You have decided to automate a business process. The problem is that the tooling landscape is fragmented and every vendor claims to solve every problem. Zapier, Make, n8n, Power Automate, custom code — they all look superficially similar until you try to do something specific, and then the differences matter a lot.
This post breaks the landscape into four meaningful categories, explains when each one is the right tool, and gives you a practical way to make the decision before you commit time and money to the wrong approach.
Why platform choice matters more than most people think
The wrong automation platform does not just create a bad experience — it can make your process more brittle and expensive than the manual version it replaced.
Over-engineered solutions break in ways that are hard to debug. Under-powered tools hit their limits just when the workflow starts getting value. And some platforms charge per task at a rate that makes the economics look completely different at real production volume than they did during the trial.
Getting this decision right upfront is worth more than the time it takes.
The four categories of workflow automation tool
Rather than comparing individual products, it is more useful to understand the categories. Most of the major platforms fall cleanly into one of these four.
1. Hosted iPaaS platforms (Zapier, Make)
Integration Platform as a Service (iPaaS) tools like Zapier and Make let you build automations through a visual interface, connecting pre-built app integrations called triggers and actions. You do not write code. The platform hosts and runs everything.
These are the right choice when:
- Your automation connects established SaaS tools (Gmail, Slack, HubSpot, Google Sheets, Airtable, etc.)
- The logic is straightforward — trigger an event, do one or two things in response
- You need to be running within hours, not days
- You do not have developer resources or do not want to maintain infrastructure
Where they struggle:
Pricing is the biggest practical constraint. Both Zapier and Make charge based on task volume — the number of individual actions your automations run. What costs $50/month at a few hundred tasks per month costs significantly more at ten or a hundred thousand. This is fine for low-volume automations but becomes a genuine cost driver at scale.
Both platforms also have limits on how much logic complexity they can handle cleanly. Deep conditional branching, multi-step error handling, and iterating over large datasets are technically possible but quickly become hard to build and harder to debug.
Make (formerly Integromat) is generally more powerful and more flexible than Zapier for complex logic, at a lower price per task. Zapier has a larger app library and simpler onboarding. For straightforward integrations between common tools, either works.
2. Self-hosted open-source platforms (n8n, Node-RED, Windmill)
Tools like n8n and Node-RED offer a similar visual workflow builder to Zapier and Make, but you run the software yourself — on a server you control, or through a managed deployment.
These are the right choice when:
- You process sensitive data that cannot go through a third-party SaaS platform
- Your automation volume is high enough that per-task pricing would become expensive
- You need more customisation than hosted iPaaS allows (custom code steps, complex logic, direct database access)
- You have some technical capacity — a developer, a technical co-founder, or a managed hosting provider — to handle the infrastructure
Where they struggle:
The tradeoff for control and cost is maintenance overhead. You own the infrastructure: server management, updates, backups, monitoring, and failure handling. If something breaks at 2am, that is your problem.
n8n in particular has built good community support and a large library of pre-built integrations. For organisations with even modest technical capacity, it is often the most cost-effective path once automation volume grows beyond what hosted iPaaS pricing supports.
Self-hosted platforms are not the right choice if nobody in your organisation has the technical background to keep the infrastructure running reliably.
3. Native automations in tools you already use
Most modern SaaS platforms have built-in automation features. HubSpot has workflows. Salesforce has Flow. Slack and Microsoft Teams both have built-in automation builders. Linear has rules. These native automations do not require a separate tool and often handle the most common use cases without any integration layer.
These are the right choice when:
- The automation exists entirely within one tool (or primarily within one ecosystem, like the Microsoft 365 suite)
- You want to minimise the number of systems you operate and maintain
- Your team already uses the platform, so there is no additional adoption barrier
Where they struggle:
Native automations hit their limits the moment you need to involve systems outside the platform. A HubSpot workflow that sends an email when a deal closes is native. A workflow that also creates a project in your delivery tool, logs the deal in a spreadsheet, and sends a custom webhook to your billing system is not.
The other limitation is portability. Native automations are tied to the platform. If you migrate away from the tool in the future, those workflows do not come with you — they need to be rebuilt.
