Company-wide discovery: Knowledge Pilot

In the design phase, AI tools have a broader audience: they can quickly produce a working draft of a document based on the brief, which is then fine-tuned. The greatest reach, however, comes from the discovery tool — the internal Knowledge Pilot. Internal documents are uploaded into the system, from which the AI composes answers and at the same time indicates which sources it relies on. The same principle helps outside development as well, for example in the HR department or IT support, where the answers are based on guidelines, methodologies, and procedures.

The Eviden Machine One platform and the question of planning

All the tools mentioned are part of the Eviden Machine One for Delivery suite. The solution can be run as a service or deployed directly at the customer site; it consists of a portal interface and plugins, a server component, and connections to cloud language models. Azure is used most often, but it also works with Google or Amazon clouds. When deployed at the customer, documents do not leave their environment and are not used for general model training.

The lecture also touched on planning, which accompanies all phases of a project. Simple tests — adjusting a number in an equation to yield a different result, or writing a poem where the last line is the opposite of the first — were handled by AI. With a touch of hyperbole, this implies that if we know how to use it properly, it will help almost everywhere. Nevertheless, it remains up to people to set meaningful tasks and critically evaluate the results.