▶ Watch ↗15:38
AI Cut Our Research Time From 7 Days to 15 Minutes — Gareth Yoder, Vanguard
Read the full talk →Key ideas
Scroll to read ↓Gareth Yoder explains how Vanguard modernized a cash ledger through parallel data, producer, and consumer migrations—and made its legacy code and migration guidance available to AI tools.
- Exposing legacy code through MCP let the team ask AI tools research questions directly; Yoder reports a feedback loop falling from seven days to 15 minutes.2:12 ↗
- Daily migration audits compared both row counts and individual attributes. Expected differences from fixes made only in the modern system needed allowlisting.3:11 ↗
- A common gateway gave 35 producers a modern write path, while queryable migration documentation reduced the team’s reported support burden.4:12 ↗
- Real-time, batch, and streaming consumers had distinct access patterns; at-least-once event delivery made safe retry handling important.6:38 ↗
- Consumer business justifications can connect an application’s technical failure to its impact, while metrics, alarms, and traces remain essential for daily operations.9:29 ↗