The Platform
Most AI tools answer questions. Causality.tools builds understanding, through connecting your organisation's information into a single, growing layer that can be queried, monitored and acted on by the people who know your domain best.
Causality.tools acts as a set of composable capabilities, provisioned around each organisation's data, workflows and risk profile. Some require secure research and reporting; others require live monitoring, entity intelligence or agentic workflows, but each implementation is composed only of the capabilities it needs.
Capabilities
Plain-English explanations first. Technical depth belongs in discovery, not on a marketing page.
The structural foundation of the platform. Entities — people, organisations, assets, locations, events, documents — are extracted from source data and mapped into a graph that encodes the relationships between them. Crucially, the graph is seeded from a defined ontology at implementation, so incoming data is structured against known categories from the point of ingestion rather than accumulated arbitrarily.
Retrieval built on a connected knowledge graph retrieves by relationship, not just proximity. Graph RAG answers complex queries by traversing entity connections, temporal context and causal chains. Outputs are grounded in the client's own information, traceable to source, and reflective of the relationships the graph encodes.
Across any large, multi-source dataset, the same entity appears under different names, identifiers, spellings and formats. Resolution identifies when disparate records refer to the same underlying entity and consolidates them within the graph. Because resolution is anchored against the preset ontology, the process is consistent across sources and auditable at each step.
The platform turns complex intelligence into concise, evidence-linked outputs for decision-makers, analysts and operational teams. Reports are structured around the entities, relationships and changes that matter — surfacing what has happened, why it matters, what has changed, a confidence level, and the items requiring human review. Every claim traces back to its source.
Many risks are spatial — they emerge in places, spread across regions, cluster around assets or follow movement patterns. Entities, events and risks are geo-referenced and indexed for spatial query, then joined back to the knowledge graph so place becomes a first-class dimension of analysis.
Structured memory gives the platform a governed way to retain useful context across tasks, users, documents and workflows — known entities, previous analysis, decision history, open questions, risk history and workflow state. Memory is keyed to entities and permissioned, so context persists safely rather than through uncontrolled chat history.
The system monitors sources, processes incoming data, updates the knowledge graph and surfaces relevant changes autonomously. Human experts are embedded at the points in the workflow where judgement matters — the system handles the continuous, high-volume work of watching, ingesting and contextualising while your team handles the decision-making.
Deployment is configured around your sensitivity and regulatory requirements rather than a default configuration: hosted, private cloud, client-controlled, hybrid, or fully local infrastructure. Where required, models run entirely on client-owned hardware with no data leaving your environment.
The moral centre
Every claim on this platform can be traced back to a source, a confidence level and a plain-language reason; the discipline that runs through each capability above, from the graph itself to the reports it produces. A reasoned edge is not just a line: it carries direction, a plain-language reason, a confidence score, and a pointer to the evidence behind it. An analyst can inspect and challenge the connection, rather than accept it on trust.
Evidence chain · select a node
Architecture
The platform separates reference knowledge, shared intelligence and sensitive case data, so each implementation can reuse insight while preserving case boundaries.
The same engine supports different domains through domain-specific vocabulary and framing, without rebuilding the core.
The fastest way to understand the platform is to scope it around a real goal you have. Book a discovery call or request a workshop.