<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Harry Slagel on Medium]]></title>
        <description><![CDATA[Stories by Harry Slagel on Medium]]></description>
        <link>https://medium.com/@slagelh?source=rss-ea1c7ba65b84------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*l5rglWNxjTwu0uhpCIk28A.jpeg</url>
            <title>Stories by Harry Slagel on Medium</title>
            <link>https://medium.com/@slagelh?source=rss-ea1c7ba65b84------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 22:36:27 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@slagelh/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[War Broke Out While My Family Was in Dubai. An AI Agent I Built on OpenClaw Got Us Home.]]></title>
            <link>https://medium.com/@slagelh/war-broke-out-while-my-family-was-in-dubai-an-ai-agent-i-built-on-openclaw-got-us-home-eaa65d730bee?source=rss-ea1c7ba65b84------2</link>
            <guid isPermaLink="false">https://medium.com/p/eaa65d730bee</guid>
            <category><![CDATA[openclaw]]></category>
            <category><![CDATA[war]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <dc:creator><![CDATA[Harry Slagel]]></dc:creator>
            <pubDate>Fri, 06 Mar 2026 12:52:01 GMT</pubDate>
            <atom:updated>2026-03-07T16:18:24.063Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>I set up an AI system to monitor a crisis, track flights, and find a way home. It worked.</em></h4><p>On Tuesday morning I read a report generated overnight by an AI system I had built.</p><p>Dubai’s airspace had been closed for days. We already had two flights booked to London that we were fairly sure would be cancelled.</p><p>But the report showed something interesting.</p><p>A handful of planes had finally managed to leave Dubai.</p><p>Paris. Gatwick. Manchester.</p><p>That information made me try something I would never have thought to check otherwise.</p><p>Thirteen hours later my family and I were sitting on a flight out of Dubai.</p><p>If I hadn’t been running that system, I’m not sure we would have made it out when we did.</p><h3>Saturday midday, by the pool, the war starts</h3><p>Last Saturday I was in a swimming pool in Dubai with my young family when the war in the Middle East started.</p><p>Within minutes we began hearing loud bangs across the city.</p><p>At first you don’t quite process what you’re hearing. Then the news alerts arrive and the reality settles in.</p><p>Reports were coming in that Iran had launched drones and missiles across the region in response to US and Israeli strikes. We kept hearing loud bangs across the city. I can’t remember two hours going by without one.</p><p>Airspace across the UAE was suddenly closed. Flights cancelled. Uncertainty everywhere.</p><p>When you’re travelling with young children your mind immediately shifts into a different mode. The only thing that matters is getting your family somewhere safe. I wanted my full focus on them.</p><p>But instead I constantly refreshing the news, checking airline websites. Scanning X for the latest government advice. Trying to piece together what was actually happening.</p><p>Information was chaotic and constantly changing. Every hour brought new rumours, contradictions, and dead ends. I didn’t know what to trust or believe.</p><h3>Building a crisis monitor from a hotel room</h3><p>Before this trip I’d been experimenting with a new type of AI system called OpenClaw. Not a chatbot you ask questions to, but a system that runs in the background, remembers context, connects to external tools and keeps working even when you’re not there. I’d mainly been using it for business tasks. I had no idea it was about to help me get my family home.</p><p>Once the airspace closed and we realised we could be stuck for days, I turned the system toward the crisis.</p><p>I started by getting it to monitor news feeds and government advisory pages. What was the latest travel advice. What were the evacuation options. What should we actually be doing.</p><p>I also asked it to track live airspace status, monitoring for any signs of when Dubai’s airspace might reopen.</p><p>It was compiling everything into structured situation reports delivered to me via Telegram every hour.</p><p>I could go to sleep and wake up to a briefing. I could take my kids to breakfast and check my phone for the latest update. I didn’t have to sit there refreshing five websites hoping to catch something.</p><p>That’s when it started surfacing things I would never have found on my own. The Oman signal came from this early monitoring. More on that below.</p><p>Then when the airspace finally did reopen and the report told me test flights were starting, I asked it to add something new: live departure tracking. Scheduled flights, cancelled flights and most importantly, which ones actually took off.</p><p>That last part turned out to be everything, and what got us home.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hn68tHaG2dgrg4uqadzQeg.jpeg" /><figcaption><em>Hourly crisis updates generated by the AI monitoring system.</em></figcaption></figure><h3>The signal that pointed to Oman before the news did</h3><p>One of the most interesting insights the system surfaced had nothing to do with flights.</p><p>At the time, most of the news suggested evacuation routes would go through Saudi Arabia.</p><p>But the AI picked up something different</p><p>It found signals that British Airways had quietly evacuated some of their cabin crew via Oman.