Weather-based automation examples

The best way to break down all the hype around AI agents 🤖 and really understand them is to get your hands dirty and build one. An AI agent, as opposed to automation, reasons with an LLM, remembers context and uses tools to act. I recently built an AI agent as I want to learn more about art every day 🎨. Thought I would share the process here for anyone who hasn't built an agent yet. You don't need to be technical to do this. After you decide what you want your first agent to do, here is how you can build it in 5 steps: 1️⃣ 𝗦𝗶𝗴𝗻 𝘂𝗽 𝗳𝗼𝗿 𝗮 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝘁𝗼𝗼𝗹 𝗹𝗶𝗸𝗲 n8n Create a new workflow and drop in the AI Agent node. 2️⃣ 𝗔𝗱𝗱 𝗮𝗻 𝗔𝗜 '𝗯𝗿𝗮𝗶𝗻' Every agent needs intelligence, usually from an LLM like OpenAI. Sign up for OpenAI API access and in n8n, add the “OpenAI Model” node and choose the model. 3️⃣ 𝗚𝗶𝘃𝗲 𝘆𝗼𝘂𝗿 𝗮𝗴𝗲𝗻𝘁 𝗺𝗲𝗺𝗼𝗿𝘆 + 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 Add the 'Simple Memory' node in n8n and your data source. In my example, I want my agent to send an email to a mailing list every day with new art inspiration. For my data source, I have a Google sheet with the target emails. 4️⃣ 𝗖𝗼𝗻𝗻𝗲𝗰𝘁 𝗲𝘅𝘁𝗲𝗿𝗻𝗮𝗹 𝗔𝗣𝗜𝘀 This is the most involved step, depending on your agent's logic. With n8n, you can easily call APIs for weather, stock prices, images, etc. Just add an HTTP Request node or use a pre-built n8n integration (there are many!). To make my agent a little more complicated, I made the art piece it chooses for the person every day dependent on their local weather. I connected a weather API, a Harvard Art API and the MediaWiki API from Wikimedia commons to fetch an image for the art piece chosen for every user every day. And lastly, an API to send emails. 5️⃣ 𝗪𝗿𝗶𝘁𝗲 𝘁𝗵𝗲 𝗹𝗼𝗴𝗶𝗰 𝗶𝗻 𝗽𝗹𝗮𝗶𝗻 𝗘𝗻𝗴𝗹𝗶𝘀𝗵 The beauty of AI agents is that you no longer need code to tell them what to do. Just write a prompt to describe how you want your agent to act, which APIs you want it to use for what purpose and give it step-by-step instructions. Pro tip: use ChatGPT to make your prompt more detailed and specific. (i.e. use AI to write the prompt for an AI agent!) In my case, I told it to use OpenAI to find a new acclaimed work of art every day for every user → get an image for it → write a personalized description for every user about the artist, the work and how it's viewed in art history → then, put it all together in an email and → send it. ✅ And voilà, your agent is done! Now test it in n8n, tweak as needed and when you are happy with the results, set the trigger for when the agent should run, and activate it. ✨ In under an hour, you can build your first AI agent. It’s the best way to move from 𝘈𝘐 𝘩𝘺𝘱𝘦 → 𝘈𝘐 𝘩𝘢𝘯𝘥𝘴-𝘰𝘯.

