RingCentral Replaced the Manual with a Hackathon. Everyone Builds With Codex

A corporate hackathon instead of PowerPoint training

While most corporations spend months drafting guidelines for AI usage, RingCentral opted for the opposite extreme. The Office of the CEO announced an internal “AI-Native Challenge.” Every participant, whether an engineer or a salesperson, was given ChatGPT Work and Codex. The only prompt: build a complete software project from scratch.

OpenAI points to this case as a model for adoption. Though the primary source was blocked, corroborating corporate PR confirms the massive scale of the event involving thousands of employees.

Those who don't build won't understand limits or potential

Forcing non-technical staff to build end-to-end projects carries a deeper logic. People who have never seen code realize they can create things thanks to Codex. More importantly, engineers discover exactly where these tools speed up workflows and where they fall short.

The foundation of this approach is the premise that reading about AI leads nowhere. A genuine understanding of how models hallucinate, or where they save hours of work, only arises when you are forced to drag a functional output across the finish line with them.

The limited transferability of isolated projects

It remains unclear how the company intends to maintain the codebase and security of thousands of “hobby” projects built by salespeople using AI assistants. There is an enormous gap between a disposable script that wins a hackathon and code that can run in production while handling client data.

OpenAI highlights RingCentral to demonstrate rapid deployment, but transitioning from internal toys to standardized tools is the harder, duller, and far more expensive phase.

The question of genuine operational integration

The key signal to watch will be how these isolated experiments translate into actual company operations. It remains to be seen whether thousands of projects will just stay in the drawer as exercises, or if RingCentral can build a process to securely adopt and scale the most successful internal tools.

Lilith's verdict

RingCentral is running an uncontrolled laboratory. Releasing employees into unconstrained AI creation without technical limits makes for great PR, until the accounting department finds a script trying to send production data to a third party.