AI Agents as a Management and Business Strategy Issue: The Future Vision of Financial Institutions Seen Through the Customer Interface | Insights

AI agents are expanding the scope of generative AI beyond simply answering questions, providing information, or suggesting ideas for users, and are increasingly extending into autonomous judgment and execution support based on the user's intentions and scope of authority. As a result, the focus of AI utilization at financial institutions is shifting from operational efficiency to the design of customer experience and the creation of new added value. Against the backdrop of trends both in Japan and overseas, this article examines the key issues and future outlook for financial institutions as they seek to position the utilization of AI agents as a management theme.

* Originally published in Shukan Kinyu Zaisei Jijo, Aug 25, 2026 issue.

About the Author

Yudai Suzuki

Director
Sho Ibamoto

Manager

AI Utilization Moves into the Judgment and Execution Phase

AI (artificial intelligence) utilization at financial institutions has entered a new phase in the few years since the emergence of generative AI. Until now, the primary use case has been to answer user questions, provide information, and suggest ideas. On the ground today, however, interest is rapidly growing in AI that understands objectives, organizes the necessary steps, and works in conjunction with multiple systems and external services to "autonomously judge and execute" — in other words, the AI agent.

Many companies are already advancing their consideration of how to make use of such AI agents. What matters here is that AI agents should not be regarded merely as tools for efficiency and automation. The automation of internal operations remains a promising area of application. At the same time, the essential issue from a management and business planning perspective lies in how AI agents are incorporated into the customer interface and financial transaction processes.

It should be noted that this article does not address general-purpose AI agents offered by providers such as OpenAI, nor their use by non-financial entities in the fields of agentic commerce and payments. Rather, the focus is on how financial institutions themselves can incorporate AI agents into the financial services they provide, and how they can redesign their customer touchpoints and financial transactions accordingly.

As AI agents increasingly permeate customers' daily lives, financial institutions will need to design which customers to serve, in what scenarios, and what kind of financial experience to offer. This is not a theme confined to DX (digital transformation) and AI planning departments alone; rather, it should be positioned as a management theme that cuts across the departments responsible for medium-term management plans, channel strategy, revenue models, and risk management.

A Collaboration Model Between People and AI in Financial Transactions

How will the use of AI agents transform financial services? Financial services consist of a continuous series of data-based judgments and procedures. Services that continuously present customers with optimal options based on customer attributes, transaction history, income and expenditure, asset balances, life events, and risk tolerance are highly compatible with AI agents.

One conceivable use case is a financial concierge for individuals, built around life events such as household budgeting, asset formation, home purchases, education funding, retirement preparation, and inheritance. By inputting account information, transaction history, asset status, various life events, and information from within and outside the corporate group, and as AI agents deepen their understanding of the customer, customers will be able to receive proposals for financial products and support for procedures tailored to their circumstances, rather than having to search for such products or procedures themselves. When financial institutions sell financial products under this shift, it will encourage them to provide ongoing support for customers' life events and decision-making.

For corporate clients, it is conceivable that AI agents could present optimal options and provide execution support under constantly changing conditions, based on management and business issues such as cash flow, credit, settlement, foreign exchange, and risk hedging. In doing so, it will become important for financial institutions to build and maintain ongoing points of contact with corporate clients, starting from their management challenges.

The role of employees will also change. As AI takes on routine inquiries, initial explanations, and administrative processing, people will be able to concentrate on complex cases, exception handling, specialized consulting, and relationship building. In other words, a realistic solution is a "collaboration model" in which AI broadens the base of customer touchpoints, freeing people to focus on higher-value-added areas.

That said, this transformation will not happen overnight. The use of AI agents is expected to expand in stages — starting with proposal support and the provision of information, then moving to decision support, and eventually to the execution of procedures and transactions based on user approval or predetermined conditions and scope of authority.

The forms of execution will range from those premised on human approval to those in which the AI acts autonomously on the user's behalf within a certain scope of authority. In the future, however, this is likely to develop beyond single transactions or procedures, into a form in which AI agents deployed by financial institutions coordinate across services within the group and external, including non-financial, services to support users.

As the role of AI agents expands from proposals to the execution of procedures and transactions, on-chain finance utilizing stablecoins, tokenized deposits, and RWA (real-world asset tokenization) offers a highly compatible execution infrastructure. The AI agent, as the "thinking entity," judges the appropriate course of action based on the user's intentions and changing circumstances, while the blockchain and smart contracts serve as the "execution infrastructure," processing remittances, settlements, asset transfers, and conditional transactions in real time. There is significant room for this to develop into a financial service that operates 24 hours a day, 365 days a year.

In this world, financial institutions need not remain solely product providers. They can also become providers of financial APIs, authentication, custody, and on-chain execution infrastructure that can be securely connected to via AI.

In this way, on-chain finance enables program-based automatic execution and real-time asset transfer; when combined with AI agents, it becomes possible to make ongoing judgments that reflect the user's intentions and changing circumstances, and to connect those judgments to real-time execution. This further expands the potential for realizing new financial services that operate around the clock.

These changes will shift the value that financial institutions provide, from "selling products" to "supporting customers' decision-making and continuously supporting the procedures and transaction execution based on those decisions." As financial institutions advance their use of AI agents, designing services that deliver value through ongoing engagement with customers and their AI agents — in addition to product provision — could become a source of competitive advantage.

