cult.fit Codex has become the central platform behind cult.fit’s push to simplify engineering, democratize data access and scale AI-driven member experiences across its operations. The India-based fitness chain, which runs 708 centers in 77 cities and serves nearly one million paid members, unified development workflows around Codex to keep processes consistent as the business expanded.
Before the transition, developers at cult.fit used a mixture of coding agents and assistants with inconsistent repository conventions, agent instructions and testing practices. That fragmentation made collaboration harder and increased the cost of reusing work across teams. By standardizing on Codex, the company established common rules for repository structure, testing, code review and feedback loops. CTO Santhosh Kumar said the shift let the company “build our software development lifecycle around a single platform.” The move has enabled teams to reuse patterns and components rather than reinventing them for each project.
cult.fit extended Codex into focused engineering workflows. Site Reliability Engineering agents now pull observability data and logs to investigate incidents and can even generate pull requests with proposed fixes for engineers to review. Security agents identify vulnerabilities, while on-call agents help probe recurring production issues. The company is also moving toward a spec-driven development model, where a formal specification guides work through to a pull request, creating a repeatable path from requirement to working code.
One tangible illustration of the platform’s impact came from a recent social feed feature for the cult.fit app. Kumar estimates the team previously might have taken more than a year to build this capability; using the Codex-centered workflow, the company designed, developed and shipped the feature end to end in two months.
A key enabler for these engineering gains was preparing company data so Codex could access and analyze it directly. That removed a long-standing bottleneck that forced even small queries to route through a central analyst team. With data available to the platform, functional teams can run routine analyses and find answers without waiting on specialized analysts. As a result, more than 20 internal applications have been developed by non-engineering staff in recent months, covering needs from HR to operations.
Non-engineering groups have used Codex to automate workflows and build internal tools. For example, HR created a Codex workflow to manage performance reviews, scheduling and feedback aggregation. To accelerate adoption, cult.fit held a large hackathon and ran a one-day session where managers from across functions built tools that mattered to their teams. Engineering supports these efforts through twice-weekly office hours and by embedding engineers in selected functions to provide technical and domain expertise.
The platform has also delivered immediate cost savings. When the price of an internal dashboard product rose, one engineer recreated equivalent functionality with Codex over a weekend, avoiding more than $100,000 in annual software costs. That example underscores how platform-level standardization can translate into both speed and direct financial benefit.
Beyond internal tooling and engineering, cult.fit is applying the OpenAI API to member-facing products. The AI Trainer, originally developed several years ago, provides guided, personalized workouts for members who do not use a personal trainer; this is an important product given that more than 45% of members are first-time gym users. Roughly 20% of cult.fit’s gym users currently use the AI Trainer. Recent model improvements have made the trainer more conversational and interactive: the latest version introduces a voice-first experience, integrated with GPT Realtime via the OpenAI API, offering coaching cues and supporting workout logging through two-way voice interaction.
AI has also shifted customer support workflows. Previously many cases required human review; now AI agents assess requests such as penalty-waiver appeals for missed group classes by checking contextual signals like local weather and class attendance patterns before deciding whether to waive a fee. These agents operate inside the company’s Sprinklr support platform and are connected to backend systems, moving support toward an AI-first model and reducing manual workload.
cult.fit reports concrete outcomes from its Codex rollout and OpenAI API integrations: more than 20 internal applications built by non-engineering teams in recent months, over $100,000 in annual software savings from rebuilding an internal capability, and a greater than 50% reduction in customer support costs. Looking ahead, the company plans to embed AI more deeply across operations and member experiences. Kumar describes the operational ambition as an “autopilot operations org,” where AI handles more support tasks for frontline teams, and a member-facing goal as a “trainer in your pocket” that scales personalized guidance beyond what human trainers alone can deliver.
As employees continue to build applications, analyze information and automate workflows, cult.fit intends to evolve from isolated AI projects to an organizational model in which AI is integral to how work gets done. The Codex standardization effort appears to be the structural change enabling that shift, aligning engineering practices, speeding delivery and broadening access to AI capabilities across the company.
Source: Read the original source
Leave a Reply