Every prediction is only as good as the knowledge behind it. Feed a model shallow data and you get shallow answers. This is why Foretide starts every simulation by building something most prediction tools skip entirely: a knowledge graph constructed directly from your documents.
What Is a Knowledge Graph?
A knowledge graph is a structured representation of real-world entities and the relationships between them. Unlike a database table where data sits in rows and columns, a knowledge graph captures how things connect.
For example, instead of storing "Company A" and "Company B" as separate entries, a knowledge graph represents that Company A is a supplier to Company B, that they share three board members, and that Company B recently acquired a subsidiary that competes with Company A. These connections are what make predictions meaningful.
How Foretide Extracts Knowledge from Your Documents
When you upload documents to Foretide -- reports, memos, market analyses, organizational charts, strategy decks -- the system does not just index keywords. It performs deep entity and relationship extraction.
Entity Recognition
Foretide identifies the key actors in your documents: people, organizations, products, markets, regulations, and events. Each entity gets a structured profile with attributes pulled directly from the source material.
Relationship Mapping
Next, Foretide maps how these entities relate to each other. Who reports to whom? Which company supplies which product? What regulation affects which market? These relationships form the edges of the knowledge graph, creating a web of connections that mirrors your real-world context.
Contextual Enrichment
Beyond simple connections, Foretide captures the nature and strength of relationships. A partnership announced last week carries different weight than one established five years ago. A competitive relationship between two firms is fundamentally different from a collaborative one.
The Temporal Dimension: Relationships Change Over Time
Here is what makes Foretide's approach different from a standard knowledge graph: time matters.
Most knowledge graphs are static snapshots. Foretide builds temporal knowledge graphs where relationships have a time dimension. A supplier relationship that ended six months ago is treated differently from one that is active today. A regulatory change scheduled for next quarter is modeled as a future event that will reshape connections.
This temporal awareness is critical for simulation accuracy. When agents run their simulation, they do not just know who is connected to whom -- they understand how those connections have evolved and where they are heading.
How the Knowledge Graph Powers Agent Intelligence
The knowledge graph is not just a visualization tool. It is the foundation that gives every simulated agent their understanding of the world.
When Foretide generates agents for your simulation, each agent receives a slice of the knowledge graph relevant to their role. A simulated market analyst knows about market trends and competitive dynamics. A simulated regulator knows about compliance requirements and enforcement patterns. A simulated consumer knows about product alternatives and price sensitivity.
This means agents do not operate on generic assumptions. They make decisions grounded in the specific context you provided, which is why Foretide's predictions reflect your reality rather than abstract theory.
What Makes Foretide's Approach Different
Traditional AI prediction tools treat documents as input data to be summarized or queried. Foretide treats them as the raw material for building a living model of your world.
The difference shows up in the results. Instead of getting a single number or a trend line, you get a full simulation where thousands of agents interact within the context extracted from your own documents. The knowledge graph ensures that every agent decision is anchored in real relationships and real dynamics.
If you want to understand the full process from document upload to simulation results, visit our how it works page to see the pipeline in action.
