In the rapidly evolving landscape of artificial intelligence, our interactions with AI are moving beyond simple operational tasks like getting preference-based recommendations or filtering emails. Many of us now use LLMs for creative planning and complex brainstorming and so LLMs are also becoming a viable assistant to leaders, managers, and decision-makers.
However, these interactions are often limited to purely chatbot-style text interfaces that can make structured analysis difficult to track. When it comes to the “big” questions—the high-stakes, strategic decisions that define an organization’s future—our tool StrategicAI presents a Proof-of-Concept for how these advanced capabilities can be accessed through an easy-to-use, point-and-click visual interface.
We presented StrategicAI at the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2025) as a new way forward, demonstrating how a Decision Support System (DSS) for complex strategic decisions could work using cutting-edge AI methods.
The Problem: The Complexity of “The Big Picture”
Strategic decisions are notoriously difficult to automate. Unlike operational decisions, which are often repetitive and data-rich, strategic decisions are unique, ambiguous, and require a deep understanding of broad contexts. Traditionally, this meant leaders had to spend weeks manually breaking complex problems apart, gathering external data, and vetting assumptions.
The Solution: StrategicAI
StrategicAI changes the paradigm by combining human expertise with the analytical power of LLM agents, retrieval-augmented generation (RAG), and uncertainty quantification. As a Proof-of-Concept, it showcases how modular AI architectures can tackle problems that were previously thought to be exclusively the domain of human cognitive intuition.
1. Logic Tree Decomposition
The core of StrategicAI is the use of logic trees. The system helps leaders recursively break down a complex problem (e.g., “Should we enter the Southeast Asian market?”) into smaller, manageable sub-questions. This mirrors the methodology used by top-tier strategy consulting firms but is accelerated by AI.
2. Human-AI Collaboration
We believe strategic decisions should never be “black box” outcomes. StrategicAI follows a collaborative philosophy:
- User Control: Leaders can manually define parts of the logic tree where they have specific expertise.
- AI Support: The AI can suggest decompositions for unfamiliar areas, acting as a tireless brainstorming partner.
StrategicAI also complements all of its suggestions with natural-language explanations and a percentage representing its confidence in the suggestion.
3. Data-Driven Insights with RAG
To ensure decisions aren’t made in a vacuum, StrategicAI uses a multi-agent system. It actively retrieves facts from:
- Internal Files: User-uploaded documents, reports, and data.
- Online Sources: Live web data to ensure the analysis reflects the current market reality.
Why This Matters
The acceptance of this work at ECML PKDD 2025 highlights a shift in the AI community. We are moving beyond LLMs as “chatbots” and toward using LLMs as reasoning engines inside vertical-specific integrated tools.
By serving as a successful Proof-of-Concept for strategic decision support, StrategicAI demonstrates that complex organizational decisions can benefit from the rigor and speed of modern generative AI, allowing leaders to spend less time “doing the math” and more time debating the implications and weighing the values that truly drive success.
Explore the Research
You can further explore our work through the following links:
- Read the Paper: Springer Link
- Watch the Demo: YouTube Video
- View the Code: GitHub – StrategicAI
Presented at ECML PKDD 2025, Porto, Portugal.