News

We are offering a 3-month free Proof of Concept (PoC) to the first 5 companies.

Unovia Co., Ltd. (Headquarters: Shinjuku-ku, Tokyo, hereinafter referred to as Unovia) has launched a new service that uses AI to continuously analyze text on customer support platforms such as Zendesk (which also supports JIRA, Redmine, etc.) to visualize customer dissatisfaction and signs of churn.Unovia SentinelWe will begin offering "Sentinel (tentative name)" from May 2026. Prior to the launch, we are seeking companies to participate in a 3-month free PoC (Proof of Concept) implementation.Limited to the first 5 companiesWe are recruiting.

Anyone with extensive experience in customer support will likely agree: there are almost always warning signs before a complaint occurs.

The language used in inquiries gradually becomes more formal. Phrases like "I apologize for bothering you during your busy schedule" disappear. "Understood" changes to "I acknowledge." The rhythm of replies slows down. These warning signs are certainly visible if you carefully read the Zendesk tickets.

The problem is that nobody has the time to read it.

Sentinel is a service created solely to solve the problem of "not having time to read." Unovia's natural language processing engine continuously inputs ticket text, comments, and attached emails, and calculates a "dissatisfaction score" for each customer and case over time. The moment the score exceeds a threshold or deteriorates rapidly, it immediately notifies the CS team and their supervisor.

It's an unassuming product. However, the fact that the customer service department can take action days before cancellations or online backlash occur is by no means insignificant for those on the ground.

Service Overview

What I'm doing

• The ticketing system's API continuously retrieves tickets, comments, and attachments, and then performs text analysis on them.
• Customer-specific/project-specific "level of dissatisfaction"Time series scoreCalculated as
• Automatic notifications via email/Teams for score threshold exceedance and sudden changes.
• Visualize sentiment trends by department, individual, and time period using a dashboard.
• Based on industry and customer dataCustom Keyword DictionaryYou can add

From the second phase onward, support for email (Microsoft 365/IMAP) and Microsoft Teams will be added, and by 2027, sentiment analysis of voice calls (ASR integration) will be incorporated.

Contents of the preliminary PoC

• Duration: 3 months
• cost:Free(Limited to the first 5 companies)
• Includes: Keyword dictionary tuning tailored to your ticketing system / Weekly operational fit verification with the CS department / Continued use at a special price after the official release.
• Required preparations: Issuance of an API key for the ticketing system / Ticket history for the past 6 months or so (for accuracy evaluation)

Guidelines for companies suitable for participation

• Over 500 inquiries per month (threshold at which the signal-to-noise ratio stabilizes)
• We are already using quantitative indicators such as NPS and CSAT, but we are concerned about "dissatisfaction that cannot be expressed numerically."
• The head of the CS/CX department can make the decision to participate in the PoC.

FAQ

Q. How does this differ from existing NPS surveys?
NPS is a metric that "listens" to customers, while Sentinel is a metric that reads from text "written" by customers. The two are complementary, and it is not uncommon for customers with a good NPS score to actually reveal their dissatisfaction in their inquiry messages.

Q. How are personal and confidential information handled?
The SaaS version will operate in the Azure Tokyo region. Text will be anonymized before being sent to LLM. An on-premise version (to be discussed) is also planned.

Q. What about support for languages other than Japanese?
The first version will be available in Japanese only. English and Chinese support is planned for the second half of 2026.

Q. How long does it take to implement the system?
The ticketing system can be used simply by integrating with the API key. The shortest turnaround time is 3 business days. Even including data-specific tuning, we aim for completion within 2 weeks.

Q. How does it integrate with existing CRM/BI tools?
We plan to gradually add support for CSV export, Webhooks, Salesforce, and HubSpot integration.

Company overview

Unovia Co., Ltd.We are a startup that supports business transformation for companies, focusing on enterprise IT, AI, and Web3. Based on our track record of working with the Dentsu Research Institute Group, we are developing AI-powered solutions for all aspects of business processes, including AI-based expense fraud detection, SES (System Engineering Service) talent matching, and AI-powered automated expense reporting.

• Address: 4th Floor, AiAi Building, 7-6-8 Nishi-Shinjuku, Shinjuku-ku, Tokyo 160-0023, Japan
• URL:https://unoviax.co.jp
• Inquiries and PoC applications:https://unoviax.co.jp/contact/

Recruit

Unovia is currently recruiting AI Engineers, Data Scientists, and Product Managers (PdMs). If you find the question of "how to structure user feedback" interesting, and if you have experience working with Zendesk, Snowflake, PyTorch, and LangChain, please apply.

Consultation:https://unoviax.co.jp/contact/

* "Unovia Sentinel" is a provisional name. The official name will be announced when the product is released.
*The specifications, availability, and pricing of this service are subject to change without notice.
*The company names and product names mentioned are trademarks or registered trademarks of their respective companies.