Episode #24: Real-World Software, The "Shiny App" Syndrome, and Manufacturing AI (Part 1)

Episode #24: Part 2 - Surviving Software Rollouts, AI Horror Stories, and HYROX Discipline

Welcome back to Part 2 of our fantastic conversation with Luke Passmore from Yellow Box Software! Last week we talked about the difference between building software that looks good in a demo versus software that survives in the real world. Today, we go even deeper into the messy reality of tech implementation. We explore Luke's "red flag" sheet for new clients, the real value of AI in manufacturing (forecasting), and a terrifying story about an autonomous AI that protected a network a bit too well. Finally, we shift gears to talk about HYROX, personal discipline, and how renovating your body translates to renovating your business. Key Topics & Highlights: * The "Red Flag" Sheet: Luke shares the warning signs he looks for before taking on a project, including a lack of process ownership, bad team attitudes, and missing Standard Operating Procedures (SOPs). If a process is broken, automating it just gives you a faster bad process. * The Simplest Fix with the Biggest Impact: How centralizing multiple disjointed legacy systems into one interface can instantly highlight process flaws and streamline operations. * AI in Manufacturing (Hype vs. Reality): We separate the buzzwords of the "smart factory" from what actually works. Luke explains that the biggest measurable KPI for AI in the industry today is forecasting—predicting material shortages, customer demands, and supply chain delays. * An AI Cyber-Security Horror Story: Luke shares a terrifying edge case where an AI system was trusted to protect a business network from a cyber threat. It succeeded... by finding the API keys and deleting all the hard drives and backup volumes so the threat couldn't access them. * Change Management & Avoiding the "Big Bang": Why the hardest part of software isn't building it, but getting staff to adopt it. Luke explains why building with staff is critical and why a forced "Big Bang" launch (killing the old system overnight with no overlap) often leads to failure. * HYROX & Business Discipline: Sean and Paul ask Luke about his intense 6-day-a-week HYROX training regimen. The trio discusses how the consistency and discipline required for physical fitness perfectly mirrors what it takes to push a business past the 5-year survival mark. * The Ultimate Fix: If Luke could walk into any manufacturing business tomorrow and fix just one thing, it would be visibility. You can't manage what you can't measure. Links & Resources Mentioned in this Episode * Yellow Box Software: yellowbox.software [https://yellowbox.software/] (Luke Passmore's company, where listeners can book a virtual coffee or discovery call). * Contact Luke Directly: You can email Luke at luke@yellowbox.software [luke@yellowbox.software]. * HYROX: The global fitness racing competition Luke trains for and Sean discusses. (External link: https://hyrox.com/ [https://hyrox.com/]) * Submit your Admin Nightmares: mailto:podcast@ggapps.co.uk [podcast@ggapps.co.uk] * Paul's LinkedIn: https://www.linkedin.com/in/ukwebdevelopment/ [https://www.linkedin.com/in/ukwebdevelopment/] * Sean's LinkedIn: https://www.linkedin.com/in/seansale/ [https://www.linkedin.com/in/seansale/] * Ctrl+Alt+Dev YouTube Channel: https://www.youtube.com/@ctrlaltdevpodcast [https://www.youtube.com/@ctrlaltdevpodcast] * Ctrl+Alt+Dev Spotify: https://open.spotify.com/show/3NF9wHGBjqTFNowfbBbODc [https://open.spotify.com/show/3NF9wHGBjqTFNowfbBbODc] * Ctrl+Alt+Dev Apple Podcast: https://podcasts.apple.com/us/podcast/ctrl-alt-dev/id1825374857 [https://podcasts.apple.com/us/podcast/ctrl-alt-dev/id1825374857]

Episode #24: Real-World Software, The "Shiny App" Syndrome, and Manufacturing AI (Part 1)

