What are the key features of Sales Retriever's key person research technology?

By evaluating a combination of target companies, departments, positions, names, and source evidence, it prevents contamination from other company information and lack of evidence, achieving high accuracy and comprehensiveness.

What are the verification results when compared to general AI models?

In simultaneous research of 20 companies, it recorded 6.1 times more correct answers compared to GPT5.5, reduced execution time to one-fifth, and maintained an answer accuracy of over 90%.

Why does the performance of other models degrade in simultaneous research of multiple companies?

Because general models find it difficult to make accurate judgments for each company, and are prone to insufficient citations and confusion with other companies.

In what scenarios is this technology utilized?

It contributes to the efficiency of sales activities by accurately extracting responsible party information from complex organizational structures in large companies.

Is the patent number publicly available?

Yes, the patent number 'Tokukai 2026-52570' has been obtained.

Sales Retriever Achieves 6x Performance of Latest OpenAI Models in Enterprise Key Person Research with Proprietary Patent

Published May 25, 2026 7:10 PM ・ Updated Jun 14, 2026 4:15 AM ・ 5 min read ・ Source: PR TIMES

Sales Retriever has announced that it achieved 6x the performance of the latest OpenAI models in enterprise-focused key person research using its patented technology (JP 2026-52570).

Sales Retriever, provider of AI for enterprise business development (Headquarters: Chiyoda-ku, Tokyo; Representative Director: Nariyuki Matsumoto; hereafter "Sales Retriever"), announced that it has achieved 6x the performance of the latest OpenAI models in enterprise-specialized key person research using its patented technology (JP 2026-52570).

Background of the Patent In sales activities targeting large enterprises, accurately grasping the target company's organizational structure and identifying the appropriate departments and key persons is crucial. However, large enterprises have complex organizations including group companies, subsidiaries, and affiliates, and department names vary, making it difficult to maintain accurate contact information. While generative AI and web search-based corporate research are spreading, general AI research often faces issues such as mixing information from similar companies, insufficient evidence sources, or decreased accuracy when surveying multiple companies simultaneously. Sales Retriever developed a department research technology to solve these issues by evaluating target companies, department names, positions, names, and evidence sources combined to extract actionable contact information.

Verification Results Compared to Latest AI Models In this verification, five departments (HR, IT, General Affairs, Sales, and Production Planning) were surveyed across scales of 1, 5, 10, and 20 companies. Sales Retriever was compared against representative models including GPT-5.5, Gemini 3.5 Flash, and Claude Opus 4.6 in terms of "number of correctly extracted contacts," "accuracy rate," and "execution time." Sales Retriever demonstrated superiority over general AI models across all metrics. In a 20-company simultaneous search, it achieved 6.1 times the number of correct contacts compared to GPT-5.5, the best-performing general model. While other models showed significant performance degradation when searching multiple companies, Sales Retriever maintained a high accuracy rate of over 90% in all scenarios. Regarding execution time, it consistently completed searches in approximately 60 seconds, which is 1/5 of the time taken by GPT-5.5.

FACT BOX

Source: PR TIMES
Category: Survey
Organizations: OpenAI

Sales Retriever Achieves 6x Performance of Latest OpenAI Models in Enterprise Key Person Research with Proprietary Patent

⚡ Key Points

FACT BOX

Editorial & Verification Standards

FAQ

Cite this article — HOW TO CITE

AI CRAWLER ACTIVITY