n8n and Make automation templates: what they are, how they differ and where to start

n8n and Make automation templates: what they are, how they differ and where to start

Automation is when services are wired to each other and the work between them happens by itself: an enquiry comes in from the site, lands in a spreadsheet, and a notification goes to the manager. Nobody moves anything by hand.

Automation is when services are wired to each other and the work between them happens by itself: an enquiry comes in from the site, lands in a spreadsheet, and a notification goes to the manager. Nobody moves anything by hand.

A template is one of these chains that somebody has already built. All that is left is to plug in your keys and run it.

Below: how it works, how n8n differs from Make, how to put a model into a chain, what to check before importing someone else's template, what people are paid for this work and where to start so you get a result in an evening rather than drowning in theory.

The guide is live: the platforms and the template catalogue change, and this piece changes with them.

The short version

A chain has three parts: what starts it, what happens, where it lands. Everything else is detail.

Start from a ready template, not a blank canvas. Most tasks are typical and the solution already exists: the automation templates section holds ready n8n and Make process files — import, enter your keys, done.

Describe the process first, automate second. Otherwise you cement the mess and make it harder to fix.

The saving comes from frequency, not complexity. A ten-minute daily task is about forty hours a year.

n8n or Make is a choice between control and simplicity, and on eighty percent of tasks there is no difference.

There is money here. On the market a system or a pipeline runs $490–1 460, against a median one-off job of around $60.

Part 1. How n8n and Make work

Three parts in any chain. Understand them and you can read and edit anyone else's template.

What starts it. The event everything begins from: a time arrives, an email lands, someone submits a form, a new row appears in a spreadsheet.

What happens. Steps in sequence. Each takes the previous result and passes it on: fetch the data, call a service, write a row, send a message.

Where it goes. Data flows from step to step. The last step puts the result where it is needed.

It is assembled visually, in blocks, with the mouse. The entry threshold is roughly that of setting up email filters, not of programming.

A template is one of these chains prepared for a typical task. For example: "new enquiry from the site → enrich the data → write to the CRM → notify whoever is responsible". Import it, plug in your access, run it. Minutes instead of designing from scratch.

Part 2. n8n or Make: which to choose

Both solve the same problem — visual automation. The difference is in emphasis.

n8n is about control. It can be run on your own server, take your own code, and keep the logic and the data with you (official documentation).

Make is about simplicity. More ready connections to popular services out of the box and a lower entry threshold.

Choosing in two minutes

Four questions, and the answers decide it without reading a single review.

Are you doing this for yourself or to sell? For yourself — take whichever you will learn faster. To sell — whichever you can hand over and explain to a client.

How many operations a month? Hundreds — there is no difference in money. Tens of thousands — do the maths, that is where it appears.

Is there data that cannot leave your building? If yes, the question is settled: you need the one you can deploy on your own server.

Are you prepared to dig in if a ready block does not exist? No — visual builder. Yes — the platform you can extend with code.

The general rule: start with the simpler one. Move when you hit a specific limit — "there is no block for this", "too expensive at this volume". "People say the other one is more powerful" is not a limit.

Your own server: when it is justified

The main distinction of n8n is that you can run it yourself. That is not a feature for enthusiasts but the answer to three situations.

Privacy. Data and keys do not go to a third-party service. For sensitive scenarios that is not an advantage but a condition.

Control. You manage the version, the limits and which blocks are available.

Economics at volume. For intensive automations, running your own instance often costs less than paying per operation.

The price for it is that you maintain the server yourself.

Part 3. How to put a model into the chain

A chain becomes "intelligent" when a step calling a model is added: it takes the data from the previous steps, sends it to the model and passes the result on.

That is how automation stops being a rigid if-then diagram and starts working with text, images and everything else that does not fit neatly into fields.

The rule that saves money

Not everything should go to a model.

Automation runs on rigid rules. Predictable, cheap, does not make mistakes. But it does not understand content: it cannot read an email and decide whether it is a complaint or an order.

