Major Life Insurer: Technical Support for a Company-Wide Operations Transformation
Global Consulting Firm
Context and Problem
A major life insurer was transforming its company-wide operations toward work that assumes AI agents.
Joined as one of a two-person tech team and advised each department at its meetings. Set up the team's task management on GitHub Projects.
Checking Whether an AI Agent Can Run the Work
- Breaking down the closing work: split the tax department's closing manual sentence by sentence into 13 operations.
- Checklist: built an HTML checklist to pre-assess and then test hands-on whether an AI agent could run each of the 13 operations. For each operation it records understanding (OK or question, with a memo), success-criteria checks, and the hands-on result (OK, NG, or partial, with a memo); it autosaves in the browser and exports JSON.
- Guide and instructions: wrote an HTML guide that explains the workflow with diagrams, and the work instructions handed to the AI agent.
How It Was Verified, and What Changed
Observed: hands-on tests showed fixed row/column instructions broke when table layouts changed; revised the AI agent instructions to locate data by headings and seek plan approval before running.
Agent Skills Primer
Toward agent-first operations, systematized how to define and convey work procedures in a form AI can follow (Agent Skills) and wrote a primer for non-technical client staff.
Technologies
AI AgentAgent SkillsHTMLJavaScriptGitHub Projects