We spent 11x more tokens on reasoning and got 0% better answers on revenue data ... until we added a knowledge graph. TLDR: our latest research shows that improving data representation increases agent retrieval accuracy more than upgrading the model. We ran 3,100 runs across 8 models answering revenue questions, like deal amounts, contacts, and identifying customer champions. We compared two ways of storing the data: 1. a normal database (SQL) VS. 2. a knowledge graph (relationships pre-mapped) Frontier models hit 8.9% accuracy on the questions using SQL over a relational schema. Cranking Claude Opus 4.8’s reasoning effort from minimum → maximum accuracy did not help. However, swap raw Salesforce data for a knowledge graph built on lakehouses like @databricks, @Snowflake, @googlecloud's Big Query, or @Azure Data Fabric ... and accuracy jumps from 8.9% to 99.9% - even using a 27B open-weight model at 1/20th the cost. This research shows throwing more compute at your agent cannot fix bad data structure. And is proof a revenue-specific knowledge graph is key to making revenue agents work at scale.

@AnthropicAI and @OpenAI want your product living inside Claude or ChatGPT. @Google wants the model to redraw your UI from scratch. Both are wrong about who should own the pixels, since it diminishes beautiful product experiences to basic text. We measured a 3rd option: up to 100x cheaper, and your product stays whole. → Introducing Rox Tether: An alternative to MCP Apps, A2UI, and ChatGPT Apps. TLDR: the product manages its own pixels, and the agent operates it via reference. The results: - Up to 100x token reduction - No new protocol - No browser API needed - No cooperation from the chat host Breakdown below:

Introducing ask-web: Rox’s in-house web search agent. ask-web sits on the cost-per-accuracy pareto frontier of the hyper-parameter grid when compared to frontier labs and commercial search agent providers. The agent delivers 91.3% accuracy at 1.03 cents per query on real production prompts. It has been running in production for more than 6 months with continuous evals. Inference partners: @togethercompute, @baseten, @modal Commercial Search vendors benchmarked: @perplexity_ai, @ExaAILabs, @p0. Frontier Search vendors benchmarked: @OpenAI, @AnthropicAI Exa, OpenAI and Anthropic excel on accuracy. Parallel and Perplexity are cost-efficient. Here’s the breakdown:

Introducing Rox Governance, the industry’s first unified governance layer. 2.5 years ago, we bet on two waves: enterprises moving customer data from CRMs into warehouses, and AI transforming every revenue org. Both are in full force. But access controls stayed behind in the CRM while the data moved. That's kept enterprises stuck at "no" on revenue agents. Rox Governance fixes that. One set of rules for every system your team and agents interact with. Set the rules once. The CRO sees everything, a director sees their team, a rep sees their book. All systems go →

Revenue agents are the main course. Data enrichment is on the house. B2B orgs spend $3.1B a year for employees to find contact and company data. Now agents need it too. So we made a call: enrichment is a cost of doing agentic work, not a product to mark up. It's infrastructure that revenue agents need to run. Today, we’re opening up our in-house data collection and enrichment infrastructure to all. No extra contracts. No credits to count. Just revenue, served.

Agents now Run Revenue, End-to-End. Rox Autopilot is here 🚀 The world’s largest enterprises now run their critical revenue systems on Autopilot. Coding and support agents had their moment. Now it’s time for Revenue Agents No SaaS. No enrichment. No CRM. Just an Agent. We’re offering 2× free agent actions for the next 30 days. Start Autopilot at Thank you to @sequoia, @generalcatalyst, @googleventures, @marcbhargava, @vedantsuri, @davemuni, @MongoDB, @cloudsoftware, @togethercompute, @Microsoft, @databricks, and all of our other customers & partners!