Building Multi-Agent AI Systems with Vanchhit K

Last call. This Friday, April 17 at 1 PM ET, I am going live on RSAC's Cybersecurity Learning Program. The topic: "Intelligent Agents: Let's Build AI That Can Think Together." I will walk through how to build real AI systems where multiple agents work together like a team. Not slides about what agents could do someday. Actual systems. How they coordinate. How they fail. How you secure them. This falls under RSAC's April theme of Securing AI Agents, and I could not be more excited to talk about it. It is free. It is live. And registration is still open. Register here: https://lnkd.in/egeUMEAQ See you Friday. #RSAC #AIAgents #Cybersecurity #AgenticAI #SecuringAI

Watch this quick demo to see how we build an "NYC collision hotspots" analysis in CARTO Workflows and instantly package it as an MCP Tool for an AI Agent. This allows anyone to ask "Where are the collision hotspots?" and get an immediate, visualized answer on the map, turning a complex workflow into a simple conversation. See how the CARTO MCP Server makes it possible:https://hubs.ly/Q049fc0Y0

"If your competitors start doing Level 3 AI and you don't, your projects, your quotes, and your timelines are going to look vastly different from what someone else is promising." Well said by Srikrishnan Ganesan - one of many moments in last week's webinar that made me realize how many companies are already behind! We covered a lot of ground in "Beyond the Mandate" - AI adoption patterns, outcomes-based pricing, and the two diverging strategies I'm seeing across SIs and proserv teams right now. A huge thanks to Sri for bringing his ability to see around corners - and into a whole new market - to a pretty wild moment for the industry. Recording in the comments if you missed it.

Tracking agentic interactions (n=3,959) in an Edge Delta pre-prod staging environment, there are interesting results for observability use cases when the incident count decreases: large models show some reduced consistency, but small models degrade considerably. What this finding means for Observability in the AI Era: - Noisy prod environments: small models can still hold their own - Small/quiet prod environments: small models lose value fast, go larger This will probably change quickly (as things do these days), but I thought I'd share.

Search engines and AI systems don’t interpret industries the way humans do. Instead of reading pages individually, they map relationships between entities like: • companies • products • installers • applications • locations These relationships form what’s known as a knowledge graph. This diagram illustrates how the artificial turf industry can be understood through structured relationships. See the full model: https://lnkd.in/eNr6JYWr

Specialized agents working together can solve problems a single agent simply can't. They also unlock a new class of operational headache. What gets harder as the system grows is knowing what's actually performing, and what to change when something quietly starts to drift. Agent Graphs in AI Configs is now generally available. Monitoring lives directly on the graph so you can see latency, invocations, and tool calls per node at a glance. When you spot an issue, the configuration for that node is already in LaunchDarkly. Update it, roll it out gradually, or set a fallback without touching a deployment Stop hotfixing at 2 am - https://lnkd.in/ggubWy8D

exciting to be on the front line solving some of the newest problems we're seeing in the AI space today. if this is relevant to a challenge you're working to fix today, let's chat! 🙏

In today's Business Briefing Live, Phil Wilson covers the reasoning behind the top GHL performance measurement approach using TITON Control Self-Assessments (CSAs) to monitor the Provisioning Set-Up and Deployment of GHL features, functions and services. The objective is to institutionalize the Signal --> Action --> Proof as a closed loop TITON process which is the hallmark of business excellence in the AI Automation Agency (AAA) proof-driven operations ecosystem. 25 Min. 18 Sec.

We're proud to partner with Kenley to power the workflows behind their AI agents for consultants: from the 30-minute agent runs that produce client-ready PowerPoints and Word docs, to the background embedding pipelines that make fast multimodal retrieval possible. 💙

To make AI actually useful—we have to constrain these probabilistic models with deterministic rules. Look at how the language of compliance, the structure of time, and the frameworks of authorization running your favorite AI tools: https://lnkd.in/g9zjAg9w