Crops have a temperature story—thermal #drones tell it. #Thermal + #AI + #drones are transforming the way we manage farms. By detecting subtle changes in crop temperature, they reveal problems before they become visible, helping farmers take action early. --Irrigation Optimization:  • Thermal sensors measure canopy temperature.  • Canopy Temperature Difference     (CTD = T_canopy – T_air) highlights water-stressed zones.  • AI generates variable-rate irrigation maps to save water     and improve efficiency. --Pest & Disease Detection:  • Leaf temperature anomalies     (ΔT = T_leaf_anomaly – T_healthy_leaf)     indicate early stress from pests or disease.  • AI classifies hotspots, enabling precise intervention and     reducing chemical usage. --Yield Prediction:  • Thermal patterns combined with NDVI and soil data     allow prediction of low-yield zones.  • Formula example:     Predicted Yield = f(CTD, NDVI, Soil Moisture, Weather) --24/7 Monitoring:  • Works day and night, under shadows and clouds.  • Temporal stress trends     (Stress Trend = (CTD_t2 – CTD_t1) / (t2 – t1))     help track crop health over time.  • Example: A 0.5°C rise in canopy temperature in     maize → targeted irrigation → prevents 10–15% yield loss. I’ll continue sharing posts on how thermal and multispectral cameras are being applied beyond agriculture, from infrastructure inspection to environmental monitoring. Stay tuned! Note: the image is generated by AI

I experimented with a LangGraph-based multi-agent AI system for intelligent travel planning and expense forecasting, capable of ingesting user goals and generating real-time, end-to-end itineraries with cost estimates, weather data, and activity recommendations. This system demonstrates agent orchestration over stateful message-passing, with each agent encapsulating domain-specific logic and tools. Agents operate asynchronously and pass control through a Coordinator Node that enforces deterministic execution order and fallback logic when needed. System Capabilities: - Real-time weather forecasting using Gemini Flash + DuckDuckGo search - Destination activity mining via search-based scraping - Hotel cost estimation using static heuristics and dynamic duration inputs - Currency conversion with live exchange rate retrieval - Expense aggregation engine with per-day and total cost breakdowns - Itinerary synthesis using contextual reasoning across date windows, weather conditions, and travel themes Summary generation via Gemini LLM Agent Topology: - CoordinatorAgent — DAG controller, manages traversal and ensures dependency resolution - AttractionAgent — parses user interests and maps to top results via search APIs - WeatherAgent — handles current_weather() and forecast(days) routines - HotelAgent — dynamic pricing model (flat or tiered) x trip duration - CurrencyAgent — pulls FX data, handles conversions to native units - BudgetAgent — executes reduce(sum) across line-item costs - ItineraryAgent — plans N-day schedule using weighted scoring from activity/time/weather match - SummaryAgent — final layer invoking Gemini for formatted itinerary report Tech Stack: LangGraph: StateGraph-based agent orchestration Google Gemini Flash: Contextual generation + low-latency semantic reasoning DuckDuckGo Search API: Open domain search interface Python: Modular OOP with TypedDict state containers for agent interoperability Execution Traceability: Internal logging per node traversal + state snapshots for debugging Design Rationale: The goal was to build a deterministic and inspectable agentic system with real-world applicability and plug-and-play agent extensibility. LangGraph’s node-based topology allows for tracing, recovery, and fallback handling, making the design production-compatible while maintaining full modularity. The result is a functional and reproducible architecture for real-time decision support, with clear separation of concerns and minimal coupling between agents. GitHub: https://lnkd.in/g3WiMifG

🌦️ 𝐏𝐨𝐰𝐞𝐫 𝐁𝐈 𝐱 𝐀𝐏𝐈𝐬: 𝐀 𝐏𝐞𝐫𝐟𝐞𝐜𝐭 𝐌𝐚𝐭𝐜𝐡 𝐟𝐨𝐫 𝐃𝐲𝐧𝐚𝐦𝐢𝐜 𝐑𝐞𝐩𝐨𝐫𝐭𝐢𝐧𝐠 While building analytical dashboards, I ran into a challenge: I wanted to analyze how weather impacts sales, attendance, and operations. Internal data didn’t exist, so I had to pull live data from outside sources. I explored  𝐎𝐩𝐞𝐧𝐖𝐞𝐚𝐭𝐡𝐞𝐫𝐌𝐚𝐩 and 𝐖𝐞𝐚𝐭𝐡𝐞𝐫𝐀𝐏𝐈, two free APIs offering JSON data for temperature, humidity, wind, precipitation, and forecasts. After creating accounts and generating API keys, I connected them in 𝐏𝐨𝐰𝐞𝐫 𝐐𝐮𝐞𝐫𝐲 using dynamic parameters and 𝘞𝘦𝘣.𝘊𝘰𝘯𝘵𝘦𝘯𝘵𝘴().| The journey wasn’t without hurdles: ❌ Parsing nested JSON structures ❌ Handling null values and missing fields ❌ Ensuring historical data stayed intact while refreshing new records After debugging and iterating, I built a fully automated workflow that: ✅ Pulls live weather data dynamically ✅ Cleans and structures it for analysis ✅ Combines current and forecast data into a single table ✅ Uses incremental refresh to maintain historical records The outcome? A self-updating weather dataset inside Power BI, ready to be merged with business metrics, turning dashboards from static reports into actionable insights. This project wasn’t just about weather. It taught me how APIs and Power BI can expand analytics beyond internal data, and how automation and proper error handling can make complex workflows scalable. #PowerBI #DataAnalytics #PowerQuery #APIs #BusinessIntelligence #Automation