A model understands. More expensive, more complex, less predictable.

The split is simple: mechanical work to a rule, work requiring understanding to a model. An error in either direction costs money: a model where a rule would do is expensive and less reliable; a rule where live language is involved simply does not work.

And the practical conclusion: most of what people reach for agents to do is solved by an ordinary chain.

The next step up is an agent step: the model does not just answer but chooses which tool to call. How that works, what it costs and how to run it safely is in the AI agents hub. And the knowledge of "how to perform this procedure properly" gets packaged into skills — the Skills for Claude Code hub.

Part 4. The five tasks people automate most

Almost every working chain comes down to five classes, and the library has ready templates for each.

Taking enquiries. Contacts from every source — form, email, messenger, social — are gathered into one place and whoever is responsible gets a notification.

Content. On a schedule or an event, material is assembled (often with a call to a model) and published to the right platforms.

Notifications. The chain watches a source and warns the team when something happens.

Data synchronisation. Rows across spreadsheets, CRMs and databases are kept consistent without manual copying.

Reports. On a schedule, figures are collected from several sources and turned into a summary.

The value of ready templates is precisely that all five have already been built and checked by somebody. No need to design from scratch — take the closest scenario and adjust it.

Part 5. One template taken apart in full

So that you can read any other one. Take a typical case: enquiry → data enrichment → CRM → notification.

What starts it — a form on the site: on submission it sends the data into the system.

Step one takes the enquiry fields: name, contact, message.

Step two enriches them: data is pulled from an external service against the contact so the manager can see straight away who this is.

Step three writes the enquiry to the CRM.

Step four notifies whoever is responsible: it takes the link to the created record and sends it to a messenger.

Data flows between the steps: the CRM takes fields from the enrichment result, the notification takes the link from the CRM result.

What to check before launch

What happens if the external service is down. Enrichment is a nice-to-have, not a requirement. The enquiry must reach the CRM regardless, not disappear along with the outage.

What happens if the form is submitted twice. Someone clicked twice — two identical clients must not appear in the CRM.

Where the keys are kept. In the platform's secure storage, not in the scenario text, which can be exported along with the credentials by accident.

Take one template apart like this and you can read any of them: the "trigger → steps → data" structure is the same everywhere.

Part 6. Reliability: what to do about failures

A chain is useful exactly as far as it survives failures. External services occasionally answer with an error or stay silent, and without preparation such a failure breaks the chain quietly — you spend a week thinking the process is running when it has stopped.

Three techniques.

Retries on steps that call unstable services. A short pause and a repeat often resolves a temporary failure.

A separate branch for errors. A step fails and the chain does not stop but takes the fallback route: it sends a warning and stores the data to retry later.

Duplicate protection. A repeat run on the same data must not create a second record.

And the main rule: any chain must be able to report an error. A failure notification matters more than the automation itself — without one you learn about the problem when a client asks why nobody replied.

A good template already contains an error branch. If it does not, add one, or the first failure will mean quietly losing data.

Part 7. Security: keys and access

A chain holds the keys to your services, so its security is first of all access management.

Minimum permissions. Give a scenario access only to what it genuinely needs, not "just in case".

Keys in storage, not in the text. A secret written directly into a step travels out with the exported scenario.

Read other people's templates before importing. Look at which services the scenario calls and what access it asks for. That is exactly why descriptions in the catalogue are open before download — you assess the template and its appetites in advance.

If the scenario contains an agent step, the same measures apply as for agents in general: isolation and confirmation on destructive and payment operations. Covered in detail in the AI agents hub: automation with an agent inherits all of the agent's risks.

Part 8. When you have many automations

What comes to the fore is not the individual scenarios but maintaining them.

A run log. The execution history you use to take failures apart. For critical scenarios, a warning on failure.

Versions. So you can roll back if a new edit broke something.

Separate environments. A test one for checking edits, a live one for work. Experiments must not touch what is running.