Use Cases Being Explored in Japan and Overseas

Momentum around the use of AI agents is expanding by the day, but the stage of development and areas of focus differ by country and by company. In the financial industry as well, a wide range of initiatives are beginning to emerge, from operational efficiency to customer-facing applications, with each company exploring its own use cases.

Overseas financial institutions are advancing initiatives that use AI assistants and AI agents in addition to sophisticating operations through generative AI. The scope of application is expanding beyond operational support to research and analysis, customer support, and the provision of other services tailored to customer needs and circumstances, with new service models being explored in which people and AI collaborate to enhance customer value.

In Japan, meanwhile, financial institutions are examining authentication, authority management, accountability, and institutional arrangements from the standpoint of how to safely incorporate AI agents into financial services that handle customer assets. In addition to inquiry response and sales support, companies are exploring their own future visions through different approaches, including new customer-facing financial services enabled by AI agents (see chart).

That said, the primary battleground at present remains operational efficiency and productivity improvement, and cases in which AI agents themselves handle financial transactions remain limited. In Japan in particular, emphasis is placed on preserving trust in financial institutions as social infrastructure that handles customer assets. This requires a more cautious approach to customer protection, accountability, allocation of responsibility, personal information protection, model risk, data governance, and security.

Of course, such a cautious approach may constrain the speed of adoption. On the other hand, if a framework can be designed that reconciles customer protection with convenience, it could also become a source of competitive advantage for new, trust-based financial services. In addition, there are many issues that are difficult for individual financial institutions to resolve on their own — such as data linkage, connections with external services, rules for coordination between agents, and how to protect users and verify authority — requiring industry-wide alignment of thinking and the formation of common rules.

In this area, in addition to rule-making by technical standards bodies and platform operators, institutional arrangements by industry associations and relevant authorities are also important. The "Basic Policy on Economic and Fiscal Management and Reform 2026," approved by the Cabinet on July 21, positions the promotion of AX (AI Transformation) as an important theme in growth strategy. The Financial Services Agency is advancing demonstration projects for AI agents, and discussions on on-chain finance are progressing within the ruling Liberal Democratic Party, among other developments, raising expectations for further progress in establishing the necessary environment going forward. At the customer interface, it will be essential to differentiate while also distinguishing between areas of competition and areas of collaboration.

The Potential to Change the Premise of the Customer Interface

As the development and adoption of AI agents progresses, the first task for financial institutions is to define, as a management issue, the purpose for which AI agents will be used. If the sole objective is cost reduction, efforts are likely to remain confined to the partial optimization of existing operations. From a business planning perspective, institutions should articulate a clear hypothesis for how AI agents will contribute to customer value creation — such as customer acquisition and deepening of transactions — and to top-line growth.

Second, institutions need to narrow down use cases based on their own strengths. A financial institution with strength in retail might start "from everyday touchpoints connecting household finances, asset formation, and payments," while one with a strong corporate client base might start "from providing cash-flow forecasting and proposal support tailored to management challenges." The starting point should be chosen not for the ease of proof of concept (PoC), but based on the institution's own assets, in ways that lead to monetization and customer value.

Third, governance needs to be designed from the earliest stage. Because AI agents may be involved not only in providing information but also in making proposals and executing transactions, incorrect responses, inappropriate solicitation, or the exceeding of authority could result in the loss of customer assets or trust. It is therefore necessary to organize, on a use-case-by-use-case basis, matters such as authority design, authentication, connections with external systems, log management, human oversight and intervention, complaint handling, and model evaluation, and to build in appropriate controls. Such governance requires a cross-organizational structure that involves legal, compliance, risk management, systems, and sales departments.

Fourth, institutions should begin with small-scale implementations that keep actual service delivery in view. Only through verification in a live-use environment will it become clear what customers are willing to entrust to AI agents and where they feel uneasy. It is important to start with a limited set of customers, products, and authority, and to continuously improve based on usage data and customer feedback. Building on this, it is essential to look beyond the PoC stage and draw up a roadmap that anticipates MVP (minimum viable product) development, commercialization, and functional expansion. Operations, organizational structures, and system infrastructure must be developed in a planned manner, taking into account business strategy value hypotheses, use cases, governance, and talent development, among other considerations.

Ultimately, the use of AI agents is merely a means to deliver value. There will naturally be areas where existing systems or human handling remain more suitable. Precisely for this reason, financial institutions need to determine the division of roles among AI, people, and existing systems, and decide, as a matter of management, which touchpoints should be redesigned on an AI-first basis. The starting point is to consider customers, revenue, risk, and required investment together, and to articulate at an early stage a path toward implementation.

Just as the internet and smartphones transformed financial channels, AI agents have the potential to change the very premise of the customer interface. In the age of AI agents, the axis of competition will not be whether or not AI has been adopted, but whether an institution can build a service model that customers choose precisely because it is built on AI. In the future, the expanded use of AI agents at the customer interface may transform not only the relationship between financial institutions and their customers, but the very form in which financial services are delivered. With this transformation in view, what management must undertake is not merely to forecast the future, but to decide where to begin redesigning its own customer touchpoints.

Insights TOP