TempestAI Inc. (Headquarters: Shibuya-ku, Tokyo; CEO: So Ikeda; hereinafter "TempestAI") has launched the "Personalized Financial AI Concierge Platform," which evolves apps in banking, securities, insurance, credit cards, and wellness into AI agent-type interfaces optimized for each individual customer.

This platform is an AI agent foundation for incorporating natural language consultations, advice based on asset and household finances, financial proposals tailored to life events, guidance on various procedures, and support for preparation before transaction execution into existing financial apps.

Meanwhile, with the advancement of frontier AI models such as Claude Mythos, awareness of AI security risks is also rapidly increasing. Therefore, TempestAI provides an AI agent platform with safety and auditability for financial institutions.

Background: Financial Apps Evolve from "Tools to Use" to "Partners to Consult"

Until now, financial apps have primarily offered functional and search-based experiences, such as balance inquiries, transfers, statement viewing, financial product listings, and FAQ chatbots.

However, what customers truly desire is not just information display, but continuous support that aligns with their lives, assets, and life events, such as:

"What is optimal for me right now?" "How should I improve my spending or asset situation?" "What preparations are necessary for the future?"

Especially in the financial domain, advanced personalization is required, taking into account the customer's age, family structure, income and expenses, asset status, investment experience, risk tolerance, life events, and past transaction history. Simultaneously, risk management unique to finance is essential, including preventing incorrect answers, excessive solicitation, inappropriate product descriptions, advice outside of authorized scope, erroneous remittances, and mishandling of personal information.

To address these challenges, TempestAI has developed the "Personalized Financial AI Concierge Platform," which combines an AI agent platform specialized for financial operations, LLM guardrails, and AI Harness.

Overview of the "Personalized Financial AI Concierge Platform"

This platform is an implementation foundation for embedding an AI concierge within financial apps, enabling customers to receive consultations, inquiries, procedures, and decision-making support through natural language.

Unlike traditional chatbots that merely guide users to FAQs, it understands the customer's context, references necessary information, coordinates multiple specialized AI agents, and provides answers, proposals, and procedural support in accordance with the rules of each financial institution.

For example, if a customer asks, "How much should I prepare monthly for my child's education expenses?" or "I want to start NISA, but I want to know an approach that suits me," or "I want to know why my spending was high this month," or "I want to change my address," the AI concierge will provide answers and procedural guidance tailored to the customer's situation.

Feature 1: Advanced Personalization with Financial AI Agent Platform

This platform utilizes conversation history with customers, trends in account balances, transaction details, held assets, investment experience, and life event information, based on the financial institution's consent, terms, and authority design.

The core "Parent AI Agent" understands the customer's intent and context, and when necessary, delegates processing to specialized agents such as a Household Finance AI, Asset Building AI, Market Analysis AI, Procedure Guidance AI, and Product Description AI.

This enables natural dialogue, explanations, proposals, and procedural support tailored to each customer's situation, rather than just general answers.

Feature 2: Risk Management with Finance-Specific LLM Guardrails

Simply introducing high-performance LLMs is insufficient for utilizing AI in the financial domain. Multi-layered controls are necessary to prevent incorrect answers, hallucinations, prompt injections, inappropriate advice, proposals that do not meet suitability requirements, unauthorized operations, and improper handling of personal information.

TempestAI's LLM guardrails control AI behavior across the following four layers:

1. Input Layer Detects and controls prompt injections, malicious instructions, requests outside of authority, and inappropriate inputs including personal information.

2. Data Layer Applies authority management, reference scope control, information masking, and data usage policies to customer information, transaction data, product information, internal documents, etc.

3. Execution Layer Incorporates human verification, approval workflows, operation logs, and pre-execution checks for execution tasks such as transfer preparation, product explanations, and procedure guidance.

4. Output Layer Provides grounds for answers, checks for prohibited expressions, confirms suitability, explains risks, suppresses definitive statements, and corrects wording in accordance with internal rules.

This supports AI utilization that meets the safety, explainability, and auditability requirements of financial institutions.

Feature 3: Secure External System Integration with AI Harness

This platform provides "AI Harness" for securely connecting AI agents with existing financial institution apps, CRMs, core banking and information systems, product databases, FAQs, policy documents, external APIs, and more.

This allows for safe operation when AI interacts with external systems, in line with the financial institution's risk management policies.

Main Use Cases

1. Asset Building Assistant / Life Simulation Supports simulations and organization of ideas for education funds, home purchases, retirement funds, NISA, and insurance reviews, based on the customer's age, family structure, income and expenses, asset status, and risk tolerance.

It explains not just general product descriptions, but also "why certain options are considered" and "what risks to be aware of" in natural dialogue, tailored to the customer's situation.

