
THE CHALLENGE
AI spend climbed out of sight.
Cloud and AI bills were climbing as AI adoption spread, and the drivers were hard to see across 700+ AWS accounts. Generative AI had become the largest line item. Extended Support fees hit out-of-date databases, a 252-node dev cluster ran around the clock, and 195 file systems kept all their data in the most expensive tier.
THE APPROACH
Start from the bill, not a checklist.
Daily billing data, sized by service and account, became 14 findings, each with a confidence level. We worked each change with the client's cloud and platform teams and measured it against its own trailing baseline. A saving counts only once it shows up in billing.
WHAT WE BUI LT TOGETHER
Sized from the bill, measured in it.
AI efficiency: prompt caching on the largest AI application, cost per unit of work down ~60%
Storage: S3 Intelligent-Tiering on pipeline buckets, $174K/mo measured vs. a $231K estimate
Support costs: database upgrades off Extended Support, 8 of 11 completed by client engineers
Cleanup: backup storage down ~94%; one app loop cut config requests ~2.2M → ~9K a day
A weekly spend impact report showing identified, in-progress and measured savings side by side
BUSINESS OUTCOME
The work sits with the teams that own it.
15 initiatives are now measured in billing and worked with the client's cloud and platform teams. The client's own engineers drove many of the database upgrades, and open calls, like when an idle dev environment gets removed, now sit with engineering leadership and security.
