OpenRouter's AI Agent Doubles Sales Conversion and Saves 600 Hours Monthly

Image: Illustrative image · FoxTPNL · CC BY 4.0 · Source

OpenRouter deployed its AI agent Rasp in its five-person sales team, which halved deal times and more than doubled conversion rates while saving 600 work hours monthly. The AI handles initial client contacts, briefing, CRM updates, and call analysis, enhancing sales efficiency.

OpenRouter has integrated its proprietary AI agent, named Rasp, into the operations of its commercial sales department consisting of five staff members. According to an October 2026 report by iXBT, this agent is powered by the Ori engine and now manages key functions such as qualifying incoming leads, performing first client interactions, preparing briefings, summarizing call transcriptions, and synchronizing with the company's CRM system.

The sales team once burdened by administrative duties, as described by sales division head Michael Haimovitz, now saves about 600 hours each month thanks to Rasp's automation capabilities. Rather than replacing sales managers, the AI agent clears administrative hurdles, allowing personnel to focus more on direct sales activities.

One of the significant impacts has been reducing the handling time per demo call from 103 minutes down to 44. This efficiency gain is due to Rasp automatically researching client companies before meetings, generating draft notes, and updating CRM records with relevant information. As a result, each sales manager can conduct roughly two additional calls daily without compromising preparation quality.

Rasp autonomously sends 93% of initial emails to clients, ensuring responses within a minute of a prospect's inquiry. The remaining 7% involve complex or borderline cases, which the AI flags for human review. These enhancements have led to a 34% acceleration in closing sales cycles and a 2.6-fold increase in sales conversion rates. Moreover, 57% of the current sales pipeline now originates from opportunities enabled by capacities that the team previously lacked.

Financially, Rasp represents a substantial cost reduction compared to its predecessors Ace and Dove, which cost about $800 daily. Rasp performs comparable workloads for roughly $30 per day, including $18 for inference using the GLM 5.2 model. This savings is attributed to Ori's routing layer, which automatically assigns cheaper models for routine subtasks without diminishing quality. OpenRouter anticipates further cost drops, potentially lowering inference expenses to just 50 cents daily by shifting standard tasks to even more affordable models.

Importantly, Rasp does not take final decisions independently. It drafts recommendations, classifications, and compliance checks, but all final actions require human approval, maintaining oversight and quality control.

This implementation highlights how advanced AI-driven task automation can significantly boost commercial team productivity and sales performance while cutting operational costs, positioning OpenRouter as a case study for AI augmentation in sales environments.