Integrated Value Network (IVN): AI-powered governance intelligence and regulatory traceability platform

The problem

Government operates through thousands of interconnected laws, regulations, policies, programmes, grants, contracts, budgets, and performance measures. These elements are often managed in separate systems, making it difficult for public servants to understand how requirements connect, where obligations originate, and how changes in one area affect others.

This fragmentation creates several practical challenges. Staff spend significant time interpreting and reconciling requirements across documents. There is limited visibility into how policy is implemented and its downstream effects. Effort is duplicated across agencies and programmes. Conflicting requirements are difficult to identify. Compliance obligations are hard to trace back to their authoritative source. And transparency is reduced for citizens, grantees, contractors, and frontline staff who need to understand what is required of them.

Traditional legal and policy research tools focus on individual documents rather than the relationships between them. An analyst asked to assess how a new grant requirement relates to existing reporting obligations, for example, has no central way to map those connections. In practice, this means building a spreadsheet of citations by hand, a process that can take weeks and still risk missing important links.

The difficulty grows with the size of the collection. Each new document can relate to many of those already gathered, so the number of possible connections rises far faster than the number of documents. At the scale of a national government's rules, mapping these links by hand, or in an ordinary spreadsheet, becomes impractical, and it is hard to be confident that every relevant connection has been found.

The solution

The Integrated Value Network (IVN) uses artificial intelligence and graph-based knowledge management to create a connected view of governance. It has been in active use at two US federal agencies and piloted at several more.

The platform takes in laws, regulations, guidance documents, policies, grants, contracts, and performance measures and identifies relationships among them. Large language models help extract authorities, obligations, definitions, dependencies, and implementation requirements from unstructured text.

A knowledge graph is a way of storing information not as a list of documents but as a network of connections, where each piece of information is linked to the things it relates to. In this case, a regulation is connected to the law that authorises it, the programmes it governs, the funding it is tied to, and the reporting obligations it creates. These connections allow users to navigate across organisational boundaries rather than searching document by document. The system does not settle the connections on its own. Using widely available AI assistants, it proposes candidate links, which a subject-matter expert then accepts or rejects, and only confirmed connections are added to the graph. This expert review is what enables reliance on the results.

Using the platform, staff can trace requirements back to their originating authority, identify duplication and overlap across programmes, detect potential conflicts between requirements, understand how a policy change may affect downstream operations, and map connections between legislation, regulations, funding, and service delivery. The same method reaches beyond laws and regulations to the wider body of governance material: strategic plans, programmes, performance measures, goals, directives, instructions, manuals, and even the informal rules of thumb that staff work by.

Previously, consolidating this information was done manually. A single analyst spent close to a year of full-time work building the first dataset. The dataset was held first in a Microsoft Excel spreadsheet and then in Power BI, a data visualisation tool. Both were eventually outgrown: Excel could no longer track the growing number of connections, and Power BI could not display them all. Only artificial intelligence made it feasible to work at this scale and to aim for a complete set of verifiable connections rather than a sample.

The goal is to provide public servants with a transparent, searchable, and explainable view of how government requirements are connected.

The impact

1. Proof of concept demonstrates the ability to connect complex governance documents

The project is currently in the late prototype stage. Early proof-of-concept work demonstrates the ability to transform complex governance documents into a connected knowledge graph that supports traceability, analysis, and decision support.

2. Tasks that took weeks could be completed in hours

In one illustrative scenario, an analyst asked to assess how new grant requirements relate to existing reporting obligations used the platform's filters to map all dependencies. The graph view revealed that one grantee reporting requirement was duplicated across two programmes, enabling a recommendation for consolidation. A task that would previously have taken weeks of manual work was completed in a few hours, with a clear, evidence-backed explanation of how requirements connect and what the downstream effects would be.

3. Surfacing relationships that were previously unclear

Government rules not only set out what agencies must do; they also dictate how governance documents must connect to one another. One such rule might require that every strategic objective include both a set of identified risks and a set of performance measures. Because these connections are mandatory, they create a consistent structure across the documents, and once that structure is captured in the graph, users can ask more specific questions. For example, for any given strategic objective, the graph holds both the risks recorded against it and the performance measures set for it. That makes it possible to ask whether risks actually relate to how the objective's success is measured. Often they do not, because the risks and the measures were drawn up by different teams under different chains of command that may not be coordinated. The platform brings that disconnect into view.

4. Expected benefits across compliance, coordination, and transparency

The long-term vision is a shared governance intelligence platform that supports policy design, compliance management, service delivery, auditing, and public transparency. Expected benefits include reduced time spent on compliance and policy research, improved coordination across agencies and programmes, more informed policy development, and stronger evidence-based decision-making.

Lessons learned

• Graph-based representations reveal relationships that document review cannot. Storing governance information as a network of connections rather than a collection of documents makes it possible to identify relationships, dependencies, and conflicts that are difficult to spot through traditional methods.

• AI significantly accelerates the extraction of information from complex regulatory text. Large language models can process unstructured policy and legal documents and extract structured information, such as authorities, obligations, and dependencies, at a speed that would be impractical to achieve manually.

• Governance documents are often ambiguous and context-dependent. Establishing consistent data models and relationship definitions requires substantial policy and legal expertise. The meaning of a requirement can change depending on its context, and this cannot be fully automated.

• Validation and quality assurance remain essential. Even when AI-assisted extraction is used, human review is necessary to ensure accuracy. In hindsight, the team would have built human-review workflows into the system from the start.

• Engage operational users earlier. Compliance professionals and operational users should be involved in prototype testing earlier in the process. Their practical knowledge of how requirements are applied in practice is essential for building a system that reflects real-world use.

• Introduce governance and ontology standards early. Formal standards for how information is categorised and how relationships are defined should be established at the beginning of development rather than retrofitted later.

The project is designed within the context of U.S. federal governance, including statutory authorities, administrative regulations, grant management requirements, procurement rules, and agency policy frameworks. Design considerations include transparency, auditability, accountability, records management, and emerging responsible AI principles applicable to public-sector systems.