NTT Data Advanced Technology Verifies the Use of Generative AI for Financial System Development Using a Local LLM Environment
NTT Data Advanced Technology and NTT Data Financial Technology conducted a Proof of Concept (PoC) for a local LLM in a closed on-premises environment to safely streamline financial system development.
📋 Article Processing Timeline
- 📰 Published: April 15, 2026 at 00:30
- 🔍 Collected: April 14, 2026 at 16:01
- 🤖 AI Analyzed: April 19, 2026 at 15:28 (119h 26m after Collected)
NTT Data Advanced Technology Corporation (Headquarters: Chiyoda-ku, Tokyo; President and CEO: Tooru Fujiwara; hereinafter 'NTT Data Advanced Technology'), in collaboration with NTT Data Financial Technology Corporation (Headquarters: Minato-ku, Tokyo; President and CEO: Yoshiyuki Hosoya; hereinafter 'NTT Data Financial Technology'), has conducted a technical verification (Proof of Concept, PoC) of utilizing a local LLM (Large Language Model) in a closed on-premises environment, anticipating the use of generative AI in system development for financial institutions. This PoC was conducted with the aim of strengthening governance in financial institutions and improving predictability by fixing costs. Premised on the protection of highly critical confidential information, the verification tested the effectiveness of utilizing generative AI to check the consistency of design documents, improve quality, and accelerate modification work.
Based on the results of this technical verification, NTT Data Advanced Technology will proceed with further considerations and initiatives toward the realization of generative AI utilization that emphasizes security, governance, customizability, support systems, and cost predictability. Furthermore, in response to the needs of financial institutions, we will promote the use of generative AI that guarantees the protection of confidential information and security, while also supporting the use of generative AI in other industries that require similar requirements.
[Background]
In recent years, generative AI technology has attracted attention as a means of improving operational efficiency. However, particularly in financial institutions, in addition to increased costs due to pay-as-you-go billing and the difficulty of budget management, there are significant hurdles such as data protection due to the highly confidential nature of the information handled, and ensuring governance based on laws, regulations, and industry rules. Therefore, there is a growing need for an infrastructure where generative AI can be utilized safely and systematically.
In this verification, we built a generative AI platform on an on-premises environment, constructed a local LLM environment that does not transmit any various data externally, and confirmed the feasibility of applying generative AI to business operations in financial system development. Through this verification, we have confirmed that it is possible to develop systems utilizing generative AI while minimizing information leakage risks and ensuring security and governance in line with corporate security policies and internal controls.
Figure 1: Configuration image of the local LLM environment
[Overview of Technical Verification]
In order to verify whether a local LLM that does not transmit any various data including confidential information externally can be utilized for the development of financial systems, we defined three expected use cases: 'Automation of consistency checks', 'Improvement of document quality', and 'Acceleration of modification work'.
Figure 2: Main Use Cases
In this instance, we particularly conducted a verification of 'Use Case 1: Automation of consistency checks' for design document creation tasks in financial institution system development. Utilizing a local LLM built within an on-premises environment, we confirmed that it is possible to automatically check and correct without deviating from rules such as security policies, internal control mechanisms, and industry standards, thereby reducing the review process in design document creation. Moving forward, we will continue verification on the remaining two use cases.
Additionally, in this verification, we built a highly customizable generative AI platform where LLM models and configurations can be flexibly selected according to business content and objectives, and we conducted comparative verifications by executing use cases using multiple actual LLM models. Consequently, we confirmed the effectiveness and utility of being able to select the optimal LLM models and technologies such as RAG (Retrieval-Augmented Generation) to utilize knowledge specialized for financial systems.
Furthermore, assuming that the introduction of the generative AI platform will start small with a limited scale in the initial stage, we also evaluated an operational model that allows the scope of use to be expanded gradually. As a result, we confirmed that it is possible to transition to full-scale introduction while monitoring the effects on business operations, all while mitigating the investment risks at the time of generative AI introduction.
