Joint Stock Company Core Retail Releases "Agent Readiness Framework" Based on Proprietary Research of Five Major AI Companies

Generative AI is evolving into a tool that helps companies compare, recommend, and even handle reservations and inquiries.

Joint Stock Company Core Retail (Minato-ku, Tokyo; Representative: Kenji Sasaki) is releasing the "Agent Readiness Framework (DRA Model)," an evaluation standard for the AI era, based on proprietary research conducted on five major AI companies.

This Framework is designed as the second phase of research, following the "AI Search Visibility Survey 2026" announced in June 2026, and is a new framework for evaluating the process by which AI "understands, compares, recommends, and acts" on behalf of companies.

The full survey, evaluation criteria, and design philosophy are available on Research Hub as the "Agent Readiness White Paper 2026."

What is the Agent Readiness Framework?

This is an index for evaluating the state in which a company is understood, compared, and recommended by AI, leading to further action, encompassing SEO (visibility in search results) and GEO (citation by AI), and extending the evaluation to the state where AI can actually complete reservations and inquiries.

The core of the framework is the DRA model (out of 100 points), which consists of three layers: Discovery, Recommendation, and Action.

Being found and being chosen are separate issues. Being chosen and being acted upon are also separate issues. The Agent Readiness Framework organizes these three stages as independent evaluation axes.

Roles of the Framework and Index

The Agent Readiness Framework is the theory and evaluation standard for assessing companies in the AI era. On the other hand, the Agent Readiness Index is a diagnostic indicator that visualizes the current status of each company on a 100-point scale based on that Framework.

The Framework serves as the "evaluation criteria," and the Index serves as the "evaluation result (diagnosis)."

This press release provides only an overview. The design philosophy of the DRA model, each evaluation item, and scoring method are explained in detail in the "Agent Readiness White Paper 2026."

Survey Overview

This Framework was designed based on empirical research conducted on five major AI companies (ChatGPT, Claude, Gemini, Perplexity, Grok).

We sent 100 prompts to each company asking, "What information do you prioritize when recommending companies, products, or services?" and obtained a total of 5 companies x 100 questions (500 responses). A team of seven experts, including AI researchers, data analysts, and critical reviewers, conducted a comparative analysis.

The purpose of this survey is not to elucidate the internal logic of each AI. The objective is to extract the tendencies and commonalities in the thinking of AI when discussing company evaluations, and to use this as the basis for the Framework's design.

Seven Evaluation Principles Common to the Responses of the Five AI Companies

Comparative analysis of the five companies' responses revealed the following seven common items.

In this survey, 100 questions and 500 responses were comparatively analyzed to extract common principles. The White Paper also includes a comparison of the responses from the five AI companies and part of the analysis process.

Key Findings

Common Approach to Disclosure of Weaknesses

AI companies tended to explain the disclosure of weaknesses and unsuitable cases as elements that improve reliability from the perspective of "information completeness."

Note that this does not indicate that "AI will preferentially recommend companies that disclose their weaknesses." The causal relationship with actual recommendation behavior has not been verified at this time and is positioned as a challenge for future empirical research.

Basis for Adopting the DRA Model: Independent Presentation of a Three-Layer Structure

Claude and Perplexity, without referencing each other's responses, independently presented essentially the same three-layer evaluation structure.

Claude: Discovery / Recommendation / Action

Perplexity: Information Layer / Structure Layer / Action Layer

Although the terminology differs, the structure is essentially the same. The research team interprets this convergence as one of the bases for adopting the DRA model. It should be noted that this does not prove that this three-layer structure is the only correct evaluation model, but is positioned as a reference finding.

Survey Transparency

The survey design, analysis methods, evaluation criteria, and replication policy are published in the Methodology.

https://readiness.coaretail.com/methodology.html

Causal relationships such as Schema implementation leading to improved recommendation rates are currently unverified. They are categorized within the research report as hypotheses.

Content Available

Content

Description

URL

Agent Readiness Research Hub

A research platform that publishes research, surveys, and evaluation methodologies, with the Agent Readiness Framework at its core.

https://readiness.coaretail.com

Agent Readiness White Paper 2026

In addition to the full details of this survey, - Framework design philosophy - Comparison of five AI companies for 30 questions - Rationale for evaluation items - Excerpts from the Methodology Handbook - Future verification roadmap are included.

https://readiness.coaretail.com/whitepaper

Methodology

Publishes survey design, analysis policy, and evaluation methodology.

https://readiness.coaretail.com/methodology.html

Agent Readiness Index (β version)

A diagnostic indicator that evaluates companies on a 100-point scale based on the Agent Readiness Framework (beta version, paid service).

https://readiness.coaretail.com/report

Regarding the White Paper Research Edition (paid version)

In addition to the summary version (free version), Research Hub offers the "Agent Readiness Report 2026 Research Edition," which provides a more detailed summary of the survey results.

This report includes the following:

Background of the DRA model design

Full text of the Agent Readiness Framework

Comparison of five AI companies (30 themes)

Rationale for the 100-point evaluation items

Excerpts from the Methodology Handbook

Future empirical research plan

This research report, approximately 20 pages long, includes the design philosophy of the Agent Readiness Framework, evaluation methodology, comparative analysis of five AI companies, and empirical data.

Furthermore, as the highest-tier version that summarizes research methods and evaluation logic in greater detail, we also offer the "Methodology Handbook."

The Methodology Handbook (approx. 30 pages) is a document for experts that systematically summarizes the research foundation supporting the Agent Readiness Framework, including evaluation methodology, analysis process, design philosophy, and scoring logic, in addition to the content of the Research Edition.

Both can be downloaded from the Agent Readiness Research Hub.

Future Developments

Representative's Comment

What was impressive in this survey was the common consideration regarding the disclosure of weaknesses across the five companies. The perspective of AI processing this not as a moral judgment of dishonesty, but as an issue of information completeness, was observed in each company, albeit with differences in expression.

I am also paying attention to the fact that Claude and Perplexity independently presented the same three-layer structure without referencing each other. "Why did different AIs arrive at the same structure?" is a challenge to be explored in future research.

The Agent Readiness Framework is a system of hypotheses. We will improve its accuracy through empirical verification, such as "whether AI recommendation wording changes before and after Schema implementation" and "how evaluation changes due to disclosure of weaknesses."

Kenji Sasaki, Representative Partner, Joint Stock Company Core Retail