🚀 Agentic AI in GDP/GMP Warehouses: Real Use Case or just a Wild Idea?   This week, I had an interesting conversation with a customer about temperature monitoring and mapping across several GDP- and GMP-compliant warehouses in Germany. As usual, we discussed the basics: Alert templates, excursion handling, standard operational workflows. ❄️📦 During the last minutes of the call I had to sharpen my ears: The client had special alert settings active only for the warehouse manager, focused on very specific morning temperature bands, completely untied from those required for regulatory requirements. When I asked why, the explanation was very interesting: 🌡️ The customer uses morning temperature data and manually matches these with the expected weather data to pre-adjust the warehouse climate control system. Based on the customer’s experience with thermal inertia of the warehouse during summer, the system gets “pre‑configured” early in the day to avoid temperature excursions later at known hotspot zones. So essentially: Human expertise + historical temperature data + expected weather → manual climate optimization. 💡 That sparked a thought: What if we took this to the next level? 🤖 Agentic AI for predictive warehouse climate control: ☁️ Cloud-based, real-time warehouse temperature monitoring 🌦️ Live and forecasted weather data 🔗 An Agentic AI agent integrated into the monitoring platform 🏭 Direct API or Modbus connection to the BMS or HVAC system The AI could autonomously: - Predict hotspots before they emerge - Adjust climate systems proactively - Prevent excursions 24/7 - Optimize energy usage   🔍 Question to the GDP/GMP Community managing pharmaceutical warehouses: Is this a real operational need? Would love to hear your insights and experiences!

When the Storm Hits - AI’s Next Frontier Hurricane Melissa. Iberia heat shutdowns. 600+ extreme weather events in 2024 alone. For utility crews, these moments mean 4am alerts, Walmart parking lot staging areas with paper instructions, and critical infrastructure sitting in water while coordination happens manually. IFS Nexus Black just previewed something different at Industry X here in NYC: AI-powered emergency response that cuts power restoration times by 40%. Here’s what changes: Before Storm Landfall → Personalized crew deployment notifications with predicted damage assessment and required equipment lists. No more generic high-level alerts causing duplication and wasted resources. During Crisis → AI command centers analyzing storm patterns, inventory across utility territories, and mutual aid opportunities in real-time. Automated coordination that previously required armies of people making manual calls. The Breakthrough → Long-range planning AI capabilities that coordinate multiple tasks simultaneously. Not just answering questions or solving math problems, but orchestrating complex, multi-variable responses. SoCalGas Head of Emergency Response Brian Burdette, MAOM, CSM (who manages disaster response where January wildfires hit): “This could help us cut restoration times by getting crews to the right places with the right equipment before the event even hits. That means hospitals stay powered and families return to warm homes faster.” The technology leap: From reasoning models (March 2024) to planning-capable AI that handles the “worst call” crisis managers describe—turning single points of failure into distributed intelligence networks. Tony Alloway from BGIS nailed the operational reality: “AI isn’t going to replace you. Somebody who knows AI will replace you.” What’s Your Edge? Identify your organization’s highest-stakes coordination challenge—where minutes matter and manual planning creates bottlenecks. Start with a 2-day rapid scoping workshop (not a 12-month requirements doc). Ship a working prototype in weeks, not years. The Enterprise Edge Mark Moffat Matt Kempson Oliver Pilgerstorfer Christian Pedersen Adam Gillbe Andy Oliver Kriti Sharma #powerofpartnerships #industrialAI #proofpoints #reasoning #AI #utilites #energy #infrastructure #highstakes

