Skip to content
EN
Start a project

00 / Workflow automation · n8n · AI

Workflow & AI automation.

An email comes in, and the rest runs on its own: the details land in your inventory system, POS or ticket system, and you get a report. We build these workflows with n8n or as a program of our own. An AI model only goes where someone has to understand freely written text.

Start a projectWith AI or without

Intro call €0Cost estimate in four questions

01 / What we automate

Automating processes someone handles by hand today.

It pays off where the same steps come up every week: typing things up, forwarding, copying into a second system. We build these workflows from Duisburg, for shops, practices, newsrooms and support teams across Germany.

The whole AI and automation area ↗
By handEmailRetypingPOSAutomatedEmailWorkflowPOSReport after every run
  • 01

    Processing emails and attachments

    Lists, invoices and forms arrive by email. The workflow reads the attachment, pulls out the details and enters them into the right system as separate fields. If something is unclear, it asks back by email.

  • 02

    Connecting systems without an API

    Many POS systems and portals have no open API. Then our program operates the interface in the browser, the way a person would. That’s how we enter items and prices into the Reacard POS at Turkish Home.

    Turkish Home inventory ↗
  • 03

    Sorting tickets and inquiries

    A model reads every new message and returns three values: category, department, urgency. That’s how AI ticket routing works at a client. Sorting by hand is gone there.

    View AI ticket routing ↗
  • 04

    Content for social media

    From phone clips and a briefing, an AI pipeline cuts a reel for Instagram, with text and music in the account’s style. An n8n workflow takes every job and passes it on.

  • 05

    Workflows in n8n

    n8n connects inbox, database, calendar and business software through ready-made building blocks. We run it on our own server instead of someone else’s cloud. Every run is logged.

  • 06

    Connected to your systems

    The result goes where you work: into the POS, the ticket system, the database. If there’s no screen to review and approve, we build one too.

    Web apps & platforms ↗

02 / References

Four workflows we have automated.

From a shop to a tax office. Each one says what a model takes on and what fixed rules handle.

  • 01

    Inventory and POS from one email

    At Turkish Home the shop sends a supplier’s list by email. The system checks it against the stock, enters new items and prices into the Reacard POS and sends back a report with an audit log. AI is only needed for unknown list formats and for replies to follow-up questions, also in Turkish.

    Turkish Home inventory ↗
  • 02

    Instagram reels from phone clips

    For a social media team, a pipeline cuts a finished reel from up to ten clips and a briefing. n8n checks the job, and a model reviews the footage, cuts it and adds text and music. A few minutes later the video is ready. A change request produces the next version.

  • 03

    Workflows at a tax office

    For a tax advisory firm we are automating several workflows in day-to-day office work. Some need an AI model, others run on fixed rules. Because client data is involved, every workflow is built around the GDPR.

  • 04

    Tickets to the right department

    For a client under NDA we turned a language model into a classifier. It reads every new ticket and returns category, department and urgency. That saves several hours of sorting every week.

    View AI ticket routing ↗

03 / Tools

Automation with n8n, with our own code, or both.

We build many workflows in n8n. The tool comes from Berlin, and we run it on our own server. Where n8n reaches its limits, we write that part as a program of our own.

Is there an API?yesn8nInbox, CRM, calendarnoOur own programPOS in the browserOften both in one workflow
  • 01

    n8n is enough

    The workflow connects services that have an API: inbox, database, calendar, CRM. Rules and texts can be changed in the workflow without rebuilding the program behind it.

  • 02

    A program of our own

    A POS or a portal has to be operated in the browser, or there’s a lot of logic inside, such as stock that gets recalculated every day. That’s the case at Turkish Home.

  • 03

    Both together

    In the reel pipeline, n8n takes the job, checks it and writes the instructions for the model. The cutting happens in our program. The rules for the cut can be changed in n8n.

04 / With AI or without

AI only where a rule fails.

At Turkish Home, reading the lists, matching, the POS and the emails all run on fixed rules. A model joins in at only four points. We tell you what fits your case in the intro call.

Free textModelFixed formatCode checksPOSNo access to the POS or inbox
  • 01

    A fixed rule is enough

    If the invoice number is always in the subject line, you don’t need a model. A rule spots it and sends the email to accounting. It’s quicker to build and does the same thing every time.

  • 02

    People write freely

    A rule can still recognize a reply like “1 evet, 2 hayır”. If someone writes a whole sentence, a model understands it. The same goes for a list in a format the workflow doesn’t know yet.

  • 03

    The model delivers, the code acts

    With us, a model never writes into the POS or ticket system itself. It returns details in a fixed format, and our code checks them and carries them out. It has no access to the POS, shop or inbox.

  • 04

    A mistake would be expensive

    Where a wrong value costs money or a deadline, the workflow doesn’t decide alone. At Turkish Home every new item lands in a separate group in the POS, and the owner reviews it.

  • 05

    There are hardly any examples

    A classifier of your own learns from cases that have already been decided. For AI ticket routing, that was tickets from many years. If you only have a few, we start with rules and keep collecting.

  • 06

    You want answers, not a workflow

    If your staff should ask questions of manuals, contracts or minutes, that’s a knowledge assistant. It finds the right passage and names the source.

    AI knowledge assistant ↗

05 / Privacy

GDPR-compliant automation.

A workflow that reads emails sees names, addresses and sometimes client data. We settle where that data goes before we build.

