GEICO spent $1,000 to map legacy claims in a day

GEICO ran a one-day, $1,000-token war room to map legacy claims code and ship agent APIs, using Anthropic's Fleet Kit and code-modernization plugin.

Engineers collaborating around a screen showing legacy code diagrams
AI Engineer

AI Engineer featured Ford Prior, GEICO senior staff engineer in claims tech, detailing a one-day, $1,000-token war room that mapped a legacy claims domain end to end.

GEICO spent $1,000 to map legacy claims in a day - AI Engineer
GEICO spent $1,000 to map legacy claims in a day, AI Engineer

Prior said GEICO, a 300-engineer claims group inside Berkshire Hathaway, found engineers had 25 percent or less bandwidth for major V1-to-V2 work after patches, so it now routes minor work to agents to free them.

He ran the war room three times and Prior said the rentals uplift extended an API to save 8,500 calls per month, the damages transformation added new APIs, and the salvage reimagine produced a green-lit migration plan that also surfaced a prod bug he declined to detail.

The stack paired Anthropic's harness, model and Fleet Kit plus a code-modernization plugin with GEICO's own context, tools, methods and judgment gates, and while Prior described Fleet Kit as markdown rules for the Claude harness, Anthropic's v1.0.0 code-modernization plugin for Claude Code is publicly documented as built to modernize legacy codebases and manage technical debt.

The demo required local access to Prior's laptop as runtime, relied on Fable models where Sonnet and Opus-class systems did the heavy lifting, and burned budget fast when Claude running Fable set up crons that spawned sessions.

Prior said the next wins are stronger gate enforcement, better decision models and searchable context beyond markdown, but he left unanswered how the factory model scales beyond a hand-picked roster without losing the human judgment that approved every step.

© 2026 StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training requires a license. See our terms.

Watch

The videos behind the news, one flip at a time

The feed mixes the videos from our articles with startup demos and you flip through them like TV channels for free with no signup.

Daniel Singer

Written by

Daniel Singer

Editor, StartupHub.ai

Daniel Singer is the editor of StartupHub.ai, a technology expert and thought leader on AI and its applications across sectors, from fintech and healthcare to developer tooling and consumer software. He writes and tests the tools covered here thoroughly and regularly, and built StartupHub.ai to give founders, operators and buyers a clearer read on what they are actually being sold.