Rippling CEO Parker Conrad built an internal dashboard correlating employee org data, Claude token spend, and GitHub metrics to flag inefficiency

One telling insight – by looking across GitHub pull request data and Claude spend – Rippling AI identified engineers with high AI spend, whose peers frequently asked them to re-do their work in code reviews 3/

Important to call out that Rippling AI is building real BI, not just svg images - you can drill down to see individual records, apply global filters and parameters, and share dashboards with either the owner’s or the recipient’s data permissions.

I asked Rippling to build a dashboard to flag inefficient spend. It joined org data with claude spend, github pull-request and code-reviews, and perf management data, and one-shotted an analysis that changed the way we are overseeing token spend internally 2/

No single metric can measure engineering output. But a constellation of signals helps. The challenge is stitching data across systems where identities don’t match (“coder3” on GitHub vs. “adam@yourco.com”). Rippling’s employee-centric data model makes those joins trivial. 6/

“Spend per pull request” is interesting but not a great measure – our top and bottom performers are roughly tied for the lowest token spend per pull request. Neither is “total spend” something you want to drive down – our highest performers also have the highest total spend. 4/

@parkerconrad 🫣 not me

@parkerconrad the signal isn't the Claude spend. it's that reviewers are doing the vetting the engineer skipped