Work

Built, not promised.

This is what I have implemented in real companies. I name clients when they have agreed to it. Otherwise I describe the industry and the task.

Florian Fürst speaking at Berlin AI Day 2026 in front of a screen showing the fuerst.ai logo

Case study · 1NCE, IoT connectivity, Cologne · since 2026

A team of AI agents for first-line support

  1. Requestby email or customer portal, in several languages
  2. Triage agentrecognises topic and language and passes the request on
  3. Six specialist agentsdraw on the knowledge base and all systems of record and write the reply
  4. The AI decidesand checks how confident it isConfident: the reply goes out.Unsure: a person checks it first.
  5. Second levelreceives, with all the information, whatever the first stage cannot solve

Starting point

1NCE connects devices worldwide over the mobile network. Requests come in by email and through the customer portal, in several languages. The support team answered many of them again and again in similar ways. The systems were well documented and had interfaces: ideal conditions for achieving a lot quickly.

What I built

As the n8n expert in 1NCE’s AI Transformation Squad, I built a system of AI agents. The diagram shows how it works.

What it achieves

  • A fraction of the former 32 hours is how long requests now wait in first-line support.
  • €2 million a year, almost: what 1NCE is on track to save on an external service provider.

Figures from 1NCE, as of August 2026. This is how 1NCE sums up the six months:

Florian, thank you for an outstanding six months. Fully automating our first-line support […] was a huge step for 1NCE. It also showed everyone what AI-supported workflow automation can do for a business at real scale. It was a pleasure working with you!

Stephan Klink, 1NCE

More projects

Other industries, same approach.

Commodity trading, Switzerland: pricing

A small commodity trader deals in raw materials internationally, between buyers and sellers. When a bid comes in, the seller needs a price fast: with transport costs, margin, and the question of which carrier serves the route at all. That was calculated by hand in Excel every time. I built a portal that does this calculation automatically. Faster, and less prone to errors. 2024 and 2025, with Softr and make.com.

Home page of the pricing portal: latest bids from buyers with product, bid, destination and Incoterms, buyer names masked
Screenshot: the portal with test data, built with Softr and make.com

Export control: AI search in goods lists

For export control law specialists, I showed in a workshop and a prototype how an AI search goes through goods lists and cites the relevant passages. Hosted in Germany, without US providers.

Goethe-Institut: film bookings

Around 2,000 orders sat in an Excel spreadsheet. They became a data model, a portal and automated processes connected to a film database.

  • 322 films
  • 174 institutes
  • 2,035 orders since January 2024
Home page of the booking portal: key figures on films, institutes and orders, the most ordered films, upcoming orders with their status, and the film catalogue
Screenshot: the portal, built with Softr. First place, Softr Build Week 2025

la marchante GmbH: process portal

A small team, and the processes were written down nowhere. Everyone knew what they did. If someone was away, that knowledge was missing. For la marchante, the company behind the TeeFee drinks brand, I built a portal in which the team captures all its processes itself, in one place. That keeps the knowledge in the company and is also the basis of my analysis: I assessed the processes, reviewed the tools in use and created a roadmap for automation, including training for the team. Next: AI agents for the cash flow forecast and marketing.

  • 72 objects
  • 34 tools
  • 107 business processes
Home page of la marchante’s process portal: tiles for capturing objects, software, people, teams and business processes, below them the count so far: 72 objects, 34 software, 4 people, 21 teams and functions, 107 business processes
Screenshot: the process portal, built with Softr

Prizes and certificates

Awards and partners

  • 2nd place Flow4Gold at n8n’s In The Loop conference 2026 in Berlin, in a final of around 15 participants. My entry: three AI agents that follow returns until the refund arrives. They read receipts, check delivery and send reminders about outstanding refunds, but only after I approve.
  • One of three winners of the n8n #FrAIday Challenge in April 2026 with a workflow for my own family: an AI agent reads the emails from school and after-school care, attachments included, spots dates and tasks, and puts them in my calendar and to-do list
  • Adacor partner
  • 1st place Softr Build Week 2025 with a film booking portal for the Goethe-Institut
  • Softr Certified Partner
  • Certified AI Business Advisor since 2019
  • Ten years of digital projects at Frankfurt School of Finance & Management, with budgets of up to €350,000
Florian Fürst opens his prize at the Flow4Gold show at n8n’s In The Loop conference 2026
Softr’s winner graphic: 1st place Softr Build Week 2025 for Florian Fürst

First call

What would your project be?

Tell me in 30 minutes where things get stuck.

Book a first call

Florian Fürst talking to the audience at Berlin AI Day 2026
First call