An agent that reads your emails and drafts replies, another that qualifies your incoming requests, a third that keeps your sales dashboard up to date. Two years ago, this took a developer and several weeks. Today, with Make, an SME leader can build their first agents by assembling blocks on a screen. You still need to know what the tool actually does, what it doesn't, and where to start.
Key Takeaways
- Make (formerly Integromat) is a visual automation platform born in Prague, acquired by the German software company Celonis in 2020
- Its AI agents, launched in spring 2025, add decision-making to traditional automation
- A first email-sorting agent can be built in one to two hours and can save several hours a week
- Since August 2025, Make bills by credit: AI modules use more than a standard module
- Data is processed in data centers located in the European Union, a GDPR advantage, but Make isn't self-hostable
What exactly is Make?
Make, formerly Integromat, is a no-code automation platform born in Prague in 2012. It took the name Make in February 2022, sixteen months after being acquired by Celonis, a German company specializing in business process analysis. The company reports more than 400,000 client organizations in over 200 countries.
The principle comes down to one sentence: visual scenarios where each block is a module. You connect a trigger (a new email, a new row in a spreadsheet) to a chain of actions (update the CRM, send a notification, generate a document). The scenario runs every time it's triggered.
Make stands out for two things. First, its visual editor: you see the data flow moving from one block to the next, including branches and loop-backs. Second, the depth of its integrations, more than a thousand connectors, with access to connected services' functions that goes deeper than just sending a message.
AI agent vs. traditional automation: the difference
Traditional automation follows a fixed path. You define each step in advance. If a case falls outside that path, the automation fails.
An AI agent receives a goal and tools, then decides the steps itself. In practice:
| Criterion | Traditional automation | AI agent |
|---|---|---|
| Logic | Fixed path | Autonomous decision |
| Adaptability | None | Adapts to content |
| Setup | Every step by hand | A goal and tools |
| Example | "For every email, send an acknowledgment" | "Read my emails, sort them, draft replies for the urgent ones" |
Automation follows a path; the agent chooses its own.
What Make's AI agents can do
Make launched its AI agents in spring 2025. In practice, two levels are available to you.
One-off AI modules: a block that summarizes a text, sorts an email, extracts fields from an invoice. This is AI inserted as a building block into a traditional scenario.
Full AI agents: an agent block that receives an instruction, chooses its tools, and chains actions together until it reaches a result. It can read an email, query a database, prepare a reply, and submit it for human validation before sending. It's this second level that changes the work: the agent handles cases that fall outside the plan, not just the ones anticipated in advance.
The most profitable uses for an SME: sorting and replying to emails, qualifying quote requests, processing supplier invoices, updating the CRM from exchanges, preparing meeting notes. You'll find other concrete examples on our use cases page.
Building your first agent with Make
The shortest path starts with a simple case: sorting emails. Here are the steps.
Step 1: create the scenario
In Make, create a scenario. The trigger is the inbox: every new email starts the scenario. Make offers connectors for Gmail, Outlook, and most email services.
Step 2: connect the language model
Add an AI module. Make connects to the leading models (GPT, Claude, Gemini, Mistral, and others) via an API key. That's the only technical piece of data you need: a key provided by the model's provider, entered once into Make.
Step 3: write the instruction
This is the step that determines the agent's quality. The instruction spells out the role, the rules, and the expected format. For example:
You are an email assistant for an accounting firm. Sort each email into four categories: urgent, to handle, informational, spam. For urgent ones, prepare a draft reply. Never reply without explicit validation.
Step 4: wire up the outputs
Depending on the category, the scenario routes to different actions: creating a draft, adding a label, archiving, a notification in Slack or Teams. Each branch is wired up in a few clicks.
Step 5: test, adjust, monitor
Send about ten varied test emails and watch the agent work. Adjust the instruction if the classification goes off track. Once it's in production, monitor credit consumption: that's the point to watch specifically with Make.
A first agent of this kind can be built in one to two hours and can save several hours of sorting each week.
Make and confidentiality: what you need to know
This is the point to look at closely for a French SME. Make is a cloud platform: your scenarios and your data pass through the vendor's servers, not your own.
Two important nuances. First, Make is a European company, based in Prague, and processes data in centers located within the European Union. For GDPR compliance, that's a real advantage over platforms hosted in the United States. Second, the company holds solid certifications: SOC 2 Type II, ISO 27001, GDPR compliance.
But for data that must never leave the company, it needs to be said plainly: Make isn't self-hostable. A firm handling highly sensitive client files would do well to look at tools you install on your own servers. That's precisely the role of expert support: choosing the right tool for the level of confidentiality required. Law firms and accounting firms don't face the same constraints as a retail business.
Make, n8n, or Zapier: how to choose
Three platforms, three different logics.
| Criterion | Make | n8n | Zapier |
|---|---|---|---|
| Model | Cloud, visual editor | Open source, self-hostable | Cloud, the simplest |
| Integrations | Over a thousand | Several hundred nodes | The largest catalog |
| Pricing | By credit since August 2025 | Per execution, free if self-hosted | Per task |
| Confidentiality | Data processed in the EU | 100% on your premises if self-hosted | Hosted in the United States |
| Audience | SMEs that want visual power | Technical teams, high confidentiality needs | Immediate need, no technical skill |
The simple rule: Zapier to get started fast and simple, Make for visual power with data processed in the EU, n8n for full control and data that never leaves.
For an AI agent, Make and n8n are the two serious choices. Make is more accessible: the visual editor guides the build, and the learning curve is gentler. n8n requires a bit more technical skill but offers self-hosting.
FAQ
Do you need to know how to code to build an AI agent with Make?
No. The interface is entirely visual: you assemble blocks and fill in instructions in plain language. The only technical step is getting an API key from the language model's provider, then pasting it into Make. If you can set up an email rule, you have the skill level needed.
How much does an AI agent on Make cost?
Make offers a free plan (1,000 credits a month), then paid tiers starting around ten dollars a month, billed annually, for 10,000 credits. AI modules and code execution use more credits than a standard module. On top of the credits comes the cost of using the language model, billed by its provider on every call.
Is Make GDPR-compliant?
Yes, in principle. The company processes data in centers located within the European Union, holds SOC 2 Type II and ISO 27001 certifications, and provides contractual clauses. What's still up to you: choosing the right data center, documenting the subprocessor, and checking what you send within the scenarios. For the most sensitive data, self-hosting remains the only guarantee that nothing leaves.
Can Make be self-hosted?
No. Unlike n8n, Make is only available as a cloud service. If your data must never leave your infrastructure, look toward self-hosted n8n, possibly paired with a local language model.
What's the difference between Make, n8n, and Zapier?
Zapier is the simplest and offers the largest catalog of apps, but its per-task billing penalizes complex scenarios and its data is processed in the United States. Make offers the most powerful visual editor with data processed in the EU. n8n is open source, self-hostable, and the only one of the three that can run entirely on your own premises.
The right tool, at the right time
Make is a good entry point for an SME that wants to build its first AI agents without coding, with data processed in Europe. It isn't the answer to everything: for data that must never leave the company, you need a self-hosted tool.
That's where NexeAI comes in. We build your agents with you, on Make or on whichever tool fits your constraints, then we train you to become self-sufficient. No black box, no perpetual subscription, no dependency on a provider.
Let's talk about your project to identify your automatable processes.
To go further, check out our dedicated AI agents page, our automation offer, or our training programs.
