# AiAdvisors

> production AI systems (PL: produkcyjne systemy AI)

AiAdvisors is the AI engineering practice of Romuald Członkowski — creator
and maintainer of n8n-mcp, the leading MCP server for n8n: an open-source
tool that connects AI agents to the n8n automation platform and gives them
a full service layer (documentation lookup, configuration validation,
workflow operations) so agents build working workflows instead of guessing
(22,600+ GitHub stars, 128,000+ users, used by teams at Deutsche Telekom,
PayPal, Mercado Libre, MIT and NYU). Founder of AiAdvisors with multiple
production implementations. He has also
worked as an advisor on World Bank projects. The practice builds production AI
systems: agentic process automation, document intelligence and RAG,
agentically developed applications (Claude Code + MCP under architectural
control), MCP servers and model fine-tuning. The site is bilingual:
Polish at the root (canonical), English under /en.

## Core pages

- [Strona główna (PL)](https://aiadvisors.pl/): Hero, problem framing, services ("what I build"), case studies preview, about, FAQ, booking CTA.
- [Home (EN)](https://aiadvisors.pl/en): English version of the landing page.
- [Case studies (PL)](https://aiadvisors.pl/case-studies): Index of client work.
- [Case studies (EN)](https://aiadvisors.pl/en/case-studies): English index of client work.
- [Services (PL)](https://aiadvisors.pl/services): Index of the four service areas.
- [Services (EN)](https://aiadvisors.pl/en/services): English index of the four service areas.

## Case studies

- [Multi-marketplace e-commerce](https://aiadvisors.pl/en/case-studies/ecommerce-operations): An automation platform for a multi-marketplace e-commerce business — from pricing and product data across 7+ marketplaces, through agentic research, to training an internal team that now builds its own automations. (PL: https://aiadvisors.pl/case-studies/ecommerce-operations)
- [Thedy & Partners](https://aiadvisors.pl/en/case-studies/tax-advisor-assistant): A tax advisor's assistant — a self-updating base of 538,000+ tax interpretations with hybrid search and a research agent. A preliminary opinion in ~2 minutes instead of 12 hours. (PL: https://aiadvisors.pl/case-studies/tax-advisor-assistant)
- [CFOpro](https://aiadvisors.pl/en/case-studies/cfopro): Two AI systems for a CFO-as-a-Service firm: an agentic content engine that researches and writes articles on its own — with an editorial gate before publishing — and a Coda-based finance operating system that replaced scattered spreadsheets, with automated invoice intake (KSeF, email, receipts) and a hub for staying on top of payments. (PL: https://aiadvisors.pl/case-studies/cfopro)
- [n8n-mcp](https://aiadvisors.pl/en/case-studies/n8n-mcp): An open-source tool that connects AI agents to the n8n automation platform and lets them build automations that actually run — 22,600+ GitHub stars, 128,000+ users, 1.1 million+ workflows built. Used by engineering teams at Deutsche Telekom, PayPal, Mercado Libre, MIT and NYU, among others. (PL: https://aiadvisors.pl/case-studies/n8n-mcp)
- [Energy House](https://aiadvisors.pl/en/case-studies/invoice-automation): An automated invoice system for a busy energy business — cost invoices that used to scatter across seven team inboxes and get gathered by hand every month now collect themselves into one organised archive and one clean register, ready to send to the accountant. (PL: https://aiadvisors.pl/case-studies/invoice-automation)
- [World Bank — Romania](https://aiadvisors.pl/en/case-studies/world-bank-romania): A two-part data engagement for the World Bank in Romania's energy sector — turning ~100,000 building energy certificates (dozens of gigabytes of photos, scans and spreadsheets) into one structured database, and turning the national census into accurate maps of how the country heats its homes. (PL: https://aiadvisors.pl/case-studies/world-bank-romania)
- [Real-estate investor (USA)](https://aiadvisors.pl/en/case-studies/us-real-estate): An automated deal-flow pipeline for a US multifamily investor — AI reads broker offerings from emails and flyers, matches properties against the database, enriches the data and prepares Letters of Intent. (PL: https://aiadvisors.pl/case-studies/us-real-estate)
- [Financial due-diligence consultancy](https://aiadvisors.pl/en/case-studies/interview-insights): An AI pipeline for a Nordic financial due-diligence consultancy that turns due-diligence interview transcripts into structured findings — answering every question on the DD agenda, checking each answer against the transcript, and surfacing what management revealed without being asked. Built privacy-first, so confidential deal data never leaves the firm's own cloud. (PL: https://aiadvisors.pl/case-studies/interview-insights)
- [Mental-health app for therapists](https://aiadvisors.pl/en/case-studies/mental-health-app): A REST backend for a therapists' mental-health app. What began as a migration from a Telegram bot to a Flutter app grew into a full clinical platform — 40+ workflows and 30+ endpoints in production, with a therapeutic AI core, RAG per therapy method, multimodal input and case conceptualizations. (PL: https://aiadvisors.pl/case-studies/mental-health-app)
- [Saudi Ministry of Industry](https://aiadvisors.pl/en/case-studies/saudi-ministry-ai-use-cases): An AI strategy for the Ministry of Industry's Seneai platform (Saudi Vision 2030) — around 60 potential use cases assessed on impact and feasibility, distilled to 7 prioritised initiatives and a phased roadmap, plus a working proof of concept for AI-assisted factory inspections. (PL: https://aiadvisors.pl/case-studies/saudi-ministry-ai-use-cases)
- [AdBridge](https://aiadvisors.pl/en/case-studies/adbridge): A creative-ops pipeline for a French agency — from raw images and videos the system generates ad copy in French, standardises creative naming and, after approval, publishes the ads to Meta Ads by itself. (PL: https://aiadvisors.pl/case-studies/adbridge)
- [Custom-trained n8n model](https://aiadvisors.pl/en/case-studies/n8n-workflow-model): Compact models I fine-tuned myself that turn a plain-language request into a working n8n workflow. On the task they match frontier models like Claude Opus — while running self-hosted on a single GPU, for a fraction of the API cost. (PL: https://aiadvisors.pl/case-studies/n8n-workflow-model)
- [World Bank — Clean Air Programme](https://aiadvisors.pl/en/case-studies/world-bank-capp): IT and AI architecture advisory for Poland's Clean Air Programme (CAPP) — one of Europe's largest thermal-modernisation programmes. Two advisory reports with recommendations and a phased AI implementation roadmap. (PL: https://aiadvisors.pl/case-studies/world-bank-capp)
- [Aurora — 3D in the browser](https://aiadvisors.pl/en/case-studies/aurora-3d): A research project testing how far the agentic way of working carries when I step into a field I know nothing about. I had never done any 3D work — three days later an interactive colony ship was running in the browser. What lasted isn't the ship; it's the process, written down and reused since. (PL: https://aiadvisors.pl/case-studies/aurora-3d)