</p><p>That got me thinking.</p><p>If airlines were already moving crew through Oman, there was a strong chance repatriation flights would eventually operate from there too, rather than directly from Dubai.</p><p>Based on that one signal, I got the system to map out a fallback plan:</p><ul><li>Routes to the Oman border</li><li>Transport companies that could take us</li><li>Hotels near the crossing</li><li>Flights leaving from Muscat</li></ul><p>Then it flagged something I would never have considered.</p><p>You cannot simply rent a car and drive across the Oman border. Rental insurance doesn’t cover the crossing.</p><p>Without that insight we could easily have driven there and ended up stranded at the border with two young children and no way forward.</p><p>A few days later, both British Airways and the UK government confirmed that repatriation flights would indeed operate from Oman not Saudi Arabia</p><p>The AI had identified it before anyone announced it.</p><p>I don’t think I would have connected those dots myself. Not while looking after two kids in a hotel room. The system picked it up because it was watching everything, all the time, and I wasn’t.</p><h3>The flight that got us home</h3><p>By Tuesday morning the departure report was waiting for me.</p><p>Several Emirates flights had successfully departed.</p><p>Paris. Gatwick. Manchester.</p><p>That was the moment. I went straight to the Emirates website and started checking those routes. We still had two flights booked to London for the following day, but we had no idea whether they would actually operate.</p><p>Paris. Nothing available.</p><p>Manchester. Nothing either.</p><p>I refreshed the page.</p><p>A flight appeared.</p><p>Within minutes I had booked my entire family onto it. It was leaving thirteen hours later.</p><p>At that point we still had no guarantee any of our flights would actually depart. All day I was monitoring the three flights we had booked and my brother Lewis had a great idea to monitor Flightradar, comparing which flights were still scheduled versus cancelled and checking which ones were actually moving. I plugged this into the crisis monitor as well to automate it. Then we got notified that both the London flights got cancelled. But the Manchester one didn’t.</p><p>I managed to get hold of Emirates and they confirmed it was indeed a repatriation flight.</p><p>A stroke of luck, yes. But without the intel from the AI I would never have thought to look. And the only reason I even thought to check it was because the AI had surfaced those early departures that morning as they took off.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LQTn49c8P0CG1gvW5bdqTA.jpeg" /><figcaption><em>Finally on the flight home after days of uncertainty.</em></figcaption></figure><h3>What would have happened without the AI</h3><p>Sitting in the car driving from Manchester back to London, with my wife and kids finally asleep in the back seat, I knew I had to write about this.</p><p>Without the system:</p><p>I wouldn’t have been able to focus my full attention on looking after my wife and kids. I would have been doing what everyone else was doing — refreshing BBC News, scrolling X, hoping something would break. The same thing millions of people do in a crisis. And it almost never gives you an edge. It just puts you on edge.</p><p>I wouldn’t have got the signals and intel that connected the dots before any human could.</p><blockquote><em>I wouldn’t have found the flight home. I’d probably still be in Dubai. I wouldn’t have been checking the Manchester route at six in the morning. That flight would have come and gone without us.</em></blockquote><blockquote><em>I wouldn’t have known that the Manchester flight was actually going to go if I didn’t use my brother Lewis’s idea of monitoring Flightradar and then asking the AI agent to alert me if anything changed, which led me to call Emirates to confirm it was a repatriation flight.</em></blockquote><blockquote><em>I wouldn’t have known — or it would have taken me much longer to find out — that Oman was shaping up to be the evacuation route. Even though we didn’t end up going that way, the system planned the entire thing out. Routes, transport, hotels, flights from Muscat. Luckily I also had some friends who started that journey, so I had some firsthand insights as well.</em></blockquote><blockquote><em>I wouldn’t have known about the rental insurance problem at the border. We could have been stranded. Instead we looked at a private transfer option.</em></blockquote><p>The AI didn’t make the decisions for me. But it dramatically improved the quality of the information I was working with, got it to me faster, and automated it. It watched everything, constantly, and surfaced the signals that actually mattered.</p><p>In a fast-moving crisis, that kind of speed and clarity is the difference between getting out and being stuck.</p><h3>The next phase of AI is already here</h3><p>So what is this system and why does it feel so different from ChatGPT?</p><p>OpenClaw is part of a new category of AI systems often called agent frameworks. Instead of being just a chatbot, it behaves more like a persistent AI operating system that can run tasks, remember context and interact with tools continuously.</p><blockquote><strong>Disclaimer:</strong> OpenClaw is still in beta and can present security risks if not configured properly. Exposed API keys, insecure network access, overly broad tool permissions etc. That’s why I run it locally on my own old MacBook at home, and run daily security audits.