Process automation
Business process automation
I connect the systems in your company that don't talk to each other, and let the routine between them run by itself — without you having to change the tools your team already knows.
Business process automation means the repetitive work between your systems stops needing a person. An order from the shop writes itself into a spreadsheet, creates the invoice, sends the notification and opens a task — instead of somebody retyping it every day.
I build it on n8n and Make.com, adding AI only where something has to understand free text. A simple process starts at €390 and is usually live in 3–7 working days; more involved integrations take longer. I quote a price after I have seen your process — not before.
Where the hours actually go
In most small companies the problem is not that people work slowly. The problem is that there is a gap between the systems, and a person is the bridge across it. It usually looks like this:
- Retyping between systems. From an email into a spreadsheet, from the spreadsheet into the invoicing software, from there into stock. The same fact entered three times.
- Answering the same questions. Availability, price, delivery date, order status — daily, over and over, the same answers.
- Checking whether anything arrived. Someone opens five applications every morning just to find out whether there is anything to do.
- Manual reports. At the end of the month, numbers are copied from three sources into one spreadsheet.
- Forgotten steps. Nobody sent the confirmation, nobody followed up, the lead went cold.
Each of these takes a couple of minutes on its own. Together it is several hours a week — and worse, they are the hours in which careless mistakes happen.
What can realistically be automated
Not everything at once. I always start with one clearly bounded process that hurts the most. These are the types I deploy most often:
Intake and triage of enquiries
A form, email or message is recorded automatically, sorted by type and routed to whoever it concerns.
Orders and stock
A new order starts the chain: invoice, label, stock update, customer notified.
Documents and invoicing
An invoice arrives by email, the data is read out and written where you need it — without retyping.
CRM and follow-up
The lead is created, assigned, and the system reminds you when nobody has replied for too long.
Reports and overviews
Daily or monthly summaries assemble themselves and arrive where you actually read them.
Internal notifications
Instead of five open applications, one message in Telegram or email when something has genuinely happened.
I add AI only where it earns its place — when free text has to be understood, classified, summarised, or data extracted from a document. Where rules are enough, AI only adds cost and unpredictability.
What I can connect it to
We do not replace the tools your company is used to. The automation attaches to what you already run:
- E-commerce: WooCommerce, Shoptet and other platforms with an API
- Spreadsheets and documents: Google Sheets, Google Drive, Excel/OneDrive
- Communication: Gmail, Outlook, Telegram, Slack, Discord
- CRM and tasks: HubSpot, Pipedrive, Notion, Trello, Airtable
- Invoicing and accounting: systems with an open API or import/export
- Calendars and booking: Google Calendar, Calendly
- AI models: Claude, GPT — via API, with control over what is sent
If a tool has no API, it can often be worked around with an export, a webhook or an intermediate layer. If it cannot, you hear that at the start — not after you have paid a deposit.
How a deployment runs
- Free consultation. You describe what eats the most time. I ask about the detail of the process — who does what, when, and in which tool.
- Proposal and price. You get a concrete plan: what the system will do, what stays with a person, what it costs to run monthly, and what the deployment costs. No hidden items.
- Implementation. I build and test the workflow on real (or test) data. You see how it looks as it goes.
- Handover. Written documentation, the credentials are yours, the system runs on your accounts. You are not dependent on me.
- Tuning. The first days always turn up an exception nobody thought of. That gets ironed out as part of the deployment.
The whole thing runs in writing — you do not have to set aside time for calls unless you want to.
When I advise against automating
This is the part most suppliers leave out. Automation is not always the right answer:
When the process is not settled yet. If the way you work changes every month, the automation will be rebuilt faster than it can save anything. Settle the process first, then automate it.
When it happens rarely. A ten-minute task once a month has nothing to return. As a rough rule: below roughly 5 hours a month, deployment usually does not pay for itself.
When the problem is organisational, not technical. Automation speeds up chaos too. If it is unclear who is responsible for what, solve that first.
