Cult.fit expands Codex from engineering to operations
India’s technology-enabled fitness company cult.fit is using OpenAI’s Codex across engineering, internal operations and customer support. The company operates 708 fitness centres in 77 cities and serves nearly one million paid members. It reports that non-engineering teams have built more than 20 internal applications in recent months, that an internally rebuilt dashboard capability saves more than $100,000 a year, and that customer support costs have fallen by more than 50%.
Cult.fit first standardised its software development lifecycle on Codex after developers had been using a mixture of coding agents and assistants. The differing tools brought different repository conventions, instructions, skills and workflows, limiting knowledge sharing and reuse. Codex became the common platform for repository structures, testing, code review and feedback loops.
A shared development foundation
Santhosh Kumar, cult.fit’s CTO, said the company could build its software development lifecycle around a single platform once it standardised on Codex. The organisation then extended the tool into specialised engineering processes. Its Site Reliability Engineering agent aggregates information from observability platforms and logs, investigates incidents and can, in many cases, create a proposed-fix pull request for an engineer to review.
Security agents help identify and address vulnerabilities, while other on-call agents examine recurring production issues. Cult.fit is also moving towards spec-driven development, in which a specification guides work through the lifecycle to a pull request. Kumar estimated that a social feed for the cult.fit app, previously expected to take more than a year, was built and shipped end to end in two months using the new workflow.
The company’s approach follows the enterprise Codex implementation work represented by enterprise Codex implementation work as organisations seek common foundations for deploying coding agents across teams. At cult.fit, standardisation is intended to make development practices more repeatable rather than leaving individual groups to maintain separate conventions.
From analyst queues to self-service tools
Cult.fit made company data easier for Codex to work with so that functions could find answers and carry out routine analysis themselves. Smaller data requests had previously passed through a central analyst team. Kumar described the change as the company’s largest step towards data democratisation.
Teams are also creating workflows for their own needs. HR built a Codex workflow for performance reviews, scheduling, feedback collection and assembling the resulting information. When the cost of internal dashboard software rose sharply, one engineer rebuilt its functionality with Codex in a weekend; cult.fit now runs that capability on its own application.
The company paired access with adoption support. It held a hackathon, brought management together for a one-day building session, runs engineering office hours twice a week, and embeds engineers in selected functions for technical and domain support.
AI in the member and support experience
Alongside Codex, cult.fit uses the OpenAI API for its AI Trainer. More than 45% of its members are first-time gym goers, and around 20% of gym users currently use the product. The latest version uses GPT Realtime through an API integration for a voice-first experience, providing coaching cues and supporting workout logging through two-way voice interaction.
In customer support, AI agents assess requests to waive penalties for missed group fitness classes. The agents can check signals including local weather and class attendance patterns before deciding whether to waive a penalty. Cult.fit integrated the agents with its existing Sprinklr support platform and backend systems, shifting a predominantly human-led chat process towards an AI-first workflow.
The practical implication is that organisations seeking similar results need to pair a common technical platform with accessible data, clear review points and support for teams turning operational needs into production workflows.