propose a server.
The same tidy process works in Power Automate, in n8n or in a shared spreadsheet; an untidy one works in none of them.
The tool matters less than the order.
If you're deciding where to build yours and want an opinion before you start, write to me or see how each line of work runs.