When every case needs judgement. If ninety percent of cases are handled individually, at most the preparation can be automated — not the decision.
If the consultation shows it is not worth it for you, I will say so. Badly deployed automation is worse than none, because people stop trusting it.
What drives the price
There is no fixed price list for automation — two companies can ask for “the same thing” and have tenfold different work behind it. The price is driven mainly by:
- Number of steps and exceptions. A straight chain versus a process with dozens of branches.
- Number of integrations. Every extra system means more testing and one more place where something can fail.
- Data quality. Clean structured data is cheap. Unstructured emails and PDFs are expensive.
- AI versus plain rules. A language model solves what rules cannot, but needs more tuning and has a running cost.
- Volume. Ten operations a day and ten thousand are different demands on the infrastructure.
- from €390One bounded process — connecting tools, automatic notifications, form handling.
- €700 – 1,200A multi-step process or an AI agent — qualifying enquiries, bookings, working with a calendar and spreadsheets.
- from €1,200Connecting several systems, an e-commerce pipeline, or an assistant over your own data.
These ranges are indicative — a frame, not a price list. On top of the deployment, budget for monthly running costs (a server or tool licence, possibly a model API); I itemise those separately in the proposal so you see the full figure.
What is actually running
Instead of percentages you have no way to check, things you can verify yourself:
- A verified client review. The tool retailer Luko Tools — automating product processing in WooCommerce and posting the products to Instagram and Facebook. The review is published on Jaspravim.sk.
- A verified review for an n8n agent. The SEO article agent for the insurance site gopoistenie.sk — every week it picks a topic from Search Console data, checks facts against official sources and prepares the article as a draft for approval. The client's review is published on Jaspravim.sk.
- A second verified review for a commissioned desktop application, published on Freelancehunt.
- Three live demo bots on Telegram you can open right now, without signing up.
- AI LAB certificate — AI Automation Expert (60+ hours of Make.com and n8n). It is a course certificate, not a vendor partnership.
Frequently asked questions
How much does automating one process cost?
A simple bounded process starts at €390. Multi-step solutions with AI fall in the €700 – 1,200 range, and connecting several systems starts at €1,200. You get a ballpark range after the initial consultation. The exact price depends on scope. If the project requires a detailed review of the systems or a technical design, the next step may be a paid technical analysis — without one it would be guesswork.
How long does deployment take?
A simple process usually 3–7 working days from approval of the proposal. Connecting several systems takes longer, and you get the date in the proposal rather than halfway through the work.
Do we have to change the tools we use?
No. The automation attaches to what you already have — shop, spreadsheets, email, CRM, invoicing. Changing tools is a separate decision and not a precondition.
What if the tool we use has no API?
It can often be worked around with a scheduled export, a webhook or an intermediate layer. If it cannot, you find out at the consultation — not after paying a deposit.
Who owns the system and where does it run?
It runs on your accounts and the credentials are yours. You get written documentation covering how it works and what to do when something changes. You are not dependent on me and can hand it to anyone else.
What happens when the automation fails?
Every process has error handling — when something fails the system does not drop it silently; it sends a notification and the case can be handled by hand. Automation must never lead to an order or enquiry quietly disappearing.
Is automation acceptable under GDPR?
It depends where the data ends up. If it is sensitive, the solution can be built on self-hosted n8n so the data never leaves your infrastructure. Where an AI model is used, I state in advance what is sent and what can be excluded. Nexflow does not provide legal advice — for sensitive processing, consult your DPO.
Do you work outside Banská Bystrica?
Yes. The whole process runs in writing and remotely, so location is irrelevant.
Sources
- n8n — official documentation — self-hosting and integrations
- Make.com — official documentation — operation model and scenario limits
Related articles
Related work
Related services
Which process costs you the most time?
Tell me what you do by hand. I will tell you free of charge whether it is worth automating and roughly what it would cost — even if the answer is that it is not worth it yet.
Get a free consultation →