I Review Make.com for AI Business Automation
I’m still yet to find my perfect AI automation platform. So far I’ve tried n8n, Gumloop, Relevance AI, Lindy AI, and today it’s time for my make.com review.
What I’m looking for in all these AI automation tools is a kind of ‘orchestrator’, something to oversee my various business API calls; something that can manage tasks like an AI employee.
Let’s see if Make.com has what it takes to run my AI agent swarm.

New here? Welcome! This is the journey of building a 100% automated AI business in 2025. You’re jumping in after we’ve already kicked things off, so you might want to catch up first.
Check out these key posts to get the full story—and don’t forget to subscribe for updates and exclusive perks:
I’m still working my way through my list of AI automation frameworks (this is 4/8 on my list!) I’ll update them all and summarise when I’m done, but if you’re new here subscribe at the top for the summary post.
My AI Automation Make.com Review
What is Make.com?
For those of you who’ve read along with my other reviews here, you get the drill. Make.com is another ‘AI Workflow Tool’. In short it let’s you use AI as a component in a multi-step cascading actions automation schema.
Make.com calls these workflows “Scenarios”.
Make.com is a workflow builder first; but last month they introduced AI Agents (still in beta).

… so I guess it’s safe to throw Make.com into the pile of AI automation workflow tools.
What do I use Make.com for?
I’m testing Make.com to see if it can be the glue around my various business API actions. It’s worth noting that Make.com seems very flexible and so it could likely be used a bunch of other ways.
But in this review my use case is primarily to encapsulate a single AI Agent in my swarm, (e.g. Directory Prospector, or Directory Builder).
My First impressions of Make.com
Make.com is goofy looking when you first get started. It’s icon-based workflows seem like something a kid made in powerpoint at first. But don’t let that fool you, there’s a lot of technical capacity under the goofy ui.
Just look at this ‘Scenario’:

It actually works really well, but look at the icons, they’re like 260 pixels wide 😨.
Anyway, here were my first impressions of Make.com:
- Goofy UI – It undermines Make.com’s quality
- AI Agent (beta) functionality literally just missing until you upgrade to paid plan (no signposting)
- Overwhelming number of ‘modules’
- New nomenclature (workflow = scenario, node = module, everything seems to be a webhook)
- That said, I kinda like it’s quirky style
Make.com AI Review: Testing
I’ve not been perfectly scientific in comparing these various AI workflow tools. I am doing slightly different tests with each; but that’s how it’s going to have to be here. This experiment, and the AI landscape is in flux.
I actually resorted to testing a new workflow tool today because I reached that point, again, where I almost built one.

I’m getting somewhere with the agent actions for my experimental first swarm (AI Directory Maker), but I’m still struggling to get a platform which can handle everything I need it to.
Onward, then, to test Make.com as an agent factory.
Test Workflow: AI Directory Prospector
My aim for this test is to work through at least the first steps of this agent. I need it to come up with viable ideas for web directories.
I need it to poll Google keyword data and find the low hanging fruit (keyword phrases that will be relatively easy to rank for, but have traffic). Essentially It’s the five steps from the following table:

(Actually I’ve managed to combine steps 2 and 3, and I’m currently skipping step 4.)
Queue Woody re-subscribing to the RapidAPI endpoints for search and domain checks

Getting Started with Make.com Tests
It took me a while to even find the AI agent functionality in Make.com. It’s actually only available for paid plans, and that’s not really said anywhere. Due to single sign on you get whipped quite quickly into the ‘app’ homepage, which on first load is pretty sparse.

After giving up my credit card, (I chose ‘Pro’), I saw that I did indeed now have an ‘AI Agents’ menu item. Great. I started there.
But wait! Making an AI agent is literally useless unless you have ‘Scenarios’, aka workflows.
So sidestepping the new name for workflows, I went in and created a scenario which retrieves Google keyword analysis directly from a RapidAPI endpoint…
Make.com Review: Creating my first Scenario
This is the page you see when you load up a new workflow.

At first I thought it was super goofy, but after a while I got used to the massive ‘module’ icons.
So where to start…
I figured I may need some ‘catch’ module to receive the keyword from the AI Agent I was going to make.
Digging around I found this:

You can actually super simply specify inputs and outputs for a scenario.
That’s very common-sense, and fits well with my FlowSpec approach.
PSA: I can’t keep typing scenario instead of workflow, from now on I’m going to call things what they are: Scenario = Workflow.
AI Agent Service
Specifying an input/output on it’s own isn’t enough to kick off a workflow.
How you have to do this in Make.com is to wrap the ends of each workflow with a webhook, like so:

AI Agent calls Webhook(s)
I don’t love calling these junctions webhooks, nor do I like that they’re effectively an open web endpoint (read: vector in).
As soon as I’d set up the above I nakedly called the webhook via postman API, and it ‘just worked’:

In one way this is awesome – it’s a super quick, no-bs way of making easy to access services.
On the other hand if these endpoints were easily exposed it’d give any bad actor a way into the system, (and a way to cost you credits).
Their solution for this is IP whitelisting, which is a good simple way of providing some protection.

