monday.com

Halloween 2026: your AI agent has an AI agent

Organised chaos? We know a thing or two about that 😅 Coach Ben Durham takes us for a lap around the base - showing exactly how the Bonds Flying Roos get things done with monday.com 🦘

Cheaper than a consultant. Similar accuracy

no pressure but somewhere a product manager just marked “flying car” as Done.

new phone who dis? Whatever you’re carrying on October 23, your monday app will fit it (iPhone duo included). We got you 🫡

Olympic champion. World record holder. Proud owner of a herd of cows. 🐄 We’re thrilled to welcome Lindsey Vonn to the Elevate by monday.com stage- swipe to meet her, then check out who else is joining us. See the full speaker lineup 👉 https://lnkd.in/daeSN35m

Yesterday, I had the pleasure of speaking at LangChain Interrupt ’26 London about how we at monday.com AI engineering are using evals to make our AI agents smarter. I've shared the monday.com vision shift to Doing the Work for our customers and how the real difference for them is not another AI feature. it's Reliability, Trust and Confidence in AI results. That's why we built a hybrid evaluation architecture connecting what happens before release with what we learn from real production usage. Two pipelines. Two questions. Offline EVALS runs pre-production against datasets with expected outcomes and is the gate for release, answering the question: Can we ship? Online evals analyze real production traces and uncover failures we couldn't anticipate. Real users bring new scenarios, unexpected behaviors, and data we've never tested. It answers the question: Is it working for real users? Focusing on three key signals: • Task Completion - Did the agent achieve the user's goal? • Trajectory Health - Did it use the right tools and execute efficiently? • User Frustration - Did the user struggle, repeat requests, or correct the agent? A high completion rate can hide slow responses or wrong tool selection. A healthy trajectory can still miss the goal. The lesson: don't rely on a single metric, but don't measure everything either. Focus on what is significant for you! We also break down the conversations by domain, family and use case. It helps us to understand our strengths and opportunities and helps us prioritize our work. Online & Offline evals must come together. Measuring quality, Latency and Cost. One cannot replace the other. When online evals uncover recurring failures, we can turn them into offline regression tests, verify fixes before release, and measure whether production actually improves. Closing the Loop: Production → Detect failures → Offline tests → Fix → Verify in production. I've also shared how we as builders are getting trust in the metrics and scores. LLM judges aren't the ground truth. We build confidence through human reviews, comparisons between different judges, repeated scoring for consistency, and real user feedback. The real value of evals isn't knowing your agent's score. It's knowing what to improve next, why it matters, and whether your changes actually made an impact. Thanks to Harrison Chase and the entire LangChain team for having me at Interrupt London and to everyone who joined the session!

35 million MCP calls to monday.com last month. Nov: 255,911 Jan: 668,534 Feb: 2,163,816 Mar: 5,496,618 May: 9,665,100 Jul: 17,958,901 Sep: 35,511,411 remember when people said MCP was dead? inspired by Zeno Rocha 's post

Breaking news: the Elevate keynote is going LIVE on LinkedIn 🚨 Our AI direction, our roadmap, and a first look at what’s next, straight from the people building it. Sign up now (because we all know the recap never hits like the real thing) 👇

Last year at Elevate by monday.com, I managed to squeeze in 17 customer meetings. This year, I'm going for 18. 😎 I'll be heading to NYC for Elevate26 on October 28–29, where I'll be speaking alongside Omer Fast! Our session, "A day in the life of an AI-empowered manager," will show how monday vibe and AI agents can help managers deal with the everyday chaos of tasks, meetings, and people. But the part I'm most excited about happens offstage. Meeting hundreds of customers, hearing what they love (and hate), and discovering all the creative ways they're using our products. This year, I'll have even more to talk about: monday vibe, AI agents, and a secret new product we've been working on. 👀 So if you're coming to Elevate, I'd love to meet you! Come show me what you've built, share your biggest frustrations, or just say hi.