GenAI has already started adding a lot of value to many enterprises. But much more is yet to come.

GenAI is based on Large Language Models (LLM) that mimic the way human beings reason and argue. Thus, LLMs can easily be used as reasoning engines to extract information, combine it or summarize it.

The main downside of LLMs is their tendency to hallucinate because they do not have a concept of truth nor any relationships to the real physical world.

Fortunately, with well defined guardrails it is possible to almost prevent hallucinations.

We build GenAI assistants that are controlled by well defined sources of truths rules.

Our agents continuously learn and eventually will specialize for each individual in the company.

We build GenAI solutions as multi-agent systems.

Multi-agent system is an old software paradigm central to AI.

It's the idea of designing complex systems as interactions of autonomous rather simple agents. Each agent is therefore responsible for one very precisely defined and limited task.

Some agents can work in semi-isolated flat groups or hierarchies. Agents talk to each other and exchange information, especially in their neighborhood, with messages.

These kind of architectures enable to build very efficient Graph-of-thought GenAIs. They are very flexible, decentralized and cloud agnostic.

A GenAI solution can be build like a lego construction, with software bricks.

Also, very interesting and advanced features can emerge almost spontaneoulsy from those architectures.

And last but not least, there is no proprietary lock-in possible anymore.