Municipal Sales Support AI 'LobbyAI' Achieves 100% Win Rate in Third-Party Comparison Against General-Purpose High-Performance LLM (GPT-5.5)

LobbyAI Inc. (Headquarters: Minato-ku, Tokyo; CEO: Kyotaro Takahashi), a developer of public information analysis platforms, announces the results of a third-party AI comparison between its municipal sales support AI 'LobbyAI' and the general-purpose high-performance LLM GPT-5.5. In anonymous A/B evaluations assessing practical effectiveness in municipal sales, LobbyAI achieved a 100% win rate when excluding ties.

LobbyAI is an AI service that analyzes and organizes information essential for municipal sales and policy outreach, built upon a proprietary public information database. This database collects and structures data from public sources such as municipal council statements, administrative plans, budgets, committee documents, and bidding information.

This evaluation confirmed that, unlike general-purpose LLMs that provide generic topic summaries, LobbyAI combines its proprietary database with specialized analysis frameworks designed for municipal sales, enabling practical tasks such as identifying target municipalities, providing evidence-based justifications, and generating actionable next steps.

The verification tested eight common use cases in municipal sales. Outputs from LobbyAI and the comparison model were anonymized and evaluated by a third-party AI, Gemini 3.5 thinking.

Evaluation was conducted across three key axes critical to municipal sales: (1) Target Selection, (2) Source and Evidence Support, and (3) Specificity of Next Actions. Each axis was assessed from three perspectives: rigorous source review, sales leadership review, and risk review, resulting in a total of 72 judgments.

The results showed LobbyAI winning 65 judgments, the comparison model winning 0, and 7 ties. Excluding ties, LobbyAI achieved a 100% win rate, with a dominance rate of 90.3% across all 72 judgments.

Summary

This comparative validation demonstrates that LobbyAI significantly outperforms the GPT-5.5 baseline in practical municipal sales applications.

Key findings include:

※ Comparison target: Output from GPT-5.5, a general-purpose high-performance LLM

LobbyAI was highly rated for organizing primary information—such as council minutes, administrative plans, and budget documents—to help sales personnel determine 'which municipality, which department, what to propose, and when.'

In contrast, the baseline output, while useful for general policy topic summaries or drafting initial meeting questions, often required additional research for critical sales decisions based on individual municipal data, such as identifying specific departments, determining budget allocation signals, or assessing procurement readiness.

Background of the Comparison

Municipal sales require reviewing vast amounts of public information before selecting a target.

Municipal websites, council minutes, administrative plans, comprehensive plans, sector-specific plans, budget and financial reports, bidding notices, and subsidy programs are scattered across different municipalities, formats, and update schedules.

Therefore, sales personnel face significant research burdens and require deep administrative expertise to determine: 'In which municipality is the issue emerging?', 'Which department oversees it?', and 'Is now the right time to propose?'

In municipal sales, merely understanding policy trends is insufficient. What is needed in practice is information directly tied to sales decisions:

- Potential target municipalities - Candidate departments to contact - Evidence of issues based on council statements, plans, and budgets - Reasons to propose now—or reasons to wait - Risks such as existing initiatives, competitor involvement, or budget constraints - Questions to confirm in the first meeting - Next actions that can be recorded in sales management or CRM systems

While generative AI has made policy topic search, summarization, and issue identification easier, translating municipal-specific primary data into actionable insights—such as target selection, contact points, timing, and first-meeting talking points—requires accuracy, source verification, and contextual understanding of administrative processes.

LobbyAI aims to support this initial research phase in municipal sales, transforming fragmented public information into actionable intelligence for sales teams.

Overview of the Third-Party AI Comparison

The comparison assigned the same investigative tasks—based on eight realistic municipal sales scenarios—to both LobbyAI and the comparison model.

Outputs were anonymized as A/B pairs and evaluated by Gemini 3.5 thinking without knowledge of their origin. After evaluation, the anonymization was lifted to aggregate results between LobbyAI and the baseline.

The eight target themes included:

Evaluation axes were as follows:

1. Target Selection: Assessed whether the output correctly identified the target municipality, responsible department, and optimal timing for outreach.

2. Source and Evidence Support: Evaluated whether council minutes, administrative plans, and official municipal documents substantiated claims, identified issues, and justified timing.

3. Specificity of Next Actions: Assessed whether the output provided concrete next steps, including first-meeting discussion points, hypothesis validation, required documents, and CRM-ready action items.

Each evaluation axis was judged from three perspectives—rigorous source review, sales leadership review, and risk review—resulting in 72 total judgments.

Comparison Results: LobbyAI 65 Wins, 0 Losses, 7 Ties; 100% Win Rate Excluding Ties

The third-party AI’s anonymous A/B evaluation showed LobbyAI winning 65 out of 72 judgments. The comparison model won 0, with 7 ties.

Excluding ties, LobbyAI achieved a 100% win rate. Its dominance rate across all 72 judgments was 90.3%.

The evaluation particularly praised LobbyAI for extracting sales-relevant hooks—not just general policy summaries or keyword overviews—from primary sources such as recent council minutes, budget committee reports, standing committee discussions, organizational restructuring plans, budget execution status, challenges in existing programs, transitions from pilot to full-scale implementation, and specification discussions in advisory councils.

In contrast, the baseline output, while helpful for general policy summaries or initial meeting questions, was criticized for tendencies to apply generic frameworks to specific municipalities, over-rely on well-known leading municipalities, lack strong source support, and fail to adequately identify risks such as existing implementations or competitor presence.

Three Key Differentiators

1. Contextual Understanding Based on a Proprietary Public Information Database

In municipal sales, knowing that 'this policy theme is nationally important' is not enough for actionable decisions. What matters is evidence explaining 'why this municipality, this department, and why now?'

LobbyAI analyzes municipal-specific challenges and planning phases using its proprietary database of public documents—including council statements, administrative plans, budgets, and committee reports—collected and structured from municipalities and central government agencies.

This enables the extraction of timing-relevant insights such as organizational changes, plan revisions, budget execution trends, transitions from pilot to full implementation, and specification discussions in councils—factors critical for sales timing.

In contrast, general-purpose LLM outputs, while capable of summarizing policy trends, often result in generic recommendations like 'check the plan' or 'review the minutes,' requiring further manual research.