  • 01

    Data processing agreement

    If we process your data, we sign an agreement under Art. 28 GDPR. It states who may do what with the data.

  • 02

    Our own server, not someone else’s cloud

    n8n and our programs run on servers we operate ourselves. If you want, the workflow runs on your own premises, like AI ticket routing. There, new tickets never leave the company.

    Hosting & email ↗
  • 03

    Only known senders

    An inbox that takes jobs only processes emails from registered senders whose origin has been verified. Everything else stays where it is and gets no reply.

  • 04

    Personal data out first

    If a model learns from old cases, we remove names and contact details before training. In production it only gets the permissions it needs for its task.

06 / What it costs

What automation costs depends on systems and data.

We bill by effort. The intro call and cost estimate cost €0.

Effort is driven mainly by how many systems get connected and whether one of them has no API. If a step needs a model of its own, the data adds to it: how many cases there are and how clean they are. If a rule is enough, the work on the model drops out.

Four questions, then you see budget and weeks.

What it’s about, how big, how demanding, how urgent. No obligation and no contact details. If you send an inquiry afterwards, you don’t type anything twice.

Get a cost estimate
Intro call and cost estimate
€0

Every payment comes after a sign-off. You can step out between any two milestones.

07 / How it works

From manual steps to a running workflow.

This is how we approach automation. Specification, sign-off and payment follow the same process as any project.

  1. 01

    Look at the workflow

    You show us how it runs today: which email comes in, who types up what, where it ends up. Add a few real examples, such as lists, tickets or documents.

  2. 02

    Rule or model

    For each step we work out whether a fixed rule is enough or a model is needed. Where a mistake would be expensive, we decide who confirms.

  3. 03

    Build

    We build the workflow in n8n, as a program of our own, or both. If a step needs a model of its own, we clean up the old cases first and measure over several rounds.

  4. 04

    Trial run

    First the workflow processes real emails but writes nothing into your systems. The reports come to us, so you can see what it would do.

  5. 05

    Go live

    It goes live when you decide. After that you get a report after every run, and every change is logged.

The full process ↗

08 / Questions

Questions about workflow automation.

What is workflow automation?
Workflow automation, often also called process automation, is a sequence of several steps that runs without manual work: an email comes in, the details are read, checked and entered into the next system, and at the end you get a report. At Turkish Home, one email with the supplier’s list is all it takes. The POS and the stock follow.
How is it different from AI automation?
In AI automation, a model takes over a step that used to require someone to read text: classifying a ticket, understanding a freely written reply, cutting a reel. Most steps in a workflow don’t need that. We only use a model where a rule fails.
What is n8n, and why do you use it?
n8n is a workflow automation tool from Berlin. It connects services through ready-made building blocks, logs every run and can be self-hosted. We run n8n on our own server, so the data doesn’t pass through someone else’s cloud. If n8n isn’t enough, we write that part as a program of our own.
Are you an n8n agency?
We’re a software studio from Duisburg and work for businesses across Germany. n8n is one of our tools, and we build many workflows in it. If it isn’t enough for a step, we write that step as a program of our own, as we did for the POS at Turkish Home.
n8n, Make or Zapier: which one fits us?
Make and Zapier only exist as the vendor’s cloud service. n8n can be self-hosted, and we run it on our own server. That matters as soon as customer or client data passes through the workflow. If a POS has to be operated in the browser, we write that part as a program of our own.
Do we need AI for this, or is a rule enough?
Often a rule is enough. If the information is always in the same place and in the same form, such as a number in the subject line, a rule spots it and is quicker to build. You need a model when people write freely and describe the same thing differently every time. We sort out what fits your case in the intro call.
Does this work if our software has no API?
Often, yes. The Reacard POS at Turkish Home has no public API. So our program logs in through the browser, enters items and prices and then reads back whether everything arrived.
Does the AI then decide on its own?
Only where you want it to. A misrouted ticket can be passed on, so the model can decide alone there. Where a mistake costs money or a deadline, it suggests, and a person confirms. We set that line with you before we build.
Is the automation GDPR-compliant?
We decide per project where the data is processed, and we sign a data processing agreement under Art. 28 GDPR. n8n runs on our own server, and if you want, the workflow runs on your own premises. With AI ticket routing, new tickets never leave the company, and personal data was removed before training.
How many examples does a model need?
For a classifier of your own, a lot. For AI ticket routing it was tickets from many years. For a task like understanding a freely written reply, a ready-made model is enough, with no training of its own. Whether you have enough, we can tell from an export before the quote is final.
How long does it take until a workflow is live?
That depends on the same things as the effort: how many systems get connected and whether one of them has no API. The order is fixed: look at the workflow, settle rule or model, build, trial run without write access. The workflow goes live when you decide. The calculator on the contact page gives you a range in weeks.
Who looks after the workflow once it’s live?
We do, the same team that built it. You get a report after every run, and every change is logged. If a list or a screen changes, we adjust the workflow. With a maintenance contract we’re reachable daily from 6 a.m. to 11 p.m. and get back to you within three hours.
What does automation with n8n cost?
We bill by effort. That is driven mainly by how many systems get connected and whether one of them has no API. If a step needs a model of its own, your data adds to it. The intro call and cost estimate cost €0. The calculator on the contact page gives you a first range in four questions.

09 / More

What else we build.

Tell us briefly about your project

The intro call costs nothing, and neither does the cost estimate after it. If what you need can be bought off the shelf, we tell you its name and send no invoice.

Start a project

We usually reply the same business day