## Services

### AI consulting & strategy

[AI Consulting & Advisory Services in Poland](https://aiadvisors.pl/en/services/ai-consulting) (PL: https://aiadvisors.pl/services/ai-consulting)

I help companies decide where AI actually pays off — and then build it. As an independent AI consultant based in Poland, I don't resell licences or anyone else's platform, so the recommendation is the only thing I have to sell.

- Process audit and AI opportunity map — I go through your operational processes and identify the ones where AI makes economic sense. I flag the ones where it doesn't just as explicitly — usually the most valuable part of the conversation.
- AI strategy and prioritisation — I turn a long list of ideas into a short list of decisions: what we do first, what waits, what we don't do at all. Every priority carries a cost, a risk and a way to tell whether it worked.
- Workshop for leadership and teams — A half-day or full-day session after which the team understands what AI can do today, what it cannot, and where the boundary sits inside your process. No marketing, no scare stories.
- First use case, delivered — I take the chosen priority to a working production system. That is the moment advisory stops being theory — and the only honest way to verify a recommendation.
- Architectural oversight — If you already have a team, I come in as an architect: reviewing the design, model choices, cost and security, and staying available for the decisions that come up.

### Process automation

[Process Automation Consulting: n8n, iPaaS and AI Agents](https://aiadvisors.pl/en/services/process-automation) (PL: https://aiadvisors.pl/services/process-automation)

I build business process automation on n8n — with AI agents where a judgement call is needed, and plain logic where a rule is enough. Where low-code runs out, I write the microservices that take over.

- Quoting and inbound requests — From an enquiry in the inbox to a finished quote: reading the requirements, matching them to your pricing and catalogue, generating the document and handing it to a salesperson to approve.
- Invoicing and back-office processes — Many suppliers, many inboxes, many formats — reduced to one register ready for accounting, with duplicate and exception handling built in.
- System-to-system integration (iPaaS) — CRM, ERP, e-commerce, spreadsheets, supplier APIs. I connect systems that were never designed to talk to each other, and keep those connections working when the other side changes.
- Operational agents — Where a process needs judgement — qualifying a lead, classifying a ticket, choosing a path — I place an agent with a clearly bounded scope and an explicit point where it hands the case to a human.
- Microservices where low-code ends — Custom APIs, scrapers, document converters and compute services — written when no ready-made block exists, or when one would become the bottleneck.
- Training your automation team — I teach your team to build their own n8n automations and stay on regular office hours to review their work and unblock the harder cases.

### Conversational AI

[Conversational AI Consulting in Poland](https://aiadvisors.pl/en/services/conversational-ai) (PL: https://aiadvisors.pl/services/conversational-ai)

I build AI assistants that answer from your own knowledge — documents, procedures, case history — rather than from whatever the model invents. And systems that analyse the conversations nobody has the hours to review.

- Assistants grounded in company knowledge — A system that answers from your documents, procedures and case history — with the source shown alongside every answer.
- Support for expert teams — An assistant that shortens a specialist's research: finding the right regulation, interpretation or prior case and drafting an answer for review.
- Customer-support automation — Classifying and triaging tickets, drafting replies for an agent, and handing the case to a human at the point where the system stops being confident.
- Conversation and interview analysis — Transcription and analysis of recordings: findings, recurring themes, risk signals. Available in a fully private variant where recordings never leave the client's infrastructure.
- Integration with the channels you use — Slack, Teams, email, an internal panel, telephony and CCaaS platforms — the assistant shows up where people already work instead of becoming another window.

