Rox

Getting excited for Shriram Sridharan's Lightning Talk at AWS re:Invent - how AI agents are changing the way revenue teams work. If you're attending, catch it at the Partner Experience Pavilion on Monday at 6:30pm. Revenue teams sit on huge amounts of data - CRM, emails, transcripts, product usage, and more, but it’s scattered across systems and stuck in disconnected workflows. That results in missed opportunities and too much time spent on manual, low-leverage work. Shriram will share Rox’s POV on what becomes possible when Revenue Agents run on a unified knowledge graph that combines internal data + external data. With full context, agents handle research, prep, risk monitoring, and outreach so sellers can focus on customers. He’ll also break down how leading companies are already using revenue agents to lift conversion rates, speed up deal cycles, and drive more revenue per rep - without adding headcount. If you want to see how top teams are transforming their revenue orgs with AI, don’t miss it!

Clever Columns has gone through four major evolutions at Rox - each one shaped by what we learned from real usage and what we realized we could push further. What began as a simple single-turn web search is now one of Rox’s flagship features: a multi-agent research system that runs across thousands of accounts with private context, structured tools, and workflow-level orchestration. ➡️ V1: Web Search + LLM The first version was straightforward: execute a web search, pass snippets to an LLM, return a summary. It worked, but it was confined to public data and couldn’t perform multi-step reasoning or touch private context. ➡️ V2: Deep Research with Third-Party APIs In this version we expanded Clever Columns into true deep research capabilities powered by multiple external APIs. To support this, we built: - A Redis-backed rate limiter using token buckets Granular per-provider allowances - Sophisticated retry logic with exponential backoff for transient failures and “bucket full” cases - Queueing to smooth bursty activity when thousands of accounts triggered research at once This version showed us how to orchestrate large-scale research reliably across many providers. ➡️ V3: In-House Deep Research + Rox Graph Integration Third-party deep-research APIs eventually capped what we could do. So we built our own research stack around purpose-built tool interfaces with explicit parameters the agent could reason over to avoid stale or irrelevant data. These tools connected directly to the Rox Graph, unlocking context external APIs couldn’t access (like pipeline status, engaged contacts, deal movement, etc.) To maintain reliability, we introduced evals that measured: - Tool-call accuracy - Multi-step reasoning correctness - Latency and cost across the entire stack - End-to-end answer quality against CCV2 benchmarks This shift gave us full control over depth, accuracy, and consistency. ➡️ CCV4: Workflows, DAGs & Data Warehouse Context The latest version adds richer orchestration and expands context beyond the Rox Graph by integrating with data residing in warehouses. We moved to Directed Acyclic Graph (DAG)-based execution & are using: - Parallel steps to improve throughput and reduce latency - Dynamic branching & conditional logic so the agent can branch into different paths mid-workflow - Flexible definitions: workflow objects are defined as structured objects the map out complex research processes - Broader context by connecting to data warehouses - A new Config Agent converts natural language prompts into structured configurations and workflows, specifying the exact tools and data sources the research agent should use. CCV4 is still only the beginning - more to come! Full eng blog: https://lnkd.in/eX7QJMcd & if this is the kind of system you want to build, come join us: https://lnkd.in/eZAMB2BD

Weekly Drop 92: Featuring personalization in Command! Shout out to Alex Derhacobian for the demo 👇 We built this because Command is where our customers spend most of their day, and agents should naturally reflect their own tone and way of working. We also dropped: Research: Rox can now leverage your CRM data for scaled account research Jerry Zhou Delta Lake Support: Rox is fully warehouse native. No matter where your CRM/product usage data lives (Delta Lake, Snowflake, Big Query, etc), Rox is able to connect to it via delta share to build the knowledge graph powering our agents. Storage is on your end and compute is handled by Rox. Damon Lin Outreach: More ways to kick off a campaign. You can start a campaign from our Discover page, a CSV with only a list of LinkedIn URLs, your existing accounts, and more. Daniel Chen, Ananya Kulshrestha Full release notes: https://lnkd.in/gFeKCZg9

