AI software that gets older adults talking with each other.

What we are

A product company. Our business is developing and licensing our own software, Concerto, to care operators.

Product

Concerto — real-time AI facilitation of group conversation (browser-based SaaS)

Customers

Elder-care operators in Japan, starting with adult day-care centers

Revenue model

Monthly subscription per facility, with an optional family-report add-on

Stage

Working MVP in a long-term field trial at a care facility in Kyoto. Commercial launch planned for March 2027.

Founded

June 2026, Kyoto, Japan

The product

Most conversational AI is built for one-on-one chat between a person and an AI. Concerto does the opposite: the AI acts as the host of a 20–45 minute talk session for 2–3 residents and keeps its own speaking to a minimum, so that the residents talk with each other.

• Bridges topics — remembers each participant's interests and past conversations, and connects people through what they have in common.
• Balances participation — notices who has not spoken and brings them into the conversation with a short, well-timed question.
• Easy to hear and follow — a 3D avatar and real-time captions support participants with hearing loss.
• Turns conversation into care information — conversation logs and participant profiles are shared with staff; families can receive a weekly report (optional).

Concerto runs in a web browser on the facility's tablet or PC, combined with a microphone for group conversation. The product includes a session screen, a staff dashboard for participant and session management, an operator dashboard across multiple sites, and a family view.

Business model

Concerto is sold to care operators as B2B SaaS on a monthly subscription per facility. Revenue grows with the number of facilities, and an optional family-report add-on adds a second stream. Staff do not need to run the conversation themselves: once a session starts, Concerto facilitates it, so a facility can offer more conversation time without adding staff.

We do not provide consulting or contract development.

Technology

Facilitating a group conversation in real time requires knowing who is speaking, when to step in, and who is being left out — in a noisy care facility where people wear masks. Our software stack, developed in-house and tested in the field, includes:

• Multimodal speaker identification — combines sound direction, voiceprints and lip movement. About 80% accuracy with four masked speakers under real operating conditions (prototype, measured in field trials).
• Real-time turn-taking prediction — predicts when a person is about to finish speaking, so the AI neither interrupts nor leaves awkward silences. Builds on publicly released academic research (VAP models).
• Participation tracking — estimates engagement from speaking balance, facial expression and gaze.
• Low-latency pipeline — streaming speech recognition, response generation and speech synthesis, tuned to keep the pace older adults find comfortable.
• Conversation memory — remembers what each participant has talked about across sessions.

Validation so far

• Long-term field trial with a care operator in Kyoto, with feedback from facility managers and care workers built into each release.
• 68.9% of family members surveyed (n=90) answered “yes” when asked whether they would want this service used if their own parent entered a care facility (our survey).
• Selected for the Sakishima Tech Lab Program, operated by the Osaka Business Development Agency (2026).
• Selected for intellectual-property and exit-strategy research support under a care-tech startup support program commissioned by Japan's Ministry of Health, Labour and Welfare (powered by CARISO, 2026).

Company

Company name

Concerto Labs Inc. (株式会社Concerto Labs)

Founded

June 5, 2026

Headquarters

Kyoto, Japan

CEO

Takuya Yonekura

Team

7 members covering engineering and on-site validation

Contact

contact@concerto-labs.com

Figures are from our prototype and field trials and may change as development continues.