Toward an Era Where 'AI Resident CIOs' Become Common in Hospitals: BAIOX to Exhibit Medical AI Agent 'MedPlato' at JAMI Spring Conference

In medical institutions, the burden of CIO and CISO duties is increasing annually due to the sophistication of cyberattacks, shortages of medical informatics personnel, and the acceleration of medical digital transformation (DX) fueled by subsidies.

BAIOX Inc. (Headquarters: Minato-ku, Tokyo; Representative Director: Tetsuya Kurihara; hereafter BAIOX) will exhibit 'MedPlato,' a medical CIO/CISO support AI agent co-developed with Associate Professor Kota Torikai of Gunma University Hospital, at the '30th JAMI (Japan Association of Medical Informatics) Spring Conference, Symposium 2026 Utsunomiya,' held at Light Cube Utsunomiya from June 11 (Thu) to 13 (Sat), 2026. At the booth, they will conduct live demonstrations of the 'DX Agent' feature, which allows on-site staff to complete hospital DX planning processes—previously taking external consultants 2 to 3 months—in just a few hours or days.

Technological Barriers Are Not the Bottleneck for Medical DX: It's the Shortage of 'Design-Capable Talent'
The environment surrounding medical institutions has changed drastically over the past few years. Several long-term medical service suspension cases due to ransomware have occurred domestically, making the assurance of cyber resilience a critical management issue. Simultaneously, themes for promoting medical DX have surged, such as work-style reforms for physicians (the '2024 Problem') and the implementation of electronic prescriptions and online consultations. The Ministry of Health, Labour and Welfare's supplementary budget for the 2025 fiscal year allocated approximately 2.3252 trillion yen for medical DX-related initiatives, including 46.2 billion yen for facility and equipment promotion, 20 billion yen for productivity improvement, and 1.5 billion yen for cyber security enhancement measures in medical institutions.

Meanwhile, there is a chronic shortage of CIO/CISO personnel in medical informatics departments to execute these projects. Structural issues such as project delays, insular work practices, and loss of knowledge due to dependency on external consultants have been pointed out. From regional hub hospitals to large-scale hospitals, the refrain is consistent: 'We want to create a DX plan, but we have no one in-house who understands it,' 'It takes 3 months to outsource to consultants,' and 'Knowledge disappears when the previous person is reassigned.'

The Concept of Implementing an 'AI Resident CIO' within the Hospital
'MedPlato' was developed as an answer to these structural challenges, serving as a medical CIO/CISO support AI agent that implements a 'knowledge partner' within the hospital. It was co-developed by BAIOX and Associate Professor Kota Torikai, a specialist in medical information systems at Gunma University Hospital.

Because it is an on-premises solution (installed within the hospital) hosted on a dedicated server based on NVIDIA DGX Spark, it can be operated without exposing patient data externally. The model is a 120-billion parameter equivalent generative AI (gpt-oss-120b), which learns each hospital's philosophy, functions, and current data to generate context-aware advice.

Its main functions consist of 13 tabs: CIO/CISO Dashboard, System Monitor, Observe (usability indicators), DX Agent (hearing, DX brainstorming, circular design methods, detailed design phase), System Architecture, Asset Management, Network Interfaces, Grand Design (DX strategy roadmap), Ability Assessment, Incident Response, Legal & Compliance, Audit & Inspection (cyber resilience audit), and Disaster Management. The highlight, the 'DX Agent' feature, allows users to input on-site issues and complete everything from cause analysis, idea generation, and detailed design to automated catalog search, subsidy matching, profitability estimation, and proposal PDF generation. Attendees can see the flow of formulating a DX plan by inputting real-world problems.

Enabling On-site Staff to Complete Consulting Work in as Little as 30 Minutes
- Embedded Domain Context: Implemented as prompts, the system includes medical field interviewing techniques, DX planning workflows, and subsidy search logic, providing proposals with accuracy unreachable by general-purpose generative AI services.
- Time Compression Effect: Based on BAIOX's workflow validation, processes that previously took external consultants 2 to 3 months can be shortened to as little as 30 minutes when operated by on-site staff (Premise: total labor hours comparison based on internal BAIOX data).
- Inheritance and Visualization of Knowledge: By templating CIO/CISO expertise, the company has created a structure where knowledge is accumulated within the institution, even when staff reassignments occur.