And a review once a quarter. Fifteen minutes to check that every chain is still needed and still doing what it should. The most insidious breakage is not a failure but a chain that faithfully keeps doing something no longer relevant.

As load grows, cost also comes to the fore: intensive scenarios calling models start spending noticeably. The same techniques work here as in token economics: cache the unchanging part, a low-cost model by default, count the result rather than the text. Covered in detail in the LLM pricing and token economics hub. It is precisely at the junction of volume and cost that your own server usually pays for itself.

Part 9. Content automation: what it looks like

A popular class in its own right, and a good illustration of how a model fits into a chain.

The typical scenario: once a day a schedule fires → a step gathers the source data (topic, source, brief) → a model step generates the text → where needed the next step calls image or video generation → the final step publishes or puts the piece in a review queue.

The key point: an "intelligent" step is an ordinary step in the chain. It takes the data from the previous ones and passes the result on, like any other.

Two pieces of advice that save money.

Keep a human check where a mistake is expensive. Fully automatic publishing only for proven low-risk formats. Material that goes out without a human eye will eventually go out badly.

Count the spend. Content pipelines with text, images and video spend noticeably. The same techniques apply: a low-cost model by default, cache the unchanging part, a draft before the expensive final.

Ready content scenarios are in the automation templates section, and assembling prompts for images and video is conveniently moved to the prompt generator.

Part 10. What people pay for this

The section missing from almost every other piece about automation. It gets written about as engineering — and earned from as a service.

The three things bought most often

Enquiries gathered into one place. A business receives contacts from four or five sources and loses some of them. The owner knows his revenue and does not know his losses: an enquiry that arrived in the wrong place exists in no report at all.

Bringing existing clients back. A reminder to people who have not been in touch for a while. The fastest payback in the whole set: these people do not need to be brought in by advertising or convinced of who you are.

Regular reporting. Not to save an hour, but so that the report starts existing. Most small companies do not have one — not because they do not want it, but because every time somebody has to be chased for it.

What all three have in common

None of them is about complex engineering. All three are about what is not being done at all, rather than about speeding up what is.

That is the key to the sale. Automation that speeds up something already happening sells badly: the client counts the saved hours and sees no gain. Automation that creates something absent sells easily.

Frame it through absence: "right now enquiries at night are lost", "right now nobody works the old client list", "right now there is no report".

How to approach a client

Not with a proposal about automation. With one question: how many enquiries arrive outside working hours, and what happens to them.

The owner almost never knows the exact answer. That not-knowing is your sale.

The order of prices

From the Parselona feed — 5 986 jobs over 4 months: a system or a pipeline runs $490–1 460, against a median one-off job of around $60.

An honest caveat: an explicit budget appears in 17% of listings, and by individual type of work there are between twenty and three hundred observations with a figure.

A quote is built as a composition, not a single number: an inventory of sources, building the chains, texts and rules, failure notifications, documentation and training, what the client pays for directly from here on, and support. Seven lines defend themselves — people argue with the lines, not the total.

And the main money is not in the first build. Services change their interfaces, chains break, the client's processes get rebuilt. Support is a recurring payment, and an honest one.

In the scrape, 23% of clients are looking for someone long-term and 12.9% talk about ongoing volume. A third of the market does not fit into one-off work.

Three mistakes when selling

Selling the platform. The client does not care what it is built on. The conversation is about the result.

Automating before describing. A process done differently every time will be cemented in exactly that state.

Promising staff reductions. Even if it is true. One mention turns the client's employees into opponents at every step where you need them to be honest about how the processes work. The right framing: the freed-up time goes to the things nobody ever got round to.

Part 11. If Make and n8n are more than you need

An honest fork, worth stating before you head into a builder.

Make and n8n exist for linking services to each other: an enquiry from the site into the CRM, an email into a spreadsheet, an order status into a notification. That is their job and they do it well.

But most "I want to automate this" requests are not about that. People want to produce: content, documents, images, clips. There the job is not to link two services but to move material through several processing steps.