</blockquote><p>It has memory, so it remembers context from previous conversations rather than starting from scratch. It has skills, repeatable capabilities like scraping websites, analysing data or monitoring changes. It uses tools to connect to external systems, allowing it to import and export data. And it has agents, autonomous workers that run tasks continuously in the background without me prompting them.</p><p>On top of all this I built what I call a mission control layer. A central interface that manages all the agents, tasks, skills, tools and memory in one place. The system actually built this dashboard itself in about an hour.</p><p>I interact with the whole thing through Telegram. I message it just like I would message a person. I also connected a skill that allows it to understand voice notes. So during the crisis I could literally just hold down the mic button in Telegram, tell it what I needed, and it would process it and act on it.</p><h3>What this means for the future with AI</h3><p>The system I built was experimental. I’d only planned to play around with it.</p><p>Then a war started while i was on hoilday and it became the most useful thing I had.</p><p>A few years ago, what this system did would have required an entire operations team working around the clock. Last week it was running on my old MacBook at home while I was in a hotel room in Dubai, directing it by sending voice notes and messages via Telegram.</p><p>AI that works for you 24/7, constantly monitoring, automating tasks, delivering the information you need in a concise way you can actually act on, and thinking outside the box, connecting the dots on things you never would. That’s the future. And the people who learn how to use it effectively, who learn how to work with it rather than get replaced by it, are going to benefit the most.</p><p>If I hadn’t been using it in that moment, I think I’d still be in Dubai right now.</p><p>It’s what got my family home.</p><h3>Appendix: How I built the system</h3><p>Here’s the full updated appendix:</p><h3>Appendix: How I built the system</h3><p>For anyone interested in the technical side.</p><p>The core is OpenClaw running on my local machine. It doesn’t have its own AI model. It uses existing ones and basically ties them together for different tasks, whichever makes the most sense. I’ve been using Claude Opus as the core brain, Grok’s API for social search, Claude Sonnet via Claude Code for coding tasks, and others for web search, voice notes and so on. You plug in whichever models work best for each job.</p><p>On top of that I built the mission control layer that manages agents, tasks, cron jobs, and tool connections.</p><p>The crisis monitoring setup used a combination of skills, tools and agents working together:</p><p>Skills: news monitoring, government site scraping, aviation data checking, report summarisation.</p><p>Tools: Telegram for interaction, various flight tracking and news data sources for monitoring.</p><p>Agents: one checking aviation departures every hour, one monitoring news feeds for new developments, one summarising changes and delivering situation reports via Telegram.</p><p>In OpenClaw, cron jobs and heartbeats are what allow the system to operate continuously rather than waiting for prompts. Cron jobs schedule tasks at specific intervals. Checking news sources, monitoring flight departures, generating reports or running research overnight. Heartbeats act like a regular pulse for the agents, waking them up periodically to check their state, look for new information and decide whether any action needs to be taken. Together they allow agents to run in the background, monitor signals and surface important updates automatically. During the crisis, that meant every hour the system was pulling fresh data, cross-referencing it with government advisories, and flagging departures that had actually taken off, not just flights that were scheduled.</p><p>I initially started to use the system for my own startup, Martee AI. Continuous lead generation through Apollo, research and strategy as a second brain, storing outputs in Notion, and building fully functional apps. I might write a separate post on that, but I’ve only scratched the surface so far. I’ve seen people setting up fully automated AI teams that go far beyond what I’ve built.t.</p><p>I’m contemplating building out the crisis app a bit more so others can use it who are stuck there. I’ll post it here if I do.</p><p><em>If you found this useful, follow me for more on building practical AI apps, agents, systems and tools. I might start posting more. And if you’re experimenting with OpenClaw or similar tools, I’d love to hear what you’re building. What would you use a system like this for?</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eaa65d730bee" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The next generation of service designers give us a run for our money]]></title>
            <link>https://medium.com/@slagelh/the-next-generation-of-service-designers-give-us-a-run-for-our-money-b8b34330b8da?source=rss-ea1c7ba65b84------2</link>
            <guid isPermaLink="false">https://medium.com/p/b8b34330b8da</guid>
            <category><![CDATA[career-paths]]></category>
            <category><![CDATA[digital]]></category>
            <category><![CDATA[service-design]]></category>
