Process automation on n8n: integrations, documents and AI agents.
I build and maintain automations on n8n: system-to-system integrations, document handling, AI agents where a judgement call is needed. The workflows are built by AI agents I equipped myself through n8n-mcp. I am accountable for them as for my own code, and I teach your team to do the same.
- 5.0 / 5official n8n Expert Partner · reviews in the official n8n expert directory
- 190,000+users of n8n-mcp, the tool AI agents use to build n8n workflows
- weeklydata on how AI agents build n8n: the n8n AI Automation Index
Automation in most companies stops at the same place: the simple flows work, and the costly ones need a judgement call, an exception or a document no form can handle. My work starts at that line.
I am not an agency. I built n8n-mcp, the tool that lets AI agents build working n8n workflows, and more than 190,000 people use it today. I work the same way for clients: agents build, I design, review and own the result. One person delivers the scope an agency needs a team for, and the knowledge of how it works stays inside your company.
What I do
System-to-system integrations
CRM, ERP, e-commerce, marketplaces, spreadsheets, supplier APIs and systems with no public API. I connect tools that were never designed to talk to each other and keep those connections working when the other side changes.
AI agents inside the process
Where the process needs judgement: lead qualification, ticket classification, routing, reading a document. The agent gets a clear decision scope and a point where it hands over to a person.
Documents and back office
Quotes, invoices, orders, tickets: many suppliers, many inboxes, many formats reduced to one register with duplicate and exception control.
Microservices where low-code runs out
Own APIs, scrapers, document converters, apps with a front end and n8n as the backend. Written when the ready-made block does not exist or would become the bottleneck.
First stage: one process in production
Instead of a proposal covering ten processes, one fixed-price stage. Afterwards we both know whether we want to continue, and you have a first process running in production.
- An n8n instance ready for the team: on your server, or taken over and fixed if one already runs
- One trial integration chosen together and brought to production
- Error handling and alerts from the first workflow
- A recommendation of what to automate next, and in what order
Who it is for
- Companies where several people spend a measurable part of their week re-typing data between systems.
- Teams that already run n8n or another iPaaS and have hit its ceiling.
- Organisations whose automation has to handle documents, not only structured data.
- Companies that want the team to build on its own instead of depending on a vendor.
How I run it
- 01Process mapping
I sit with the people who run the process today and write it down step by step, including the exceptions nobody remembers until they happen.
- 02Agents build, I am accountable
Workflows are built with AI agents working through n8n-mcp, with every change validated. I design the architecture, review the result and take responsibility for what reaches production.
- 03First flow in production
I start with one well-chosen process and take it all the way. A working flow in production is a better argument than a plan covering ten.
- 04Extending and handing over
Further processes are added on the same infrastructure, with the same monitoring and error handling. The team gets documentation, access and weekly sessions where it learns to build on its own.
Implementations
From single flows to platforms where the client now runs more than two hundred automations with its own team.
Over 400 workflows on one platform — marketplace automation across 6 countries, agentic product research, custom microservices and an internal automation team trained to work independently.
Read the case studyFor a CFO-as-a-Service firm: an agentic content engine publishing one article a day, and a Coda-based finance operating system — a counterparty registry, an income-and-expense ledger fed straight from invoices, and a hub that shows who needs paying and how much.
Read the case studyCost invoices arriving across seven team inboxes are found, read, de-duplicated and filed automatically — into a Google Drive archive and a Google Sheets register that's ready to hand to the accountant. Twelve connected n8n automations.
Read the case studyAI reads broker offerings from emails and flyers (vision), fuzzy-matches them against a property database, enriches them with neighbourhood data and generates Letters of Intent in one click.
Read the case studyFrom a file in a folder to a live ad — AI analysis of images and videos, French copy in three positioned variants, standardised naming and automated upload to Meta Ads.
Read the case studyThe instance and the team
Automations run on an n8n instance and a team keeps them going. Two separate services cover that.
Let's talk about the process you want to automate
A 30-minute online call, no commitment. Tell me how the process runs today and between which systems people copy data by hand, and I will tell you which workflow I would start with. After the call you get a proposed scope and a quote.
Prefer to write? romuald@aiadvisors.pl
Guides
How the n8n AI Agent works, when to pick n8n over Make or Zapier, and how to build workflows you can maintain.
What is the n8n AI Agent and how does it work?
The n8n AI Agent is a node in which a language model receives instructions (a system message) and a set of tools, then decides on its own which tools to call to complete a task. Without tools it is just a chat. With tools, most often sub-workflows, the agent can read a document, check data in a spreadsheet and make a judgement call no plain rule can handle. Since August 2026 n8n also has a separate Agents category: agents with their own memory, channels and skills that live outside a single workflow.
Read the guiden8n vs Zapier vs Make: which should a company choose for automation?
Zapier counts tasks per successful step, Make counts operations per module, n8n counts executions regardless of how many steps a workflow has. That billing unit matters more to the bill than the integration list on the homepage. Zapier and Make win on onboarding speed and the number of ready integrations, n8n wins on volume, data control, and code where the ready-made blocks run out.
Read the guideWhere do you find n8n workflow templates, and how do you use them with an AI agent?
n8n.io publishes more than 12,000 workflow templates, and n8n-mcp gives AI agents access to 2,352 of them through search_templates, get_template and n8n_deploy_template. Telemetry shows agents reach for a template rarely, read documentation and build from scratch far more often, then revise the result about nine times. This guide covers when a template genuinely saves time and what you have to check before you activate one on your own instance.
Read the guideWhat the data says about AI-built workflows
Every week I publish figures from n8n-mcp telemetry. Three of them explain why workflows built by agents need review, error handling and a verified setup before they reach production.
Frequently asked questions
- How much does automating a process cost?
- The first stage costs €1,000 to €2,000 and ends with one process running in production, with error handling and alerts. Further processes are priced stage by stage, based on what already runs. If you do not have n8n yet, setting up the instance is part of the same stage.
- Why n8n rather than Make or Zapier?
- n8n can be self-hosted, so data never leaves the company's infrastructure, and there is no cap on operations that wrecks the economics at higher volumes. It also lets you write your own code where the ready-made blocks end. If Make is enough for you, I will say so.
- Why not an agency?
- Because you get what an agency delivers without its overhead and without diluted accountability. The workflows are built by AI agents I equipped myself through n8n-mcp, and every change is checked by validation and by me. One person knows the whole system from the server to the agent prompt, and by the end of the engagement so does your team.
- Does automation have to use AI?
- No, and often it should not. A model is needed where something has to be understood or judged. Where a rule is enough, a rule is cheaper, faster and predictable.
- What happens when an automation breaks?
- Every flow handles errors on three levels: retries for transient failures of external services, a separate path for expected errors, and an error workflow with a chat alert for the unexpected ones. Failure is loud, not silent. Critical processes also keep a manual fallback.