There is another route for that — Cerebrum and the AI SKILLS connector for Claude. In Cerebrum you describe the task in plain words: the right Gen AI models are picked automatically, you pay from your balance for what is actually done, and the price is visible before launch. The connector (Full plan) works inside Claude: you describe the task and it generates frames, video and voice with Gen AI models and assembles an expert clip with subtitles, graphics and sounds. Single generations are also available directly in Gen AI.

Who needs what

Linking services, handling enquiries, synchronising data — Make or n8n.

Producing content, documents, visuals — Cerebrum or the AI SKILLS connector for Claude. You do not have to learn a builder.

Both — take both: they do not compete, they cover different halves of the work.

Part 12. Where to start

Better to start from a ready template for a near-enough task than from a blank canvas: you see a working chain straight away and adjust it to yourself.

The automation templates section holds over 3 000 ready scenarios with open descriptions in two languages. What a scenario does and why it is useful is visible before download; the files themselves open with a subscription. The catalogue grows daily.

The route: open the section → find a template for your task → read the description → import → plug in the keys → run on test data.

What to look for in the description

Five questions before importing:

The open card description is exactly what lets you answer all of this before downloading, rather than spending time on a scenario that needs a service you cannot reach.

Part 13. What to open for your own task

Six steps, from free to serious. Do not take the top one if a lower one is enough.

1. Ready templates — free to look at, Creator to use

Over 3 000 scenarios in the automation templates section: handling enquiries, mailings, reports, integrations between services.

Separately — ready agents. Not just if-then chains but scenarios calling a model: sorting an inbox, qualifying enquiries, preparing replies.

Descriptions are open to everyone without registration: what a scenario does and which keys it needs is visible before download. Files come with the Creator plan (and Make, N8N), advanced ones with Full.

2. Cerebrum and the AI SKILLS connector — if the task is about production

Content, documents, visuals. In Cerebrum you set the task in plain words, models are picked automatically and the price is visible before launch. The AI SKILLS connector for Claude comes with the Full plan.

3. The course “Automator” — Plus plan

Automating routine with ready AI tools: working with files (reports, contracts, books, translations), transcription. The Plus plan also opens the courses “Photo and Video” and “Editing”.

This is the answer to "where do I start" if you want understanding rather than a ready template.

4. Express course “AI Agents” — $9

From your first automatic agent to chains running around the clock. 20 ready projects and four platforms: Make, n8n, Relevance, Lindy.

Included in the Make, N8N and Full plans. Not part of the three free Basic days: a one-off purchase, a plan, or 1 000 experience points.

Take it if you want to understand agent scenarios but have not yet decided which platform to settle on.

5. The Make plan — the Make course in full

The course “Automations in Make” in full (23 lessons: the first scenario, iterators and aggregators, webhooks, error handling, the data store, three practical tutorials) plus the express course “AI Agents”. The plan includes everything from Creator, so all the regular templates in the catalogue too, CRM ones included, for example the RAG chatbot for Bitrix24.

Take it if your tasks are integrations with services and CRMs and you would rather not deal with infrastructure.

6. The N8N plan — the n8n course in full

The course “Automations in n8n” in full (31 lessons, including the “AI agents and prompts” and “RAG agents” modules) plus the express course “AI Agents”. In the catalogue for this plan, for example, the RAG pipeline for Google Drive documents and sales team monitoring across chats and calls (an advanced template: Full, or $9 one-off, or 1000 XP).

The skills you come out with differ from Make: self-hosting through Docker and a VPS, corporate document search systems on vector databases, enterprise-grade fault-tolerant automations.

Take it if you need control, privacy and your own server.

Full — both platforms and the archives

Everything from every plan: both courses (Make and n8n), all advanced templates and skills, the course “Working with AI software” (local installation and use of Stable Diffusion and ComfyUI) and the AI SKILLS connector for Claude. The level for people building private environments for companies.

Choosing between Make and N8N by plan

The same question as at the start of the article, but now about the training.