4. Custom-built integrations
Custom integrations are code written specifically for your use case: an API connection, a scheduled script, a backend service that listens for events and acts on them. This is what a developer builds when none of the above fits.
This is the right choice when:
- The integration involves systems that do not have pre-built connectors in any iPaaS platform
- The logic is complex enough that a visual builder would be impractical to maintain
- You need error handling, retry logic, and operational robustness that visual tools cannot provide cleanly
- You are embedding automation logic into a larger software system
Where they struggle:
Custom code requires developer time to build and ongoing developer access to maintain. It is the most flexible option and the most expensive — both upfront and over time. Choosing it for a problem that a hosted platform could solve is rarely the right call.
That said, for organisations with development capacity, custom integrations are often more reliable and more maintainable than complex visual workflows. A well-written script is easier to debug and modify than a sprawling multi-step Zapier automation with fifteen conditional branches.
The mistake most businesses make
The most common mistake is choosing the most recognisable brand rather than the right tool for the problem.
Zapier is the best-known name in workflow automation. It is also one of the more expensive options at any meaningful volume, and it has real limits for complex logic. Many businesses would be better served by Make, by n8n, or by a simple custom script — but they default to Zapier because it is familiar.
The second most common mistake is underestimating volume. Every automation trial starts at low volume. The economics look fine. Then the business grows, the automation gets adopted more widely, and the monthly task count climbs. At 10,000 tasks per month on Zapier's standard pricing, the numbers look very different from month one.
Model out your expected task volume before you commit to a pricing structure. Most iPaaS platforms make this easy enough to estimate. If you are unsure, round up.
How to make the decision
A practical sequence:
1. Map the automation first. Before you touch any tool, write out what the automation does: what triggers it, what data it processes, what it outputs, and where that output goes. This takes twenty minutes and prevents you from discovering mid-build that the tool you chose cannot handle a step you need.
2. Check whether native tools cover it. If the automation lives within one platform or ecosystem, check what that platform can do natively before adding a third-party tool.
3. Estimate monthly task volume. Count how many times the automation will run in a month and how many actions each run involves. Run the numbers against the pricing tiers of any platform you are considering.
4. Assess your technical capacity honestly. Self-hosted tools and custom code require someone who can maintain them. If that capacity does not exist in your organisation, factor in the cost of getting it — either by hiring, outsourcing, or accepting that a managed platform is the right choice even if it costs more per task.
5. Start simple. Whatever platform you choose, automate one step first. Verify it works in production before building out the rest. Complexity is easier to add than to remove.
When does AI fit into this?
Most workflow automation does not require AI. Artificial intelligence belongs in a workflow when the input is unstructured — a free-text email, a document with variable format, a conversation transcript — and you need to extract meaning or generate a response rather than just route structured data.
When AI is in the loop, it should handle one discrete step, not the whole workflow. The automation platform routes and orchestrates; the AI interprets or generates. This separation keeps the logic testable, the errors diagnosable, and the costs predictable.
All four platform categories can call AI APIs. The choice of automation platform is independent of whether you use AI — it is about the routing, triggers, and integration logic around whatever AI step you need.
If you are evaluating whether a process needs AI or simple rules, the workflow automation guide covers how to make that call.
Where to go from here
If you are clear on the category but unsure about specific products, these are the starting points I typically recommend:
- Hosted iPaaS, simple logic: Make (better value and more power than Zapier for most use cases)
- Hosted iPaaS, maximum app coverage: Zapier (largest library, best for connecting niche SaaS tools)
- Self-hosted, high volume or sensitive data: n8n (good balance of flexibility and usability)
- Within Microsoft 365: Power Automate (native, no extra cost if you have Microsoft licences)
- Complex custom logic or embedded in a larger system: Custom code, scoped and built by a developer
The right answer depends on your specific processes, volume, data sensitivity, and technical capacity. None of these platforms is universally the right choice, and most businesses that have been automating workflows for more than a year end up using more than one.
If you want an independent view on which approach fits your situation, that is exactly the kind of question I help businesses work through. You can find more on how I approach workflow automation engagements, or get in touch directly if you have a specific process or integration in mind.