EventsWebinar Recording
What's a CMS in the Era of Agents?
Learn how we replaced our own CMS with AI agents. We moved 850 pages off Sanity, and it took one engineer five weeks. Ben explains the pros and cons, and what to consider before making the move.
- Aired live on July 15, 2026
- 53 minutes
- On-demand recording
Get the recording
Enter your email and we'll send you a link to the full session recording.
What you'll get
- The complete session recording, free, sent to your inbox.
- Three checks to run before you leave your CMS, from an 850-page migration.
- The real cost breakdown: time, tokens, and engineering hours.
The session shows exactly how the migration worked, where the agents needed correcting, and what it would take for your team to make the same move. Ben Hofferber, SVP of Agentic Product Engineering, leads the walkthrough, and what agentic AI is and how agents actually work covers the terms he uses throughout.
Rangle is the team behind Nike's personalized AI shopping assistant and Sanofi's Elements Design System, and the system in the session publishes the very page you're reading: rangle.io has run on agent-published content since May.
What's in the recording
- Three checks to run before you leave your CMS: can an agent already edit your content faster than your CMS UI, is your team fluent enough to work this way, and how much of the feature set you pay for do you actually use.
- The real cost breakdown: what the migration actually cost in time, tokens, and engineering hours, and how that compares to the three months and three developers our last migration took. The numbers, not the pitch.
- The gotchas and the checks that catch them: schema drift, redirect debt, content quality when agents write, and the validation gates that keep each one out of production.
Who it's for
You manage a content platform for a mid-to-large organization. Your CMS contract is up for renewal, or you're evaluating a migration. You've heard the AI hype and you want production evidence before you make a call. Engineering leads, heads of marketing, technical founders: what you have in common is that you're the person who has to live with the decision.
You'll leave with a clear framework for deciding whether an agent-based content architecture is right for your team, including the cost model, the timeline, and the checks that keep agent-written content production-safe.
Frequently Asked Questions
Not necessarily. Rangle replaced its own CMS entirely: content lives in a code repository, agents handle the mechanical publishing, and CI runs the checks a CMS workflow used to. What you still need is somewhere for content to live, a review step you trust, and a publishing pipeline. The recording walks through when that trade makes sense and when a traditional CMS is still the right call.
The costs move: you stop paying a per-seat CMS licence and start paying for engineering time and model tokens. In Rangle's migration, the full move cost one engineer working part-time for five weeks, plus a flat Claude Code subscription and about $20 in batch API calls. The session breaks the numbers down.
Rangle moved 850 pages off Sanity in five weeks with one engineer working part-time. The recording covers what made that speed possible, and which parts of a migration don't compress no matter how much automation you apply.
It can, and this is where migrations most often go wrong: redirect debt, schema drift, and metadata that quietly changes shape. The session covers the redirect discipline that brought every old URL across, and the validation checks that protect content quality and rankings during the move.
Reliable enough that rangle.io has run on it in production since May 2026, including this page. The honest answer is that reliability comes from the pipeline, not the model: review gates, CI checks, and content-quality checks between the agent and the live site. The recording shows where agent output fell short along the way and what caught it.
Three things need constant attention: schema drift (content whose shape no longer matches what templates expect), redirect debt (old URLs that need to keep resolving), and content-quality regressions where agents write. The session walks through each one and the checks that catch them before they reach production.
Want to host an event with Rangle?