### AI engineering

[AI Engineering: MCP Servers, Fine-Tuning and RAG](https://aiadvisors.pl/en/services/ai-engineering) (PL: https://aiadvisors.pl/services/ai-engineering)

The technical layer the other services stand on: MCP servers, model selection and fine-tuning, document processing at scale, and software built agentically — under the control of architecture, specifications and tests.

- MCP servers — I design and build MCP servers that give AI agents controlled access to your systems — with validation, permissions, and verification of the agent's work before it goes any further.
- Model selection and fine-tuning — I match the model to the task and, where it pays off, fine-tune a custom one. The aim is usually the same: frontier-level quality at a fraction of the cost, without sending data outside.
- Document intelligence and RAG — Extraction and structuring of data from documents — PDFs, scans, mixed formats — and semantic search over very large collections. Scale measured in hundreds of thousands of documents.
- Agentic software development — Applications and backends built agentically with Claude Code and MCP, always under the control of architecture, specifications and tests. The agent writes the code; whether the system is built correctly is decided by oversight.
- Architecture, cost and security — A review of an existing solution for call costs, data leakage, single-vendor lock-in, and what happens at ten times the traffic.

Also offered across all four areas: internal automation-team training
("citizen automation") — employees build their own automations with
n8n-mcp, with regular office hours for review and unblocking.

## Questions answered on the service pages

**How is your AI advisory different from a traditional consultancy?**
A traditional consultancy stops at the recommendation. Here the recommendation is an intermediate product — I verify it by building. That keeps me from proposing things that cannot be implemented, because I would hit the wall myself a week later.

**Do you work with clients outside Poland?**
Yes. I run projects in English and Polish, remotely. Clients have included companies in France, the US and Romania, a ministry in Saudi Arabia, and World Bank programmes.

**How long does the analysis take before implementation starts?**
Usually from a few days to two weeks, depending on how many processes are in scope and how accessible the data is. The workshop itself is a single day.

**What if it turns out AI isn't needed here?**
Then I say so. Plenty of processes are better fixed with ordinary automation or an organisational change — cheaper and more reliable. I have no incentive to sell a model where a rule will do.

**Why n8n rather than Make or Zapier?**
n8n can be self-hosted, so data never leaves your infrastructure, and there is no per-operation limit to wreck the economics at volume. It also lets you drop into real code where the ready-made blocks stop. If Make is genuinely enough for someone, I say so.

**Does automation have to involve AI?**
No, and often it shouldn't. A model is needed where something has to be understood or judged. Where a rule suffices, the rule is cheaper, faster and predictable.

**What happens when an automation breaks?**
Every flow has error handling and alerting — a failure is loud, not silent. Business-critical processes also keep a manual path so a breakage doesn't stop the work.

**Can our team handle this without you?**
That is the point. The training and office hours exist so the company doesn't end up dependent on me. At several clients the team now builds most of the automations themselves.

**How do you stop the assistant inventing answers?**
Answers are built from specific passages in your documents and shown with the source. The system is also explicitly set up to say “I don't know” when there is no basis rather than improvise — and that behaviour is measured in evaluation before launch.

**Will our documents go to an external model?**
That depends on your requirements. We can use commercial models under agreements that exclude training on your data, or run entirely locally on models hosted in your own infrastructure. For sensitive data I propose the local option by default.

**Will this replace our support team?**
In practice it takes over the repetitive part and prepares the groundwork for harder cases. The team stays — it just stops answering the same question for the thousandth time.

**What about Polish?**
I build and test these systems in Polish. Polish is harder for semantic search than English, so model choice and document preparation matter more here — and that is part of the work.

**What is MCP, and is it just another trend?**
MCP is the standard that lets an AI agent use external systems in a controlled way, instead of hard-wiring a separate integration into every tool. I built one of the most widely used MCP servers, so I speak about it from the maintenance side rather than the announcement side.

**When does fine-tuning make sense, and when is a good prompt enough?**
Fine-tuning pays off on a narrow, repetitive, high-volume task — it cuts cost and lets you run the model yourself. For variable tasks or small scale, a good prompt and the right model choice win; tuning would only make changes harder.

**Can this run inside our own infrastructure?**
Yes. Some clients require that data never leaves their environment, in which case I deploy locally or into your cloud. It affects model choice and cost, so we settle it at the start rather than the end.

**Code written by agents — how is quality controlled?**
The same way as with a team of people, only faster: specification before implementation, tests, code review and multi-agent CI. Responsibility for the architecture, and for what reaches production, stays with me.

## Booking & contact

Free strategic consultations are 30 minutes, booked online.

- [Book a consultation](https://cal.com/czlonkowski/30min)
- Email: romuald@aiadvisors.pl

## Profile

- LinkedIn: https://linkedin.com/in/czlonkowski
- YouTube ("AiAdvisors | Romuald Członkowski"): https://www.youtube.com/@czlonkowski