“The conversion rate on Rox-influenced leads is higher than our standard funnel.” That’s what Rho CRO Tommy McNulty told us, and they’ve already driven more than enough deals to more than break even on Rox many times over. Rho was prioritizing higher-converting sales motions. And it gets at one of the questions we hear most often: “How do we know which sales motions are actually working?” Rho answers that by using Rox Plays. Leaders define the motions they want to test, apply them across accounts, and instantly see what’s performing … all while Rox keeps CRM data updated automatically. No toggling, no dashboards, and no extra lift for reps. See how Rox could work for your team: https://lnkd.in/gAuBh7Yf

Calling all sales leaders & AEs - how to make 4 core sales workflows 10x faster 👇 Rox prompt ideas you can start using today. Try free: rox.com/signup

Satya’s “Microsoft will become an agent infrastructure business” vision is playing out in the Global 2000 and it's why CIOs want Rox inside Copilot. After catching up with Scott Guthrie and customers at Ignite, here’s what I learnt CIOs need to navigate the shift from SaaS to Agents: 1️⃣ Vertical agents need deep platform integration, not just a marketplace listing - Marketplace listing is easy but not sufficient - Agents need to integrate with the enterprise data layer - Agents needs to be accessible from all core M365 apps 2️⃣ In-VPC agents, context can’t leave the enterprise - Agents need to run in-VPC - Revenue data has to stay inside enterprise - Metadata management is crucial 3️⃣ Data lineage is still critical - Info-sec needs to understand each data-flow - Need to cataloge data sources and systems feeding context - Pilots can’t be approved without a documented integration roadmap 4️⃣ Future proof architecture - Necessary to support major data warehouses - Adopt agent protocols like NLWeb early - Staying close to Gartner research helps CIOs 5️⃣ It's early days, eng to eng collaboration is key - IT–vendor relationships need to be deeply technical, not transactional - Vendors needs to rapidly refine architecture to IT needs - IT orgs want to actively learn from best agent teams CIOs like Srikanth, Jazz, Saikat, Robi and others aren’t just asking for demos. They’re asking for agentic systems that actually do work, integrate with existing infrastructure, and accelerate frontline execution today. Aligns exactly with the direction Satya outlined in the Dwarkesh Podcast and why we’re partnering with Microsoft. Rox’s revenue agents - already running 25M agents and 1T+ tokens in production - will soon be available directly inside Microsoft Copilot, bringing agentic automation to where Global 2000 teams already work. Read more: https://lnkd.in/guerdesN

Announcing Rox for Microsoft Copilot! Today at Ignite, Rox took the stage in the What’s New in Microsoft Copilot session. Soon you’ll be able to use Rox Revenue Agents directly inside PowerPoint, Word, Excel, and all of M365 - for drafting proposals, building decks, and much more. How it came to be: Tons of customer facing teams spend their day in M365, but the information they need lives across dozens of systems. Rox unifies all of that scattered data by being warehouse-native. Rox agents pull from the centralized source of truth enterprises already rely on and gives you an intelligent agent that can surface context and handle the heavy lifting of daily workflows. Now that power is available directly inside Copilot, where work already happens. Read more about the partnership: https://lnkd.in/gwcyBCnS

Rox Release 91: making everyday tasks quicker and smarter ➡️ Command - text selection bar: When you select text inside Command, you now have a helpful selection bar. Quickly reference specific parts to dig deeper or add to your notes. The "Saved prompts tip" also allows you to paste a text snippet inside Command and get a timely suggestion to save the prompt in their library of saved prompts. Gustav Ekerot ➡️ Clever Columns have a way to configure Model: You can now choose the right balance of intelligence and cost when running a clever column. Taeuk Kang ➡️ Outreach Improvements: Daniel Chen a. View all of your tasks across campaigns in one place b. View and manage email templates across your campaigns c. "Clone Campaigns" - copy and reuse campaign configurations in one click Release notes: https://lnkd.in/gFeKCZg9

“If your AEs are not using Rox, you will become obsolete in a few years.” - Syed Jafri, Snorkel AI Snorkel’s revenue team is proving what many GTM orgs are starting to realize: AI-driven consolidation > tool sprawl. With Rox, their AEs are now: + Replacing toggling multiple legacy tools with one AI-first workflow + Using clever columns to speed up research and spend more time with customers + Leveraging Plays to scale repeatable motions and execute fast = Stronger AE performance and more high-quality leads. See how consolidating tools can 10x your team’s efficiency: https://lnkd.in/gAuBh7Yf