When off-the-shelf software isn't enough, we build AI systems for your company, from prototype to production use. We also run our own AI products, and that experience flows into every client project.
Standard tools such as Copilot or ChatGPT cover many tasks. They reach their limits when a task combines your own data, several systems and fixed checks. At a timber group, Copilot read a product table as running text and produced wrong results. It only worked with a clean data structure and a dedicated integration.
For cases like this we develop custom AI solutions and AI agents for companies, from prototype to production system. At the core we are a team of two, with specialists from a select network where needed. So far we have delivered 32 generative AI projects with agents.
Assistants that make specialist knowledge quickly and reliably accessible. Examples are a legal-information tool for the in-house lawyers of a social-services organisation and a product-data assistant for a sales team that had to find the right specification in hundreds of data sheets.
Systems that read, check and process documents, such as construction plans, funding requests or data sheets. The result lands, structured, in the system your team already works in.
An AI agent handles a task in several steps and uses tools such as CRM, search or databases to do so. In a multi-agent system, several agents take on separate roles. Agents need clear roles, not just prompts.
AI products your customers use directly. For a Dutch HealthTech startup we redesigned a recovery app and built the agent orchestration in the backend. On top comes the experience from our own products such as Aderato, a pipeline for ad creatives.
We clarify the workflow, the data sources and the systems the solution connects to. This produces an architecture that stays maintainable months later.
A first working version your team uses to test on real cases whether the solution hits the problem. That shows early on whether further development pays off.
The prototype becomes a system for daily use: connected to CRM, ERP or your Microsoft environment, with error handling and documentation.
We introduce the solution with your team, train users and adjust it based on feedback until the team works with it independently.
Custom AI solutions run within your Microsoft ecosystem and your company environment. No data leaves for the outside, processing takes place in the EU, and GDPR data protection is an architecture decision, not an afterthought. The goal is that your team uses the solution independently and ownership stays in-house.
For an international timber group we built an agent system with CRM integration. It detects sales signals in construction plans and turns them into a quote, in minutes instead of weeks. The real hurdle was not the AI but the data foundation, which we moved into a Dataverse structure.
View projectArchitecture and development are led by Nikolay Krastev. He studied Artificial Intelligence at JKU Linz; his focus is agentic systems, software architecture and custom builds. Simon Micheler keeps the link to the business: what the solution has to do, who uses it and how it is introduced.
If a project needs more specialisation, we bring in profiles from our network and name them in advance.
An AI agent handles a task in several steps and uses tools by itself, for example it reads documents, looks things up in the CRM and prepares a result. Unlike a chatbot, it does not wait for every single instruction.
Yes, if the workflow is clear. We start with a single use case with high value and build a prototype first. That way you see whether the agent pays off before investing more.
Several AI agents with separate roles work on one task together, for example one researches and another reviews critically. What matters is that every role is clearly defined. The orchestration pattern determines quality and maintainability.
When standard tools do not process your data correctly, several systems need to be connected, or the solution becomes part of your product. Whether that applies is something we clarify in the discovery phase, before anything is built.
Within your Microsoft ecosystem and company environment, with processing in the EU. No data leaves for the outside.