Automation for small businesses: n8n, Make, and 5 examples that pay off within a week

When people say "process automation," most small business owners picture an expensive ERP system and a six-month implementation. The reality of 2026 is completely different. By combining tools like n8n and Make, custom scripts, and AI models, you can do in a few days what required a team of programmers 5 years ago. And the ROI is measured in weeks, not years.

With every inquiry, we first ask one question: n8n, Make, or a fully custom solution? And if AI belongs there, at which point. Here is the decision framework and 5 real examples of process automation that we deploy most often for Slovak companies.

Process automation: three paths (n8n, Make, custom code)

There is no "one right solution." The choice depends on complexity, volume, and data sensitivity:

1. Make (formerly Integromat)

A cloud SaaS with a visual editor. The fastest start without a technical person. It is a good fit for your first scenarios while you verify what actually makes sense to automate. The downside: you pay per operation, which adds up quickly at higher volumes. Sensitive data flows through the cloud - for GDPR-strict projects, that is often not acceptable.

2. n8n (self-hosted)

An open-source counterpart to Make that runs on your own VPS. A more expensive entry (it has to be deployed and maintained), but incomparably cheaper as you grow, and sensitive data never leaves. This is our most frequent choice for small and mid-sized businesses that want to live with automation long term.

3. Custom solution (Node.js, Python, PHP)

When the logic does not fit any visual tool - complex conditions, a specific integration, high performance, essential stability - we write it as a custom service. It runs on the same VPS as n8n, or completely standalone. More expensive once, cheaper forever - no dependency on someone else's tool, zero monthly platform overhead.

4. Hybrid (most common in practice)

The reality: most projects are a combination. n8n handles orchestration between services, a custom Node.js script does something that n8n does not handle elegantly, and an AI model in the background fills in what a person would spend five minutes on (classifying, summarizing, transcribing, searching).

AI as a quiet collaborator in automation

Today we almost always add an AI layer to automation - not as a gimmick, but as a concrete module that replaces 5 minutes of manual work on every run. A few examples:

• Lead classification - GPT-4o-mini or Claude Haiku reads the message from the form and labels it as "commercial inquiry / support / spam / partnership." n8n then routes it to the right person.
• Summarization and extraction - from a long client email we pull the date, address, and amount, and store them in the CRM as structured fields.
• Generating replies - a draft of the initial reply for the salesperson. A human still reads it and sends it, but saves 80% of the typing.
• OCR and invoice parsing - from a PDF invoice we pull the company ID, tax ID, amount, and due date, and drop them into accounting.
• Transcribing voice messages - Whisper transcribes a WhatsApp voice message into text, and AI turns it into an action item in Notion.

At lower volumes this costs us a few cents per operation with the OpenAI / Anthropic / Google API; at higher volumes we sometimes deploy our own open-source model (Llama, Qwen) on the same VPS - zero per-call cost.

5 real automations that pay off the fastest

1. New lead from the website → CRM + AI triage + Slack notification

Before: A form on your website sends an email to info@. Someone forwards it to the salesperson. The salesperson manually retypes it into the CRM. Spam goes to the trash only after being read.

After: n8n catches the lead, AI classifies it (commercial / support / spam / other), saves it to the CRM (Pipedrive, HubSpot, or even just Google Sheets), sends the client a personalized reply within 1 minute, and the salesperson gets a Slack message with an AI summary of what matters most.

Saves: ~3 minutes per lead × 30 leads per month = ~1.5 hours per month. The main benefit is not time - it is that no lead is lost and response time drops from hours to seconds. A lead that gets a reply within 5 minutes has a several times higher chance of converting.

2. E-shop order → invoice + accounting + email to the client

Before: Once a day the accountant opens WooCommerce, clicks through each order in SuperFaktúra, downloads the PDF, and sends it by email.

After: n8n watches the order status in WooCommerce. When it changes to "Completed," it creates an invoice in SuperFaktúra, adds it to accounting in Pohoda, and sends the client a PDF with a custom-branded email. For B2B, AI also generates a short personalized message ("thanks for your order; our customers most often ask about X - here is the answer").

Saves: ~2 minutes per invoice × 100 invoices per month = ~3 hours per month. At 200+ invoices a month it is measured in days.

3. Weekly report from multiple sources → 1 email on Monday morning

Before: On Monday morning the marketing manager opens 5 tabs (GA4, Google Search Console, Meta Ads, Google Ads, the e-shop), puts together a report in Google Docs, and sends it to the director.

After: n8n pulls data from all the sources on Sunday evening, AI writes a short commentary ("this week organic traffic to article X grew significantly, it is worth pushing some ads behind it"), and at 7:00 on Monday it is sitting in the inbox. With green and red arrows compared to the previous week.

Saves: ~45 minutes per week. And more importantly, the director actually reads that report - because it arrives on its own, short, with a concrete commentary, and at the same time.

