Rox goes “all in” on OpenAI

# Rox goes “all in” on OpenAI Source: [https://openai.com/index/rox/](https://openai.com/index/rox/) OpenAIBy combining commercial experience and deep LLM expertise with OpenAI’s models, Rox makes every seller a top 1% seller\. [Rox⁠\(opens in a new window\)](https://www.rox.com/)is reshaping how sales teams manage and grow revenue\. Their platform combines fragmented data into a unified system of record and delivers insights through always\-on AI agents swarms powered by OpenAI\. The team behind Rox comes equipped with deep expertise in AI and data integration\. Avanika Narayan, co\-founder and AI Lead at Rox, drew from her PhD research applying LLMs to enterprise tasks like data wrangling and workflow automation\. “We saw a rich ecosystem of humans interacting and engaging in revenue teams but realized they needed to be grounded in data,” she shared\. The result was a three\-tiered vision: a system of record, system of workflow, and AI\-powered assistance—brought to life using OpenAI’s API\. For Rox, launching in 2024 wasn’t just about timing—it was about opportunity\. The rise of usage\-based revenue models and the shift toward AI\-powered sales operations revealed gaps in many revenue teams’ systems\. OpenAI’s API provided Rox the flexibility to handle everything from data cleaning to long workflows to nuanced sales outreach for go\-to\-market strategies\. “Using OpenAI’s API allowed us to create something novel and inject creativity into a GTM workflow,” notes Amol Singh, one of Rox’s founding AI engineers\. The platform’s design rests on three layers: - **Data layer:**Rox uses smaller models like GPT‑4o mini to unify fragmented data from various warehouses, making it structured and easily retrievable\. - **Intelligence layer:**Mid\-tier models perform complex reasoning and prioritize actions, collaborating with sales reps to augment their tasks\. - **Interaction layer:**Advanced models like GPT‑4o and OpenAI’s Realtime API generate emails, manage LinkedIn outreach, and prepare voice\-enabled meeting briefs\. At the core is the “Rox Agent Swarm,” a fleet of always\-on AI agents assigned to each account\. While reps are offline, the agents monitor accounts; during work hours, they make reps 50% more productive by handling repetitive tasks and surfacing actionable insights\. ##### Performance Comparison on Sales Rep\. Tasks 82%12%6%96%3%1%94%4%2%Research and Outbound TasksWin Rate \(%\) Rox’s journey to launch was shaped by early experimentation and customer feedback\. Initially, the team explored an open\-ended chat interface but found it didn’t align with sellers’ needs; instead, “sellers needed a solution tailored to their specific workflow, with outputs that were more creative and unique,” says Alex Derhacobian, one of Rox’s founding AI engineers\. The team pivoted to an end\-to\-end configurable platform where each agent adapts to a rep’s specific accounts and workflows\. This approach, informed by partnerships with companies like Ramp, Couchbase and Confluent, also scales the expertise of top\-performing sellers across teams by building their best practices into Rox's platform\. The team credits their iterative approach—shipping updates daily—with accelerating development\. “Traditionally, this would take multiple data engineers and a whole team\. We did it with two people because of OpenAI’s models and how much easier it is to build,” says Singh\. Sales teams using Rox have reported substantial improvements: - **8\+ hours saved weekly:**Reps are reclaiming time previously spent on administrative tasks, allowing them to focus on higher\-value client interactions\. - **35% increase in customer engagement:**Agents help reps stay responsive to key account changes and engage at critical moments\. - **2x ROI on sales pipeline:**Beta clients saw a 2x increase in sales\-accepted pipeline, leading many to scale Rox’s use across their teams and increase the number of accounts managed on the platform\. Rox has gained rapid traction among enterprise clients, growing from zero to 25 accounts in seven months\.  Rox’s platform, built on OpenAI, pulls from publicly available information\. OpenAI’s models are central to Rox’s approach, enabling flexible AI solutions across tasks\. OpenAI “has a full library of models that you can choose from, with different capabilities and different prices,” says Narayan\. This flexibility allows Rox to deploy cost\-effective models for data processing while reserving higher\-tier models for complex reasoning tasks\. Rox’s co\-founders shared three key lessons for building effective AI\-based solutions: 1. **Start with a strong data layer:**“The most important thing is having your data layer set up well and the context management problem solved,” shares Derhacobian\. Indexing semi\-structured and unstructured data ensures that insights are actionable\. 2. **Focus on Applied AI:**Rox chose to work with OpenAI rather than building their own models\. “We’re not here to research or create base models\. We work with OpenAI to deliver amazing applications,” says Rox’s co\-founder and CEO Ishan Mukherjee\. 3. **Iterate quickly:**Modern applied AI allows for a faster pace\. “We pretty much ship daily,” says Damon Lin, one of Rox’s founding AI engineers, crediting their success to continuous iteration\. Rox envisions a future where revenue teams will start and end their days in the platform, fully equipped with the insights and support needed to engage customers effectively\. The team plans to expand Rox’s multimodal capabilities and long\-horizon task assistance\. A new voice\-enabled feature, powered by OpenAI’s Realtime API, is already helping sellers prepare for meetings with detailed, real\-time briefs\. With OpenAI, Rox will continue to enable human\-led sales functions that maximize GTM revenue potential\. ## Keep reading