Embracing AI for Daily Life Over the past few weeks, I've had the privilege of delivering several speeches and presentations about AI, showcasing numerous practical examples of how I integrate AI into my daily routine. One frequent request I’ve received is to share my prompt for daily clothing advice from ChatGPT. So, here it is: Step-by-Step Guide to Using GPT Vision for Your Wardrobe: Capture Your Wardrobe: Take clear pictures of each item in your wardrobe. Upload Images: Upload these images to GPT Vision. Categorize and Index: Allow GPT Vision to categorize and index each item by type, color, and style. The Prompt: "Act as my personal stylist and provide me with detailed daily clothing advice based on the weather forecast in Amsterdam and the contents of my wardrobe. Be specific about colors and types of clothing items I should wear. Ensure the advice is practical and tailored to the weather conditions. Include the following details in your advice: Outerwear: Type, color, and specific items. Bottom: Type, color, and specific items. Top: Type, color, and specific items. Footwear: Type, color, and specific items. Accessories: Type, color, and specific items. For the weather forecast, use current data from reliable sources and consider temperature, rain, wind, and other relevant factors. Based on this information, provide a cohesive outfit recommendation that is both stylish and appropriate for the day's activities and weather conditions." Leveraging AI for personalized fashion advice has revolutionized how I approach my daily wardrobe decisions, making it more efficient and tailored. I hope this prompt helps you as much as it has helped me! #AI #FashionTech #PersonalStylist #GPTVision #DailyRoutine #Innovation #TechInFashion #AmsterdamWeather Voorkans Eva Wassink Carine de Meyere Richard Bross Marc Jellema Remy Gieling  

Watching our satellites track storms in real time never ceases to amaze me. Last week, a fast-moving storm system swept across Argentina. Within minutes, Tomorrow.io’s space-powered AI model detected it, tracked it, and updated forecasts as the storm evolved - before it hit. On the left: Our satellites are scanning the system from orbit On the right: Real-time rainfall estimates, updated every 10 minutes as new data came in These AI-powered, space-based updates don’t just look good - they helped a customer in Argentina on the ground respond with greater confidence and speed. How? Most weather systems rely on static snapshots, hours old by the time they’re processed. Ours learns on the fly, turning raw satellite data into live, evolving forecasts powered by AI. This isn’t just about seeing the weather. It’s about changing the way teams act on it. - In aviation, that could mean rerouting flights before turbulence develops. - In energy, dispatching crews before wind shifts hit the grid. - In logistics, avoid costly ground delays hours in advance. This is operational resilience in action, no longer reactive, but proactive. And it's not just possible. It's live today.

Everything was underwater (except the ops team). They knew exactly where everyone was. Field service work doesn't stop when the weather gets rough. But most workforce systems do. This picture shows Alpha Foundations (part of Groundworks) navigating a stormy Miami day. They use HyperTrack to track their service technicians and automate field operations. When storms hit and delays pile up, our location intelligence automatically: ✅️ Tracks real-time position and calculates new ETAs ✅️ Triggers alerts to operations teams about delays ✅️ Provides updated arrival times and tracking links to customers The old way was someone sitting at a desk, frantically making calls, trying to piece together what's happening in the field. With live location tracking, you workforce management system already knows. No scrambling. No frustrated customers. No overwhelmed dispatchers. Field service teams deserve technology that has their back, especially on the tough days. Because when the weather gets rough, your people shouldn't have to. This is what human-first AI looks like in the real world.