At around 14, I started building my first websites in my bedroom.
Curiosity quickly turned into projects of my own: forums, communities, online shops, landing pages and websites for businesses. I rented dedicated servers, virtualised them and ran my own services on them. For outside companies I built complete web solutions – among them for restaurants, with menus, galleries, reservations and their own admin functions.
AI backgroundKai JansenFounder of FIDARYN
MILESTONES
The road to FIDARYN
14
First websites & online projects
Forums, communities, shops, landing pages and company websites.
Infrastructure
Servers, virtualisation & own services
Dedicated servers, VPS setups and self-run web services.
Engineering
Apprenticeship & technical responsibility
Early hands-on responsibility for complex technical installations.
01
Technology was never just a hobby for me.
Even back then I wanted to know how systems are built, why processes work – and where they can be made better.
Alongside this, my professional path started out in classic engineering.
During my apprenticeship I also worked at a car wash, and in time I knew its technology so well that I took over technical responsibility there once I had qualified.
Next came positions at a major energy utility, then as a field service technician for car wash systems and later as technical lead of a state-of-the-art facility.
By then it was long since about more than repairing things.
It was about keeping systems available, understanding faults quickly, improving technical processes and solving problems before they turned into downtime.
Energy utility
Industrial technical environment
Further technical experience beyond classic automotive structures.
Field service
Technician for car wash systems
Fault analysis, repair, system availability and field service.
Technical lead
Responsible for a modern car wash
Technical processes, availability and optimisation.
Entrepreneur
Own automotive workshop
Several years self-employed, focused on demanding vehicles and motorsport.
02
Meanwhile, my own business took shape.
Alongside my job I was already running an automotive workshop of my own.
What began as a side business grew into a second full-time job. In the end I went fully self-employed and ran my own workshop for several years.
The focus was not everyday servicing but increasingly demanding vehicles, special conversions and motorsport.
As an entrepreneur I learned something else, though: that the actual skilled work is often only part of the problem.
The other part is quotes, invoices, customer communication, suppliers, appointments, documentation, staff, paperwork, follow-up questions and information that gets lost somewhere between different systems.
That very experience would later prove decisive.
Digitalisation
Field management software
Paper-based processes digitised and partly automated.
03
The next step was digitalisation.
After my time at the workshop I moved back into a more technical environment and today work in critical electrical infrastructure.
There, too, I met a problem I already knew from other companies: many processes worked – but they were needlessly manual.
Paper-based time tracking.
Scattered information.
Duplicate entries.
Processes that depend on individual people.
So I started building a solution there myself as well.
The result was a field management software that I developed for my employer and which digitised and partly automated core processes – among them working times, hours, operational processes and information from field service.
Paperwork became a digital system. And that is exactly where artificial intelligence became truly interesting to me.
I did not want to know what AI can do in a demo. I wanted to know what it can achieve in a real business.
AI & automation
Own agent and software architectures
Experiments and productive use of various AI models, APIs and local systems.
16AI agents
Autonomous online retail
31 digital employees in seven departments, 16 of them AI agents with their own role, plus rule-based checkers and technical workers.
04
I wanted to know how far it can really go.
Not in a demo.
Not with a chatbot that writes text.
But in real business processes.
I began experimenting with different AI models, local systems, APIs, agents, automations and software architectures of my own. Individual experiments grew into ever larger systems.
And at some point an online store came into being that I built consistently around one question:
How far can a business run operationally when AI does not just assist, but takes on real tasks and responsibility within clear limits?
The result is an autonomously operating e-commerce system.
Today it runs 31 digital employees in seven departments – 16 AI agents with their own role, plus rule-based checkers and technical workers. Supplier orders only run with approval.
digital employees
31
departments
7
AI agents with their own role
16
Among other things, they handle:
Product maintenance
Supplier data
Prices and stock
Order processes
Supplier orders
Marketing
Customer service
Control and quality assurance
Every agent has a defined role and only the permissions it needs for its task.
Several agents can belong to one department. Within these departments, supervising agents take on additional control functions and check results before processes continue.
What matters most: these limits are not just part of a prompt. Roles, permissions, allowed actions and system access are enforced technically.
An agent can only carry out the actions for which tools and permissions have been technically granted to it.
This gives rise to a fundamental principle:
AI does not simply get access to a business. It gets a concrete task within a technically controlled scope of action.
FIDARYN
The sum of these experiences
Digital employees for real businesses – integrated, controllable and role-based.
05
That is exactly how FIDARYN came about.
While I was developing these systems, I realised that most businesses do not have an AI problem at all. They have a complexity problem.
Even a small business today works with several systems at once:
Email
Accounting
Tax advisor
CRM
Files
Calendar
ERP
Online shop
Industry software
And every additional system creates new interfaces, new logins and new information that has to be brought together somewhere.
At the same time, an ordinary business owner is suddenly expected to deal with LLMs, models, prompts, context windows, agents and APIs. That cannot be the answer.
A business owner should not have to configure an AI stack. They should be able to say: “I need someone for this task.”
And that is exactly how FIDARYN is meant to work. A business creates a digital employee. It defines:
its role
its tasks
its permissions
its systems
its area of responsibility
In the background, FIDARYN takes care of which model, which tools and which technical components are needed for it.
For the business owner, it does not feel like configuring an AI system. It feels like hiring an employee.
06
Control is not an add-on feature.
Many businesses are rightly cautious about AI. Not necessarily because AI is fundamentally uncontrollable, but because generative models work probabilistically and can produce errors.
That is why FIDARYN builds control not only into instructions, but into the architecture.
An AI employee only gets access to explicitly approved systems and actions. Roles, permissions, tool access and organisational boundaries are enforced technically.
So an employee can, for example, process invoices without automatically having access to personnel files. Or answer customer enquiries without being allowed to export company data.
Least privilege instead of full access.Defined processes instead of uncontrolled autonomy.
07
I am not building FIDARYN because AI is a trend.
I am building it because for years I kept seeing the same problem.
As a technician.
As a technical lead.
As an entrepreneur.
As a software developer.
And finally while building autonomous AI systems.
Businesses already have the people, data and tools they need. What is often missing is an intelligent layer that connects it all.
FIDARYN is meant to become exactly that layer.
And before I sell it to a business, it has to work where mistakes have real consequences: in real operation.
— Kai JansenFounder of FIDARYN
FIDARYN is not built for presentations. It has to work in real operation.
Let us talk about your business.
In development: past the idea, not yet ready for market.
