Knowledge Tool for Strategic Work

We answered exactly this question for an operator of critical energy infrastructure. As part of a pilot project, we developed an application that builds a knowledge base and automatically derives strategic insights from it, each with a verifiable source.

Paths to strategic clarity

Until now, there have been two main ways to gain strategic insights, and both are expensive. Either you tie up your own experts for weeks on end conducting research, or you commission expensive external studies that often merely confirm what the industry already expects. What’s missing here is a systematic analysis of the paths others have already taken.

Experiential knowledge itself is often unstructured and in a foreign language. What is needed, therefore, is not a “wiki” that merely stores documents, but a dynamic foundation that structures isolated facts, establishes connections, identifies insights, and clearly illustrates the range of options and courses of action available to decision-makers.

The impetus for the pilot came from us. We had developed our own approach, one that doesn’t merely aggregate knowledge but breaks it down into structures from which strategic conclusions can be drawn. This generated a great deal of interest; decision-makers wanted to move from reacting to taking proactive action. Our mission was to prove that this approach provides valuable access to global, domain-specific experiential knowledge.

How structure leads to precision

The AI agents in our methodology operate within a strictly modular, rule-based framework, the research matrix. It defines how the system views the world:

• System dimensions determine what is being examined: the prerequisites, design, effects, and framework conditions of a system.
• Stakeholder perspectives determine whose viewpoint is being considered: ranging from those who design a system to those who finance it.
• Question types determine how the questions are asked: ranging from inventory and causes to comparisons and evaluations.

This creates cells that the AI agents process systematically and in multiple steps. Because each cell goes through the same question types, the answers can be compared across markets. The AI compares markets, uncovers contradictions, and derives implications. What makes this unique is that it leads to insights into questions that users hadn’t even considered at the outset.

The choice of topics and the fine-tuning of the matrix are in the hands of the users. Before the first step of the search, subject matter experts at the client’s organization determine which dimensions, perspectives, and question types apply. The AI transfers these specifications to the knowledge base. If the result in the application is still too broad, it can be refined in a targeted manner. For example, by delving deeper into individual cells or narrowing down time periods.

No statement without a reliable source

Every answer, every data point, and every deduction is linked directly to an original document. A verbatim quote serves as a verification anchor. Sources are technically verified before being included. The system rejects unusable references and prompts the AI to search for reliable sources.

If the system finds no evidence, it states so. This honesty, combined with the ability to click through to the original source with a single click, instilled in the client’s domain experts the trust that an AI system needs, especially in a regulated company.

The knowledge base

In a short period of time, an application was developed that extremely quickly built a knowledge base consisting of over 4,000 individual pieces of information, each backed by sources. These were distributed across four knowledge spaces and range from a global market analysis to an assessment of the company’s own situation. From this, the system independently distilled approximately 200 strategic implications. The matrix logic reveals systemic connections that one would likely never have specifically sought out.

• The time it takes for the application to independently build up comprehensive knowledge and hundreds of implications and deliver initial value is reduced to just a few hours. (Time-to-Value)
• When working with content, every statement is just one click away from its source. Those preparing a decision can verify information rather than search for it, and see at a glance where the evidence ends. (Time-to-Evidence)
• Cost-effectiveness: The ongoing costs for research and model use when building such a knowledge base are in the double-digit euro range.

We see great potential for the next phase of the application’s development in systematically analyzing the strategic implications: If the proposed implications were to become part of a clear and coherent action strategy, they could make an enormous contribution to the quality of future decisions.

We won't build the world of tomorrow with the tools of yesterday

Anyone who wants to use knowledge with AI must adopt a modular approach to knowledge management. The output of AI is only as good as its context. Modules make it possible to determine, for each use case, what knowledge belongs there and according to which rules it is processed. They must not rely on any specific model, otherwise vendor lock-in will result. When the next generation of models arrives, you simply switch to the new model and continue working on the same foundation. This pilot project is built precisely according to these principles, with selectable knowledge spaces and an interchangeable model.

We would like to thank our client for joining us on this journey!