            <category><![CDATA[charity]]></category>
            <dc:creator><![CDATA[Harry Slagel]]></dc:creator>
            <pubDate>Thu, 16 Nov 2017 14:47:11 GMT</pubDate>
            <atom:updated>2017-11-16T14:47:11.930Z</atom:updated>
            <content:encoded><![CDATA[<p>By <a href="https://medium.com/u/ea1c7ba65b84">Harry Slagel</a>, Service &amp; Experience Design Consultant at Transform</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*whAKF8GkCYBChiR1w-wCHg.jpeg" /></figure><p>Last month Transform teamed up with <a href="http://www.ncsyes.co.uk/">National Citizen Service (NCS</a>), for the third time since 2016, to help young adults with an interest in Service Design develop their career prospects.</p><p>NCS is a charity that gives 15–17 year olds the chance to embark on challenges that will make and build skills for work and life. We invited eight students into Transform HQ to experience a day in the life of the ever-growing and exciting worlds of Digital and Service Design. The aim of the day was to teach valuable skills and grow confidence… whilst having fun, of course!</p><p>The day kicked off with an informative introduction from Transform CEO Emma Robertson. Emma shared some of her expertise and insights on Service Design and Technology, as well as introducing Transform, to give the students a taste of what these industries are all about.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*D3G9xoh_2JxlBDAIYBdh3Q.jpeg" /></figure><p>Next up for the students was a workshop on service design processes. The session took inspiration from one of Transform’s ongoing projects with <a href="http://transformuk.com/ourwork/public-health/">Public Health England (PHE)– the “One You” campaign.</a> With the primary goal of promoting healthier lifestyles, the work we’ve done with PHE is a great example of how user research and interaction design can be combined to create a suite of tools, including a quiz; a web page; and four mobile apps to support behaviour change.</p><p>By taking the students through this case study and our service design methodology, we hoped to inspire them to think outside the box and use design-thinking to generate some fresh, innovative ideas. And so, we challenged the students to come up with their own solutions to the same “One You” brief that we received.</p><p>Splitting into four teams, with a Transform consultant facilitating each, the students carried out user research by interviewing their consultant on their daily health habits. This upfront research formed ‘personas’ where information was gathered around each consultant’s daily routine, they’re emotions throughout the day, and health goals and barriers. Using this information, students were asked to identify user needs and come up with responses.</p><p>With all the teams deciding to create a smartphone app as their solution, an array of fantastic and interesting ideas were generated. I must say, I was particularly impressed by one group who split their thinking into three pillars- knowledge, motivation and wellness- which I thought was a very nice way to build a brand image and to create a clear definition of what they intended it to do.</p><p>The conceptualisation phase let the students bring their idea to life through either designing the look of their app or creating storyboards to show a typical user journey. There was a fantastic variety of designs, displaying great creative thinking.</p><p>And finally, it was pitch time! Dragons Den style. Each group was given time to plan and rehearse a presentation before bravely pitching to our panel of Transformers (we’re not that scary!)- made up of consultants and senior management. The panel saw strong passion, enthusiasm and innovative thinking in all the presentations.</p><p>But, there could only be one winner…</p><p>After much debate, the panel decided to award the team that designed an app called “Matcha” as the winner. This was the team that cleverly split their idea into three brand pillars during planning. The first ‘knowledge’ pillar provided information on different types of exercises and equipment, including a search engine that looked for all fitness videos available on the web. The ‘motivation’ pillar gamified the user’s fitness progress, providing encouragement and motivation through notification alerts. And the ‘wellness’ aspect involved a nutrition and fitness tracker.</p><p>It wasn’t just their great idea (not to mention the creative app name) that made us select them as the winning team; but they also presented a superb pitch that was not only professional, but also full of passion.</p><p>After an extremely fun and action-packed day, CEO Emma Robertson came back to do a Q&amp;A, where students grilled her with a great variety of both surprising and impressive questions.</p><p>Overall, I was extremely impressed with how quickly the students took to service design and harnessed their creative thinking. I would have loved to have an opportunity like this available when I was their age. It really was an invaluable day for the design-curious, bright minds we met.</p><p>And best of all, I’m thrilled to see that all the feedback from the students was overwhelmingly positive. We will most certainly be taking part again next year!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NYtFltnn-I1kr4uohNaZ8w.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b8b34330b8da" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>