Strategic Expansion into AI Infrastructure

Ramp, a corporate expense management platform, has launched 'Router,' an API-based model routing service. This move signals a strategic pivot to capture the growing AI inference market while deepening the company's value proposition for existing enterprise clients. By building on its own internal infrastructure—which the company has utilized for three years—Ramp is positioning itself as a central hub for managing AI token spend and model performance.

Core Functionality and Routing Strategies

Router provides a unified interface for accessing models from major providers, including OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai. The platform distinguishes itself through specific routing strategies designed to optimize cost and performance:

• Benchmark-based routing: Allows users to route queries based on up to three user-specified performance benchmarks.
• Complexity-based routing: Enables users to direct difficult tasks to high-performance, expensive models while routing simpler queries to more cost-effective alternatives.
• Testing and Fallback: Provides a dashboard for monitoring token spend, latency, and fallback attempts, allowing teams to test new models without significant refactoring.

Data Privacy and Business Model

While the service is currently free to use through the end of 2026 (excluding inference costs), Ramp has implemented a specific data retention policy. By default, the platform records model inputs, outputs, and tool calls for one year. However, the company includes an opt-out mechanism and commits to stripping personally identifiable information (PII) before using data to improve their product. This launch serves as a dual-purpose tool: it provides a technical solution for AI-heavy enterprises while creating a new entry point for Ramp to cross-sell its core expense management and token monitoring products.