New Protocol "VOICE Ado" to Fill the "Responsibility Vacuum" and Transform Judgments into Evidence, Now Generally Available

Today, VOICE PROTOCOL (planned and operated by Japan Realter Co., Ltd., Headquarters: NS Building 3F, 1-32-4 Shinjuku, Shinjuku-ku, Tokyo, Representative Director: Shigeyuki Tsutsumi) has commenced the general availability of "VOICE Ado" (formal name: VOICE AD OPTIMEZE), a new decision-making infrastructure that observes the structure of advertising traffic, fixes human judgments as evidence, and makes them verifiable by third parties.

This is not merely an advertising analysis tool. In an era where AI and humans coexist, it is a new responsibility infrastructure designed to fix "who, observing what structure, made what judgment, and whether that judgment can be verified later."

VOICE Ado visualizes only the "structure" of behavior. Heatmaps.

【What is needed in the AI era is not "stronger AI"】

Currently, companies can accelerate a large volume of decisions through AI. Ad operations. Customer support. Screening. Moderation. Report generation.

However, at the same time, critical business decisions are proliferating without the ability to explain "why that judgment was made," "who held the final responsibility," or "whether it can be verified by a third party later." In many companies today, no one can explain, weeks later, "who approved the recommendation made by AI," "why that customer was excluded," or "why that ad was stopped."

AI has lowered decision-making costs. But simultaneously, it has explosively increased responsibility costs.

VOICE Ado is the infrastructure to fill this "responsibility vacuum." It is a new responsibility protocol that fixes the boundary between AI and human judgment, making it verifiable later. That is VOICE.

・What is VOICE Ado?

VOICE Ado is a decision-making infrastructure that observes the "structure" of advertising traffic and converts its judgment process into third-party verifiable evidence.

Ado does not track people. It only observes "behavioral bias," "concentration of behavior," and "repetition of structure." For example, it observes "structurally biased movements" such as CVs concentrated in a short period, abnormally similar user flows, extremely biased dwell times, or repetitions of the same pattern.

Importantly, Ado does not conclude these as "fraud." What Ado does is merely fix the fact that "this structure was observed."

・Responsibility cannot be attributed to models

Current generative AI lacks legal personality as an accountable entity. Therefore, automated judgments made solely by AI can pose significant audit and litigation risks for companies. The biggest risk of automated judgments by AI alone is not the false detection itself, but the inability to "explain it later."

In VOICE Ado, AI does not "judge." AI's role is limited to generating reference information, such as comparing observed structures, summarizing, explaining, and organizing trends.

The final judgment is always made by a human (Human Review). And the fact itself of "when, what structure a human observed, and how they judged it" is fixed (Freeze) as tamper-proof evidence.

・What's truly scary is not AI. It's operational accidents.

In an enterprise environment, the most dangerous thing is not AI itself. What's truly scary are "operational accidents" such as Webhook duplication, asynchronous order reversal, multi-tenant boundary collapse, cache remnants, expired evidence, and CV stoppage due to SDK failures.

VOICE Ado has invested enormous design costs into fixing these boundary conditions, even more than its analysis logic. We have structurally contained operational accidents based on the premise that "humans always make mistakes."

Check on the management screen

Visualization by AI, etc.

Always human intervention

Clarification of responsibility

Third-party verifiable

Tamper detection possible

Does not hold PII

Does not acquire personal information

・"Non-destructive SDK" that doesn't stop client sites

When introducing marketing tools, the IT systems department of companies most fears that "external scripts will stop purchases or applications on the production site." VOICE Ado's SDK is designed with top priority on non-interference with client sites.

* Fully asynchronous, non-blocking design: The SDK does not block the main thread or CV processing.

* 2-second timeout: Even if an Ado-side failure occurs, observation processing is voluntarily discarded, and site processing continues.

* Kill Switch: Can be stopped immediately if necessary.

* DNT (Do Not Track) compliant: Respects browser opt-out signals.

* CSP compliant: Designed to operate even in strict CSP environments of financial institutions and large corporations.

VOICE Ado is not a "client site processing system." It is an asynchronous observation and judgment assistance layer.

・Does not hold PII

The best way to reduce information leakage risk