By FinTech Working Group Members Ganesh Raman & Michael Paull.
The introduction of AI will result in a generational shift in the way in which large, complex systems will be managed. This will see existing systems transition from collections of siloed business functions to unified data management platforms. This transition will be driven by AI’s demand for “Always On” data.
Delivering “Always On” data will require the introduction of a new control plane providing “Always On” visibility over the source and destination of data. If the system administration teams cannot easily identify how data is generated, transformed, and consumed, managing the vertical integration of AI into the business and data management processing layers of legacy systems will become an insurmountable challenge.
This is not a system engineering challenge. This is a system observability problem, one that imposes new demands on system reliability, scalability, and performance.
At the Digital Twin Consortium (DTC), we are driving the development of this new control plane to deliver this observability. This is the Process Digital Twin of the live environment, which, in the financial services vertical, we are calling the “FinTwin” of the live system.
Background
The financial services industry is supported by large, complex systems. These systems have matured over a series of “generational changes” from the introduction of client-server implementations in the 1990s through to Cloud adoption in the early 2000s.
Artificial Intelligence is introducing the next generational shift. This will be the migration of the legacy systems from collections of isolated, batch-processing business functions to unified data management platforms. This transition will be necessary to meet demand for “Always-On” data required by the AI engines.
“We trust our data. We trust our AI output.”
“Always-On” data introduces new demands for system scalability, reliability, and performance. To meet these demands, a new control plane is needed to provide continuous observability over how data is generated and transformed. Only once this control plane is in place will the C-Suite be able to say,
“We trust our data input. We trust our AI output.”
From Data Abundance to Data Clarity
The challenge facing AI adoption is not the technology; it is its vertical integration into the existing infrastructure. Existing fragmented, business unit-specific datasets will need to be converted into consistently formatted data inputs. This, in turn, will require a data dictionary to be introduced to harmonize disparate terms into a standard set of expressions.
An example might be one system using the term “exposure,” another using “risk,” and a third using “counterpart limit.” Data harmonization will also require the terms to be aligned with their context. This “semantic alignment” will require the introduction of a data ontology.
This, however, is only the first phase of the process. Maintaining the data alignment is the challenge. This is the job of the new control plane.
“Always-On” Data requires “Always-On” Observability
The new control plane will provide visibility over:
- when new data is generated
- where data is being modified
- how data is being consumed
In other words, “Always-On” data requires “Always-On” visibility over the source and destination of data.
At the DTC FinTech Working Group, we are calling this new control plane the “FinTwin” of the financial services information systems.
Defining the FinTwin
The FinTwin is a digital replica of the executing environment, delivering continuous, ultra-low latency observability over how data is generated and modified as it transitions across an implementation. The FinTwin closes the observability gap between what we think the system is doing and what is actually happening.
Eliminating the observability gap ensures that data being provisioned to the AI engines is consistently formatted, uniformly defined, and continuously provisioned.
Introducing the FinTwin
Introducing the FinTwin is an incremental process that leverages existing:
- system knowledge
- monitors
- process models
Phase One involves examining the existing repository of system diagrams and converting these into FinTwin representations. This can be delivered by using the monitoring tools already in place, together with a software layer that sorts the system’s processing activity into its correct ordering. This output is then rendered to the modeling tool(s) currently being used. This software layer has already been built and will be demonstrated in the consortium’s test bed early in 2026.
Phase Two extends Phase One by identifying the “handshakes” that join the workflow diagrams together, enabling them to be assembled into a progressively extending system topology.
Phase Three is closing the “observability gap.” This is the bi-directional synchronization of the FinTwin with the live implementation to deliver continuous observability over how data is being generated, transformed and consumed as system changes are introduced.
Bi-Directional Synchronization
In this final phase, the FinTwin is used as a synthetic environment to test changes to the live system prior to touching production. This enables the ripple effects of system change to be identified in the FinTwin prior to any changes being introduced. Once a change is introduced, its impact can be assessed with the safety net of real-time observability of the actual impact. Should the actual impact not align with the expected impact, containment and roll-back procedures can be introduced immediately.
Summary
Transitioning the finance industry’s information systems to the AI-enabled implementations requires uniformly formatted data to be continuously provisioned to the AI engines being introduced.
The first step in this process is the semantic alignment of data and compiling the aligned data into an ontology. This investment needs to be future-proofed by the introduction of a control plane providing continuous observability over how data is generated, transformed, and consumed.
This new control plane is the FinTwin of the information system, delivering ultra-low latency, real-time “Always-On” observability over the system’s behavior.
This new control plane will enable a continuously reinforcing cycle of system optimization to be introduced within a controlled and verifiable framework.
Get involved in the FinTech Working Group.