Tasks are integrations with services and CRMs, infrastructure is of no interest → Make.

You need your own server, data privacy and large volumes → N8N.

Undecided → start with the $9 express course: four platforms at once, and the choice becomes obvious afterwards.

What sits alongside

200+ AI assistants across 14 categories — for tasks requiring understanding rather than a rule. Basic plan.

Separately, the Bot architect in the Engineering category: it writes system instructions for conversational systems, including the rules for handing a conversation to a live person.

The Skills for Claude Code catalogue — over 10 000 procedures for agents.

The Open Source section — open tools and things you can deploy yourself.

The prompt generator — if there is generation anywhere in the chain.

The current picture

The automation and agent layer moves fast: harnesses are becoming products in their own right, models get cheaper and more expensive, new solutions ship almost weekly. Fresh releases and techniques are covered in the daily AI digests.

For how agents work, see the AI agents hub; for open models, Best open LLMs 2026; for ready scenarios, the templates section.

FAQ

What is the difference between n8n and Make, in plain words?

n8n is about control: it can run on your own server, take your code, keep the data with you. Make is about simplicity: more ready connections and a lower entry threshold.

Does that matter to you? On most typical tasks, no — you will not notice the difference. It appears when the data is sensitive or the volume has grown to tens of thousands of operations a month.

Do I need to be able to program?

No. A template is imported and configured visually: you plug in your keys and data, the logic is already built by the author. Code may be needed only for non-standard processing, and not always then.

That is exactly why it is worth starting from a ready template for a near-enough task rather than building from scratch: you get a working chain immediately and learn on it.

Can a model be built into an automation?

Yes: a step calling a model is added to the chain, and for complex tasks an agent step that chooses tools itself.

But check first whether it is needed: if the task is describable as "if A then B", a rule will be cheaper and more reliable. Agents are covered in detail in the AI agents hub.

Can n8n be run for free?

It can be run on your own server — that is the platform's key distinction. This removes per-operation charges, leaving the cost of the server and its maintenance.

For a team that values privacy and control, that is often cheaper and more reliable in the long run. The price is that you hold the server, the updates and the backups yourself, so smaller projects find it easier to start in the cloud and move later. The requirements are in the official documentation, and ready templates work the same either way.

Where should a beginner start?

Not with learning the platform in general, but with one specific task you currently do by hand.

Keep a list of what you have done more than once this week. Run it for a week and automate nothing. Whatever appears three times is your first candidate.

Then find a ready scenario for a near-enough task in the templates section, read the description, import it, plug in the keys and run it on test data. In an hour you have a working result instead of theory.

How much can be earned from automation?

From the Parselona feed — 5 986 jobs over 4 months — a system or a pipeline runs $490–1 460 against a median one-off job of around $60. Those are different units of measurement: one is an operation, the other a result with stages and a handover.

And the main money is not in the first build but in the support: services change, chains break, processes get rebuilt. Almost one client in four in the feed is looking for someone long-term — and automation is exactly the kind of service that fits.

Try it

Registration is free. Three days of the Basic plan switch on with one button, no card needed: 200+ AI assistants, the prompt texts, the skill files, five programme courses (including "Sales") and six Basic mini-courses.

One action right now. Keep a list of the tasks you did by hand more than once this week. Do not try to fix anything — just run the list for a week.

The list will show you where to start more accurately than any audit.

What to take on the platform for this task

Regular templates open on the Creator, Make, N8N and Full plans.

What to read next

Make or n8n: which to choose — decided by four questions in two minutes.

Automation without a developer — where to start if you are not a developer.

Three automations people buy — what gets ordered and how to sell it.

A company knowledge base as a service — RAG agents and why they command more.

Local models for sensitive data — self-hosting as a condition of the deal.

Auditing the processes first — what is done before automating.

Digitising an offline business — where clients are actually lost.

Auditing a sales team from call recordings — a system and its junior version.

Build once, runs always — the alternative when the task is about production.

AI for business owners — what this looks like from the client's side.