My expectation is nearly every user will end up just leaving these endpoints wide open though, and IMO it’d be better to have an option ‘restrict to internal make.com traffic’ or similar, (because as it stands to use AI Agents to call this webhook I have to leave this blank).
If anyone knows a better way of doing this in make, please do comment below.
Calling API’s in Make.com
Back to the test. Setting up this API calling workflow as actually super simple once I understood the webhook idiosyncrasies. For some reason the throttling and timeout issues I had with n8n and Gumloop just didn’t show up.
Make.com’s http request module is well refined and common sense:

Making a Simple AI Agent with Make.com
Next I made a simple AI Agent which called this keyword-getting workflow as many times as it needed to.
This was super simple and only needed OpenAI API keys, a name, a model, and a prompt:

Once you’ve added that, you then have to add each scenario workflow as a system tool:

In a confusing turn of events, this new AI agent doesn’t appear to have any steps, until you look in “Scenarios”, where you can now see your AI agent as a workflow 🤷🏼?

Anyway in this new scenario-ai-agent-workflow I added a final step to output the AI agents findings to me in Slack:

Test AI Agent outcomes
I had to tweak it in place a little (making the keyword workflow ‘run immediately’), but in short order, it worked nicely:

Now this is only the first half of the AI Agent’s responsibilities, and it’s output is, as of yet, untamed. But I failed to find any blocking snags and upon testing it. Make.com is growing on me.
Will I continue to test after this Relevance AI review?
For this review test I’ve stopped here, but right now I’m trying to stand up the first 2 AI agents, and my next task is to try to do this in Make.com, fully!
Make.com Review: Pro’s
In this review I’ve poked fun at Make.com’s UI, nomenclature, and webhooks. But in all honesty it’s grown on me. In terms of actual inputs and outputs I think it’s a solid system that is definitely usable for AI business automation today.
This is what I love about Make.com so far:
- Use your own AI API keys – It’s nice to have transparent usage data
- HTTP call module covers all eventualities well (timeouts, headers, params in a common-sense way)
- Open Webhooks are good for testing (via postman api etc.)
- Version Control – Make.com has a simplistic, but useful version control system for it’s workflows
- You can export a ‘blueprint’ of a scenario
- AI Assistant is pretty helpful
- Straightforward pricing, probably the cheapest of the paid workflow tools I’ve tested
- A gazillion modules (aka nodes), and even more templates for scenarios, (the ones I tested worked well, except Google docs)

Make.com Review: Con’s
No software is perfect. It’s hard for me to take Make.com seriously with it’s primary colour icons, and this list of con’s reflects some of the friction points which almost made me not want to make automations with Make.
I think Make.com manages to do well despite these weak points. Basically I didn’t hit any actual blockers, but it did feel like a throwback experience to Windows 2000 at times.
Here’s where I struggled to enjoy reviewing Make.com:
- Nomenclature & common-sense Flows
- Personally I don’t like the nomenclature (a workflow is actually a Scenario, but an AI agent is also a scenario 😵💫)
- You can create an AI agent and assign ‘System Tools’ (which are actually Scenarios 🤣)
- It seems like everything needs encapsulated in a webhook
- Language often directly reflects API, which could be confusing; “create a message” vs “send a message”
- UI
- I didn’t find the UI intuitive, despite it’s bright colours
- Delete is hidden behind right click
- ‘Scenario’ log is basic & unhelpful; on-module logs are useful
- There’s no clear way to ‘run’ an agent… you have to make a new scenario to run the agent?
- Billions of apps to start with, no common-sense ‘triggers’
- Transparency/Signposting:
- Agents are a paid upgrade, but it’s not obvious they even exist if you just start a trial account
- If you sign up for affiliate account ever, you can’t seem to use that email for actual automations – try buying a plan… infinite redirect
- The AI assistant is good for advice, but takes forever (Make.com is behind some competitors like Relevance AI here)
- No templates for AI agents (I guess they are in beta)

Make.com Review Summary
TL;DR; Make.com appears simplistic and brightly coloured, but is actually quite good under the layer of goof.
I suspect I am not the primary market for Make.com. It’s whole ethos doesn’t gel well with me as a entrepreneur engineer. I guess it targets the larger market of no-code users.
To that end I found myself looking past the seemingly simplistic wrappings; it’d be easy to right off Make.com, but underneath it’s goof is a capable, well designed system available at a modest price.

I would recommend Make.com to anyone who’s attempting AI automations for their business. I’d advise them to suspend their disbelief as they first try it; it might well be just what’s needed, (and at a competitive price).
I will continue to test Make.com and double back here if I have anything else to add. Subscribe for updates.
AI Automation on Make.com: Review Final Thoughts
I don’t know if it’s me getting used to these workflow tools, and so utilising them better, but Make.com does seem to work well.
Each workflow tool I try in theory offers the same thing, but they’re idiosyncratic, quirky, and competent in different ways. I intend to run back through all these tool reviews and update them later.
Last thought: If I was Make.com I’d add a ‘developer’ plan and make a dev-layer like Stripe did with payments
Do you use Make.com?
I’m still getting to grips with Make.com – if you’re already automating with it, I’d love to read your thoughts in the comments below.
If you’ve not tried Make.com, give it a whirl – if you use the following link and purchase a plan I’ll get a small affiliate fee, or you can just go to make.com
Leave a Reply