4. Monitoring competitor prices / product availability

Before: Once every 2 weeks someone opens 10 competitor pages and notes in a spreadsheet where they are changing prices.

After: A custom Node.js script (n8n does not have enough power for web scraping at higher volumes) downloads the prices of selected products at competitors once a day, compares them with the previous day, and if they change by more than an agreed %, sends a message to Slack. AI adds context to the change ("this product was discounted for the first time in 3 months").

Saves: 1-2 hours per week + a reaction advantage over competitors in seconds. At a larger e-shop with 50+ key SKUs, this often translates directly into revenue.

5. Onboarding a new client / employee

Before: For every new person, HR or the office manager manually creates accounts in Google Workspace, Slack, and Notion, project managers add them to projects, and prepare a welcome pack.

After: The manager fills out 1 form (name, position, start date). n8n creates the accounts, sends the access credentials, adds them to the right Slack channels, creates their page in Notion with an onboarding checklist, and emails them. On top of that, AI prepares their first 7-day plan based on the position.

Saves: ~2 hours per employee. At a company where 1 person joins per month, that is not staggering. At an agency growing by 3 - 5 people a month, it is measured in days.

Ballpark prices - just for orientation

So that you have a rough idea of the kinds of amounts we typically talk about with automation (the numbers are only indicative, the actual price always depends on complexity, the number of connected systems, and whether an AI layer is part of it):

• A simple automation without AI (1 trigger → 2-4 actions, like "lead → CRM + email"): roughly 200 - 450 €, done within a week.
• Moderately complex with an AI layer (classification, summarization, integration of 3+ systems): on the order of 500 - 1 500 €, done within 2 - 3 weeks.
• A custom solution (custom script, custom AI model, specific logic): typically from 1 500 € upward, depending on scope.
• n8n hosting + monitoring + minor tweaks: a flat fee from roughly 39 €/month.

For comparison: a typical full account in Make costs around 25 - 50 €/month just for the tool itself (without building the scenarios). Self-hosted n8n is a fraction of that price with equal or greater flexibility - but it needs a technical partner.

We can give you a more precise number for your specific case after a short consultation - that is the only way to honestly assess whether a given process is worth automating and which path to take. The manual version can sometimes be the more economically sound option, and we will say so honestly.

When automation does not make sense

Let's be honest - not everything is worth automating:

• A process you do 3× a year. Doing it manually will be faster.
• A process that changes every 2 months. Maintenance will eat the ROI.
• A process you have not documented yet. Write it down on paper first - automating chaos is just faster chaos.
• A process where AI gets 80% of it wrong. Soberly test a sample before you put it into production.

Conclusion

Automation is not "for big companies with an IT department." It is exactly the opposite - the biggest multiplier for a small business, where every saved hour goes straight into what you do best. The key is to pick the right tool for the right problem: in some places Make is enough, elsewhere n8n pays off, somewhere the only path is custom code, and an AI layer is surprisingly often a good fit.

Do you have a process that costs you a lot of time manually? Write to us - we will take a look at it and tell you which path makes the most sense and what the estimate is. The consultation is non-binding, and we will reply within 24 hours.

FAQ

Frequently asked questions.

What is the difference between n8n, Make, and a custom automation?

Make is a cloud SaaS, easy to start with, but you pay per operation. n8n is open-source, you host it yourself, cheaper as you grow, and sensitive data does not travel anywhere. We choose a custom solution (Node.js, Python, PHP) when the logic does not fit visual tools or you need maximum performance and zero monthly overhead.

How is AI used in automation?

Most often as a quiet layer that replaces short manual work - lead classification, email summarization, extracting data from invoices, generating draft replies, OCR, or transcribing voice messages. We choose between cloud APIs (OpenAI, Anthropic, Google) and our own open-source model depending on volume and data sensitivity.

How much does building one automation cost?

As a rough guide, simple scenarios without AI run around 200 - 450 €, moderately complex ones with an AI layer on the order of 500 - 1 500 €, and custom solutions from 1 500 € upward. These are only ballpark numbers for orientation - we can refine the exact amount after a short consultation and based on the brief, always as a fixed price upfront.

What systems can it connect to?

With almost anything that has an API - Gmail, Slack, Notion, Google Sheets, WooCommerce, Shopify, SuperFaktúra, Pohoda, Stripe, HubSpot, Pipedrive, Meta and Google Ads, ChatGPT, Claude. And when something has no native integration, we add it via a generic HTTP request or a custom connector.

Is my data safe in the automation?

With self-hosted n8n or a custom solution, data never travels outside your own server - that is the main reason we recommend this path for sensitive data and GDPR-strict projects. With the AI layer, we can also deploy our own open-source model locally if there is a requirement that data not leave the infrastructure.