Findy AI+ Launches 5 New Features to Visualize AI Best Practices, Eliminating 'Individual Skill Dependency' and 'Black Box' in AI Agent Utilization

Findy Inc. (Shinagawa-ku, Tokyo; Representative Director: Yuichiro Yamada; hereinafter 'the Company'), which provides an engineer platform, announced that it has started offering 5 new features on June 4, 2026 (Thursday) for its AI product 'Findy AI+', which automatically analyzes AI tool usage in development organizations and supports the presentation and execution of improvement actions.

The new feature set consists of 3 features provided via a web dashboard: 'Session Log Analysis', 'AI Cost Efficiency Analysis', and 'AI Agent Usage Analysis', and 2 features provided via an MCP server: 'Harness Maintenance Analysis' and 'Harness Maintenance x Productivity Analysis', for a total of 5 features.

By combining AI agent session logs with GitHub data, the features visualize generative AI usage best practices within an organization and support automated improvement. These features are available within existing 'Findy AI+' plans at no additional cost.

◆ Background for New Feature Release

◾️ 'Individual Skill Dependency' and 'Black Box' of AI Utilization

While generative AI tools such as Claude Code, GitHub Copilot, Cursor, and Devin are rapidly spreading within engineering organizations, it remains difficult to see within the organization how commands and agents are being mastered and how results are being maximized, even after introducing AI tools. As a result, the productivity gap between engineers who are advanced in AI utilization and those who are not is widening, and the know-how of high-performing teams and members is not shared with other teams, failing to lift the entire organization.

◾️ Increase in AI Costs Makes ROI Visualization a Management Challenge

Furthermore, as the use of AI agents expands across the organization, the rapid increase in AI costs due to higher token consumption is emerging as a new management challenge. However, currently, there is no means to quantitatively determine which repositories or teams' AI usage contributes to output (e.g., merged PRs) or whether the cost is yielding results, making it difficult to verify return on investment.

Additionally, the state of harness (AI instruction files, AI instruction folders, etc.) maintenance, which maximizes the effectiveness of AI agents, varies greatly by repository and team. There has also been no way to numerically understand how investment in harness maintenance affects productivity.

To address these challenges, the Company is providing 5 new features that combine the GitHub data analysis technology of 'Findy Team+', which has accumulated development activity data from engineering organizations for many years, with the AI agent session logs collected by 'Findy AI+'. This multi-angle visualization of an organization's actual AI usage and automation of improvement actions helps build AI utilization, which tends to remain an individual skill, into an organizational capability. These new features are available from 'Findy AI+'.

◆ Overview of New Features

── Web Dashboard ──

① Session Log Analysis: Visualize the entire organization's AI usage status with data

Analyze conversation session logs with AI agents at the organization, team, and individual levels to continuously visualize the frequency, depth, and trends of AI usage. This allows organizations to understand their actual AI usage, which previously relied on 'intuition', as objective data.

② AI Cost Efficiency Analysis: Visualize cost-effectiveness and reduce wasteful spending

Aggregate AI usage (token volume) by organization, team, and member, and link it to actual deliverables (code changes/additions, number of pull requests, etc.) to calculate and visualize 'AI cost per output'. This addresses the management challenge of 'whether AI is leading to results commensurate with its cost' by supporting the improvement of teams with large ROI improvement potential and inefficient usage patterns.

③ AI Agent Usage Analysis: Quantitatively understand whether developed harnesses (Skills) are actually being used

Visualize how often harnesses (Skills) developed internally for Claude Code to handle specific tasks (e.g., automated code checks) are actually used, by organization, team, and user. This allows data-driven understanding of how much developed Skills like 'create-pull-request', 'self-reviewer', 'commit-message', and 'setup-worktree' are utilized in actual development workflows.

── MCP Server ──

④ Harness Maintenance Analysis: List the maintenance status and improvement points for 'environments where AI can operate autonomously' in each repository

Analyze AI instruction files and folders (e.g., AGENTS.md, /AGENT) in each repository and list the configuration status for major AI tools (Claude Code, Copilot, Cursor, Devin, etc.) across the entire organization. Identify repositories with lagging maintenance by checking the number of configured Commands, Agents, Skills, and MCP servers, and automatically recommend improvements based on the configurations of leading repositories.

⑤ Harness Maintenance x Productivity Analysis: Quantitatively understand ROI through correlation with GitHub data

Visualize the number of configured Commands, Agents, Skills, and MCP servers per repository alongside productivity metrics such as the number of merged PRs from GitHub. By showing the correlation between harness maintenance level and productivity metrics, this feature automatically supports decision-making on 'which elements of maintenance are most effective for improving productivity' and 'which repositories/teams should be prioritized for maintenance investment'.

◆ Primary Target Audience and Expected Effects

These features are primarily targeted at the following roles within engineering organizations that utilize generative AI:

- CTO / VPoE: Quantitatively understand the organization's overall AI usage level through session logs x GitHub data, optimize AI costs, and utilize ROI data for management reporting.
- EM / Tech Lead: Compare their team's command usage and harness maintenance level with other teams and execute specific improvement actions.