Cloudflare’s internal Cloudflare OS rollout
This is the operating-model companion to the Cloudflare OS launch: start from jobs to be done, build canonical org context, give engineers and non-engineers different paths, and treat AI as a toolmaker under existing permissions.
Links: Original source · Related link
Logged at IST: 2026-08-05 21:03 IST
What it is: Sam Rhea’s account of how Cloudflare rolled out Cloudflare OS internally and what it learned about AI adoption across technical and non-technical teams.
Gist: The article is less about the product surface and more about the operating model behind it. Cloudflare started cautiously, then hit the familiar inflection point: better agents made employees want production access to many systems of record. The CIO framing is that the company had to enable that energy while keeping internal systems, customer data, and permissions safe.
The principles are useful: use AI to spend more time with customers and build better technology, give everyone superpowers rather than only developers, keep humans responsible for output, invest in canonical organizational context, and never give a user more permission through AI than they already have.
Cloudflare split the rollout into two paths. For engineers, it built an Engineering Codex and put agents into review loops across merge requests, technical designs, and incident reports. For everyone else, it first ran a “magic AI email bot” manually to learn the real jobs to be done before turning those patterns into skills, context files, data connections, and workflows.
Cloudflare OS then became the browser-based place to run those skills and workflows under Zero Trust, MCP portals, AI Gateway, DLP, and model controls. The v2 shift is the important product lesson: instead of spending tokens to re-run mostly deterministic skill sessions, let users describe workflows, have agents create the code behind them, and run those apps or agents on demand, on schedule, or from events. The help-desk example shows this as a secure app around ticket metrics and draft responses, backed by gatekeepers and shared under each user’s own permissions.
Newsletter angle: Strong org-design companion to the Cloudflare OS launch. The headline is not “deploy a chatbot”; it is “learn the jobs to be done, encode organizational context, then let AI generate deterministic tools inside permissioned workspaces.”