2. Household Budget Assistant AI automatically categorizes transaction details and visualizes spending trends. It analyzes fixed costs, subscriptions, dining expenses, entertainment expenses, education expenses, etc., and provides personalized improvement suggestions.

For example, it offers specific insights such as, "By reviewing underutilized subscriptions, there is room for reduction of about 3,000 yen per month," or "Dining out expenses increased this month compared to last month."

3. Procedure Navigation AI guides customers through procedures such as address changes, name changes, card reissuance, account opening, and various applications in a conversational format.

Instead of making customers search for FAQs or web pages, it organizes necessary information, required documents, and procedural steps through conversation and accompanies them until completion.

4. Support for Bank Transaction Execution Based on deposit/withdrawal patterns and past transaction history, AI organizes details such as transfer recipients, amounts, and timing, assisting in transfer preparation and presenting confirmation items.

However, for actual remittances and important transaction executions, a human verification and approval process is always incorporated to suppress the risk of erroneous remittances and unauthorized operations.

5. Financial Product Explanation and Consultation Support Provides explanations for investment trusts, insurance, loans, credit cards, deposit products, etc., tailored to the customer's level of understanding and interest.

AI's response content is controlled in accordance with the financial institution's product description rules, prohibited expressions, risk explanations, and suitability confirmation policies.

6. Wellness Proposes wellness products and services tailored to changes in health status and lifestyle.

For example, it encourages daily app usage and engagement by promoting health and daily life with approaches like, "Since you achieved your step goal this month, a health promotion incentive will be applied," or "Would you like to review the special rider details of your attached medical insurance in line with your regular health check-up schedule?"

7. Loyalty Program AI comprehensively analyzes customer purchasing data and consumption preferences to recommend personalized benefits, discounts, and partner company services.

Beyond uniform point benefits, by integrating with the underlying business matching engine, it automates proposals that deeply resonate with individual lifestyles, such as "limited coupons for frequently visited cafes" or "new service information from affiliated stores matching past purchasing trends," maximizing both the return effect and the revitalization of partner businesses.

Utilizing Personalized AI for Service Advancement

In traditional product development, customer needs were typically understood based on surveys, inquiries to call centers, and feedback from sales representatives. However, these methods face challenges in capturing customers' latent issues and their constantly changing needs in real-time.

Through interactions with the AI concierge, this platform allows for the accumulation and analysis of customer concerns, interests, life events, changes in asset status, and reactions to services, after anonymization and statistical processing under appropriate authority management.

For instance, by analyzing consultation content such as "anxiety about education expenses," "how to prepare for retirement funds," "I want to start NISA but don't know where to begin," or "I want to improve my household finances," it is possible to continuously understand the challenges customers face.

This enables financial institutions and various companies to:

Plan and develop new financial products Improve existing products Provide personalized proposals for each customer segment Improve UI/UX Advance marketing strategies

...to proceed more rapidly and data-driven than before. TempestAI positions the AI concierge not merely as a customer touchpoint, but as a next-generation product development engine, supporting financial institutions in continuous service improvement and the creation of new value.

TempestAI's Strengths

TempestAI is an AI vendor originating from the University of Tokyo, specializing in the development and implementation of AI solutions for the financial domain. It provides end-to-end support for financial institution AI development, including business understanding, AI agent design, LLM guardrails, authority design, data integration, HITL (Human-in-the-Loop), audit logs, UI/UX design, PoC design, and production operation design.

Furthermore, TempestAI possesses implementation expertise in areas requiring high safety and reliability, such as the construction of AI avatars during presidential elections, and applies technologies for prompt injection countermeasures, hallucination suppression, output control, and ensuring auditability within the financial domain.

Future Outlook

TempestAI believes that financial apps will evolve from mere transaction and inquiry tools into "personalized total life support platforms" that continuously assist customers with their lives, asset building, and decision-making.

Moving forward, TempestAI will continue to collaborate with financial institutions in banking, securities, insurance, credit cards, payments, and asset management to promote the AI agentification of each company's existing apps, customer bases, and operational data.

Leveraging its financial domain knowledge and cutting-edge AI implementation capabilities, TempestAI will support financial institutions in achieving next-generation customer experiences, operational efficiency, and advanced risk management.

Company Overview

Company Name: TempestAI Inc.

Location: 3F, Udagawa-cho Building, 14-13 Udagawa-cho, Shibuya-ku, Tokyo

Representative: So Ikeda, Representative Director & CEO

Business Activities: Development and provision of AI solutions for the financial industry

URL: https://tempestai.co.jp

Contact: https://www.tempestai.co.jp/contact

FACT BOX

  • Source: PR TIMES
  • Category: 製品リリース