Regarding costs, we verified that by adopting a local LLM, it is possible to realize a stable cost structure (fixed costs) that is not significantly affected by the usage volume of the generative AI platform. Furthermore, we adopted an architecture that builds the generative AI platform on on-premises servers simulating customer environments. This not only promises additional cost reduction effects such as reducing cloud usage fees and suppressing network bandwidth costs, but also brings expected benefits such as easier budget management and clarifying mid-to-long-term cost outlooks.
[Future Outlook]
Based on the results of this technical verification, NTT Data Advanced Technology will proceed with further considerations and initiatives toward the realization of generative AI utilization that emphasizes security, governance, customizability, support systems, and cost predictability. Furthermore, in response to the needs of financial institutions...
Based on the results of this technical verification, NTT Data Advanced Technology will proceed with further considerations and initiatives toward the realization of generative AI utilization that emphasizes security, governance, customizability, support systems, and cost predictability. Furthermore, in response to the needs of financial institutions, we will promote the use of generative AI that guarantees the protection of confidential information and security, while also supporting the use of generative AI in other industries that require similar requirements.
[Background]
In recent years, generative AI technology has attracted attention as a means of improving operational efficiency. However, particularly in financial institutions, in addition to increased costs due to pay-as-you-go billing and the difficulty of budget management, there are significant hurdles such as data protection due to the highly confidential nature of the information handled, and ensuring governance based on laws, regulations, and industry rules. Therefore, there is a growing need for an infrastructure where generative AI can be utilized safely and systematically.
In this verification, we built a generative AI platform on an on-premises environment, constructed a local LLM environment that does not transmit any various data externally, and confirmed the feasibility of applying generative AI to business operations in financial system development. Through this verification, we have confirmed that it is possible to develop systems utilizing generative AI while minimizing information leakage risks and ensuring security and governance in line with corporate security policies and internal controls.
Figure 1: Configuration image of the local LLM environment
[Overview of Technical Verification]
In order to verify whether a local LLM that does not transmit any various data including confidential information externally can be utilized for the development of financial systems, we defined three expected use cases: 'Automation of consistency checks', 'Improvement of document quality', and 'Acceleration of modification work'.
Figure 2: Main Use Cases
In this instance, we particularly conducted a verification of 'Use Case 1: Automation of consistency checks' for design document creation tasks in financial institution system development. Utilizing a local LLM built within an on-premises environment, we confirmed that it is possible to automatically check and correct without deviating from rules such as security policies, internal control mechanisms, and industry standards, thereby reducing the review process in design document creation. Moving forward, we will continue verification on the remaining two use cases.
Additionally, in this verification, we built a highly customizable generative AI platform where LLM models and configurations can be flexibly selected according to business content and objectives, and we conducted comparative verifications by executing use cases using multiple actual LLM models. Consequently, we confirmed the effectiveness and utility of being able to select the optimal LLM models and technologies such as RAG (Retrieval-Augmented Generation) to utilize knowledge specialized for financial systems.
Furthermore, assuming that the introduction of the generative AI platform will start small with a limited scale in the initial stage, we also evaluated an operational model that allows the scope of use to be expanded gradually. As a result, we confirmed that it is possible to transition to full-scale introduction while monitoring the effects on business operations, all while mitigating the investment risks at the time of generative AI introduction.
Regarding costs, we verified that by adopting a local LLM, it is possible to realize a stable cost structure (fixed costs) that is not significantly affected by the usage volume of the generative AI platform. Furthermore, we adopted an architecture that builds the generative AI platform on on-premises servers simulating customer environments. This not only promises additional cost reduction effects such as reducing cloud usage fees and suppressing network bandwidth costs, but also brings expected benefits such as easier budget management and clarifying mid-to-long-term cost outlooks.
[Future Outlook]
Based on the results of this technical verification, NTT Data Advanced Technology will proceed with further considerations and initiatives toward the realization of generative AI utilization that emphasizes security, governance, customizability, support systems, and cost predictability. Furthermore, in response to the needs of financial institutions...