“requests and tokens tell us like usage, like merge requests and pipelines tell us about platform activity, but they do not tell us what value … providing. … the more meaningful long-term question really is like what is the future velocity and outcomes that we can measure with less friction, getting changes out safely.”
Rajesh Bhatia, Senior Director of Engineering, Cloudflare. Leads the team that built and runs Cloudflare's internal AI engineering stack, so he is in a position to know its adoption and design; Cloudflare also sells the platform products it uses internally.
Cloudflare sells AI or agent tooling, so it has a reason to publish flattering numbers about AI. The passage stays in the archive; read it with that in mind.
How this was checked
Not yet checked by an independent second reader. The wording, speaker and date come from the source linked above; treat it as a lead until it is confirmed.
First reader (reader-web lane, 2026-11-01): two passes over the original audio, a whisper transcription of the full recording and a separate pass over the YouTube caption track, agree on the words quoted, except as noted. Not yet checked by a blind second reader. Cuts: 'that they're providing' to 'that we're providing' (passes disagree, so only 'providing' is kept) and the sentences between. Both passes agree on the rest. Punctuation only otherwise.
How to cite this passage
Rajesh Bhatia (Cloudflare), “From Assisted to Delegated: Cloudflare's AI Engineering Stack (interview, Background Agents Summit 2026)”, Ona (YouTube), 2026-05-06, at 20:49. https://www.youtube.com/watch?v=MbLdrAZFQRs Archived at https://the-ai-record.com/passages/cloudflare-requests-tokens-tell-usage-not-value-2026
Context. Describing early misses with Cloudflare's internal AI code reviewer, before the team tied its findings to machine-readable engineering standards.
Context. Asked which month or model made AI coding tools start to work at Cloudflare; he had said the company began piloting coding assistants in mid-2025.
Context. Answering how Cloudflare reached 93% of R&D using AI tools; he has just said adoption was slow at first because the models were not good enough.
Cloudflare