Enhancing Effort Estimation and Transforming the Information Technology Department through the Use of AI Agents - Promoting the Efficiency and Automation of Effort Estimation Using Agentforce and a Data Platform -

Meiji Yasuda's Information Systems Department spent approximately 10,000 hours annually on effort estimation, creating a significant burden that constrained other operations, particularly during periods when the planning and review of projects for the following fiscal year were concentrated. In addition, effort estimation depended on the knowledge and experience of experts in specific fields, resulting in challenges related to dependency on specific experts and low reproducibility.

To address these challenges, Meiji Yasuda sought to build a mechanism that would enable user departments to assess the effort required for system-related requests themselves by using data from past system development projects accumulated in Salesforce. The objective was to improve the efficiency of effort estimation and eliminate person-dependent processes.

To achieve this, following discussions with ABeam Consulting, with which Meiji Yasuda had previously exchanged views on the use of Salesforce, the company concluded that Agentforce, Salesforce’s AI agent, offered the fastest path to implementation, because data from past system development projects had already been accumulated in Salesforce. In July 2025, Meiji Yasuda launched a proof of concept for an interactive AI-based effort estimation solution using Agentforce.

This initiative established an environment in which users can assess the effort required for system-related requests themselves through dialogue with AI. In addition, by using Agentforce, the initiative automated some tasks previously performed by the Information Systems Department, including presenting similar projects and automatically creating system development request records. As a result, a substantial reduction in workloads associated with effort estimation is expected.

To ensure the reliability of estimation operations, the accuracy of AI responses is a prerequisite for operational use. Improving the accuracy of AI-generated effort estimates was therefore a critical success factor. Before building the AI agent, the project began by preparing the data foundation needed to improve accuracy. The project team analyzed the Information Systems Department's project execution process in detail and expanded the data available for learning and reference by using AI to import various materials that had not previously been digitized into Salesforce.

As a result, the project improved the accuracy of AI-generated effort estimates and established a data platform that enables continuous accuracy improvement.

Although applying Agentforce to system development effort estimation was an initiative with few precedents in Japan, ABeam Consulting drew on its accumulated expertise to build an AI-based effort estimation automation solution in a short period, enabling early validation of its effectiveness.

In close consultation with Meiji Yasuda, ABeam Consulting advanced the initiative from measure design through implementation and validation in short cycles. Although estimate accuracy was initially challenged by insufficient input data, ABeam Consulting responded flexibly by organizing the required data, designing the target data model, and proposing specification improvements, thereby improving accuracy.

Furthermore, with a view toward transforming the Information Systems Department into a next-generation organization, ABeam Consulting has provided hands-on support in exploring new use cases for AI agents. It continues to promote initiatives that shift the role of the Information Systems Department from defensive to proactive operations.