In a previous post, I argued that one of the main benefits of RegDefy is that it allows you to explain your ecosystem to a wide range of stakeholders—investors, regulators, participants, and employees. If people can understand the scope, structure, and interconnections in your system, they are far more likely to trust it and work effectively within it.
In that earlier piece, I also mentioned two ways to achieve this understanding: diagramming and modelling. The distinction between them is nicely summarised in the C4 model’s section on diagramming vs. modelling. In short:
- Diagramming is about communicating a view of the system—often manually created, with emphasis on clarity and storytelling.
- Modelling is about formally capturing the structure and relationships of the system in a machine-readable way—allowing for automation, validation, and richer analysis.
When you’re building a new ecosystem—especially in a fast-moving domain like DeFi—keeping all this information coherent and consistent is not trivial. Early on, so much is in flux: business services are being defined, technical capabilities are evolving, regulations may be shifting. That’s where RegDefy’s concept of an ecosystem Digital Twin comes in. By modelling the entire system in a consistent structure, you can maintain a single source of truth that adapts alongside the project.
The Side Benefits of an Ecosystem Digital Twin
One of the big advantages of having a Digital Twin in RegDefy is that you can query it. This means you’re not just passively documenting your system—you can actively interrogate it to get meaningful insights.
We can divide queries into two broad types:
- Structured queries
- Unstructured queries
Let’s explore both, with examples.
1. Structured Queries
Structured queries are goal-oriented and guided by the structure of the model itself. The system assists you in navigating the relationships between entities to reach a defined objective.
They’re particularly useful for repeatable tasks, especially when the alternative is fiddly, multi-step navigation through a front end. Structured queries shine when you know exactly what you’re looking for and want to get there quickly and reliably.
Example: Impact Analysis
A good example is impact analysis during an incident. Suppose there’s a:
- System outage
- Data breach
- Disaster at a data centre
The immediate objective in Incident management would be to understand the impact i.e.:
- Which Important Business Services are affected
- Which regulatory clauses you are no longer compliant with
In RegDefy, you can run a wizard-driven structured query that uses modelled relationships to navigate from:
Systems → Capabilities → Controls → Policy Commitments → Regulatory Obligations
Because all these relationships are already modelled in the Digital Twin, the query is fast and reliable. You can go from a failed component to a list of at-risk compliance requirements in seconds.
Benefits of Structured Queries
- Incident response: Instant impact analysis lets you prioritise real incidents better.
- “What if” planning: Simulate potential failures to decide where to allocate resources before anything goes wrong.
- Audit preparation: Rapidly produce consistent, evidence-backed answers.
In short, structured queries are about navigating the known—they’re repeatable, precise, and highly automatable.
2. Unstructured Queries
Unstructured queries are more open-ended. Here, the user can interrogate any part of the model with any objective in mind, without being constrained to a predefined flow.
The most natural way to handle unstructured queries is to integrate the Digital Twin with a Large Language Model (LLM). This allows you to type in plain English questions like:
- “Which controls in our AML framework need work?”
- “How does the payments subsystem interact with our fraud detection capabilities?”
- “Which team owns the remediation plan for KYC non-compliance?”
The Privacy Challenge
The challenge is that some of the data in your Digital Twin is highly sensitive:
- The status or efficacy of various controls
- The exact contents of policies
- Incident and risk data
Sending all of this to a public API for processing may not be acceptable for your risk appetite.
RegDefy offers two approaches:
- Privacy-preserving local LLMs
- The model runs entirely within your controlled environment.
- No sensitive data leaves your premises.
- Perfect for highly regulated industries.
- Public APIs to state-of-the-art models
- Maximum capability and convenience.
- You accept some trade-off in privacy.
- Useful for non-confidential exploration or when strong anonymisation is possible.
This is the familiar functionality vs. security/privacy trade-off. However, it’s worth noting that the line may be shifting—especially with announcements like OpenAI and Ollama’s gpt-oss, which point towards more powerful models running locally.
How RegDefy Uses LLMs for Unstructured Queries
Once you choose your LLM approach, RegDefy can pass the most relevant slices of the model to the LLM as part of a Retrieval-Augmented Generation (RAG) pipeline. This ensures the model answers your question with reference to the latest, most accurate ecosystem data.
That means you can literally ask anything about your ecosystem and get answers grounded in the model:
- “Which of our Important Business Services depend on cloud hosting in APAC?”
- “What policy commitments would be affected if we retired System X?”
- “List the regulators we report to for each jurisdiction and the main obligations under each.”
The Future: Beyond RAG
We can see two major expansions on the horizon:
- Fine-tuning LLMs
- Tailor a model specifically to your organisation’s ecosystem and regulatory environment.
- Produces more accurate and contextually relevant answers.
- Model Context Protocol (MCP) endpoints
- Allow autonomous agents to interact with the model directly.
- Opens the door to automated workflows—like responding to a compliance questionnaire with one click.
Why This Matters
By supporting both structured and unstructured queries, RegDefy lets you get value from your Digital Twin in multiple modes:
- Structured: Repeatable, high-trust, precision queries for critical operations.
- Unstructured: Exploratory, flexible, and creative queries for analysis, decision-making, and learning.
In both cases, the core enabler is the fact that the ecosystem is modelled, not just drawn. Diagramming helps tell the story; modelling keeps the story accurate, consistent, and actionable—even as the ecosystem evolves.
Final thought:
A Digital Twin is not just documentation—it’s an active tool for understanding and managing your ecosystem. The more accessible and queryable it is, the more valuable it becomes. In RegDefy, whether you’re running a precise impact analysis after an outage or brainstorming new risk controls with the help of an AI assistant, you’re leveraging the same, coherent, evolving model.
And in the long run? You might just go from answering urgent questions in hours… to answering them instantly.
