A working layer above existing software.
FIDARYN replaces no programme. AI employees prepare work and hand cases to each other in a traceable way.

FIDARYN gives small and mid-sized businesses AI employees with a role, permissions and approval. On the software they already use. We talk to pre-seed investors who understand trades and SMEs.
Customer service, quoting, warehouse, purchasing, dispatch and accounting work on the same case, but in different programmes. At every handover, information is searched for, copied and explained again. That costs time, creates rework and slows growth, especially when skilled staff are scarce.
FIDARYN replaces no programme. AI employees prepare work and hand cases to each other in a traceable way.
The business decides what is only suggested, what needs approval and what may be done within fixed limits. Fixed code outside the language model enforces it.
Roles, rules and business knowledge live in FIDARYN, not in the model. Local model, German data centre or a later switch: the customer stays free to move.
Permissions, approvals, logging and tenant separation are part of the architecture from day one. That is exactly where AI fails in many businesses today: trust.
No migration project. That lowers the barrier for businesses with grown system landscapes.
Every role, connection and rollout can be reused at the next business. That is where the lead grows, not from a single model.
For businesses where the owner also runs dispatch and the master craftsman writes quotes in the evening. Language, processes and approvals fit them.
This section only lists what is built and automatically tested. Everything else is listed below as the next stage.
Fetching from the mailbox, classifying the enquiry, one case across several messages, reply draft, handover to another role and follow-up question. In the demo it runs with a real mailbox and a real model.
Nothing goes out without human approval. The approval is bound to exactly this draft and runs only once. Every step is logged: who, what, when, with which decision.
Hard tenant separation in the database, central permission checks, rule checks outside the AI, credentials in a vault without AI access, an isolated mail connector and sign-in with a second factor.
Every change runs automatic security and functional tests. Software components are checked daily for known vulnerabilities.
Learning from reviewed corrections, company memory with sources, connections to accounting and trade software, and sign-in via the FIDARYN app. We build them together with the first founding businesses.
Registered in Germany. Source: ZDH, key figures of the German trades, data 2025.
In German manufacturing. Source: Destatis, business register, reporting year 2024, own sum of two size classes.
Another 9 % intend to. Source: Bitkom, survey of 504 trade businesses, published August 2025.
The model: subscription with a base fee, included AI roles and a usage allowance. More roles, systems and locations extend the contract. Onboarding is billed separately. We validate prices with the founding businesses.
Models can now operate tools and multi-step processes. At the same time, skills shortages, system sprawl, mandatory e-invoicing and the wish for data sovereignty keep growing. SMEs need AI they can trust.
The figures describe the environment, not a target customer list, and are not added up.
Kai Jansen combines his own workshop and business experience with building his own software and agent systems. He knows the handovers FIDARYN takes over from his own business.
We discuss documents, product status and a live demo in person.
This page is not an offer to acquire shares. Go to the investor login
Early access for selected businesses: FIDARYN is starting with a pilot phase for founding businesses. We show you live in the demo what your business can use today.