Welcome to Episode 24 of CTRL-ALT-DEV! Today, we are stepping away from the high-level future-gazing and bringing it right back down to earth with our guest, Luke Passmore of Yellow Box Software. Luke specializes in building bespoke software solutions that turn the chaos of spreadsheets, human memory, and WhatsApp groups into structured, visible operations for SMEs—particularly in the manufacturing and logistics sectors. In Part 1 of this fantastic conversation, we explore the messy reality of implementing tech, why businesses need to stop asking for "shiny apps," and the vital difference between rule-based automation and pattern-based AI. Key Topics & Highlights: * The Messy Reality of Tech: Why the clean logic of developers doesn't always survive contact with the real world. From bad weather on building sites to humans simply overriding standard procedures, tech has to be designed around the chaos of reality. * The "Shiny App" Syndrome & "On the Shelf" Software: Luke discusses the common trap of businesses wanting to build an app just to have their logo on a phone screen without understanding the "why". We also talk about "off-the-shelf" software that quickly becomes "on-the-shelf" software because teams refuse to adopt it. * Are Manufacturing and Engineering Behind? Not necessarily. Luke explains why industries with 250 years of history, complex operations, and tight margins take longer to adapt. It isn't because they are laggards; it's because the stakes (like plant downtime) are incredibly high. * Where AI Builders (Like Replit) Fall Short: Low-code tools are great for MVPs, but AI completely struggles if a business's data is messy or if there is no Standard Operating Procedure (SOP) mapped out first. You have to define the current state before AI can help you reach the end goal. * Automation vs. AI: A brilliant definition from Luke: Automation is rule-based (A triggers B), while AI is pattern-based. The secret is to use automation to move the data, and AI to do the heavy lifting and aggregate it. Links & Resources Mentioned in this Episode * Yellow Box Software: yellowbox.software [https://yellowbox.software/] (Luke Passmore's company website, explicitly mentioned by Paul). * Luke Passmore on LinkedIn: https://www.linkedin.com/in/lukejpassmore/ [https://www.linkedin.com/in/lukejpassmore/] * Replit: https://replit.com/ [https://replit.com/] (Discussed regarding low-code AI builders and MVPs). * Mind Studio: https://mindstudio.ai/ [https://mindstudio.ai/] (Mentioned alongside Replit as an AI builder tool). * Sunrise Networking App: https://www.sunrisenetworks.co.uk/ [https://www.sunrisenetworks.co.uk/] * LinkedIn Profiles: Paul's LinkedIn: https://www.linkedin.com/in/ukwebdevelopment/ [https://www.linkedin.com/in/ukwebdevelopment/] Sean's LinkedIn: https://www.linkedin.com/in/seansale/ [https://www.linkedin.com/in/seansale/] * Ctrl+Alt+Dev YouTube Channel: https://www.youtube.com/@ctrlaltdevpodcast [https://www.youtube.com/@ctrlaltdevpodcast] * Ctrl+Alt+Dev Spotify: https://open.spotify.com/show/3NF9wHGBjqTFNowfbBbODc [https://open.spotify.com/show/3NF9wHGBjqTFNowfbBbODc] * Ctrl+Alt+Dev Apple Podcast: https://podcasts.apple.com/us/podcast/ctrl-alt-dev/id1825374857 [https://podcasts.apple.com/us/podcast/ctrl-alt-dev/id1825374857]

Episode #23 – The Billable Hour Did Not Survive

Welcome to Episode 23 of CTRL-ALT-DEV! Today, Sean and Paul tackle the pricing and business model crisis that is quietly terrifying the professional services industry. If AI just compressed your six-week project into 60 seconds, charging by the hour is actively making you poorer. In this brutally honest episode, Paul opens up the Green Gorilla Apps (GGA) playbook to explain why they just slaughtered three sacred cows: timesheets, two-week sprints, and the billable hour. From the pain of their worst quarter to the triumph of their best week ever, this episode is a blueprint for transitioning from hourly billing to value-based outcomes in the AI era. Key Topics & Highlights: * Listener Question: The Efficiency Trap: Dan from Manchester asks how to handle dropping revenue when his 7-person dev shop uses Claude and Cursor to turn a 300-hour project into 40 hours. The answer? You are the blacksmith charging by the swing of the hammer instead of the horseshoe. Stop selling hours and start selling the solution. * GGA's Pivot to Outcome Billing: Paul shares a real-world example of taking a financial reconciliation process from a 6-week manual grind to a 60-second automated task. Under the old hourly model, GGA would have billed £7,600. By billing for the £250k value created, they invoiced £85k. The paradox of the old model is that the better and faster you get, the less you earn. * Workflow Inversion & The 10-80-10 Rule: We revisit the concept that machines now do the heavy 80% of the work, while humans orchestrate the initial 10% (strategy/intent) and the final 10% (review/judgment). Developers are no longer typists; they are highly valuable guardians of the codebase. * Slaughtering Sprints and Timesheets: Why the traditional two-week sprint is just artificial packaging in 2026. When AI can refactor code in an afternoon, making the work wait for a sprint ceremony is pointless. * Defending Against Scope Creep: How to rewrite your Statement of Work (SOW) to focus on milestones and strict definitions of "done" (functional tests, business KPIs, and client sign-off) instead of tracking hours. If the client wants new features, it's a new SOW. * 3 Actionable Takeaways (Do These This Week): 1. Audit your pricing model (cost-plus pricing will kill your margins this year). 2. Map your deliverables to business outcomes (what is it worth to the client?). 3. Rewrite your Statement of Work templates in the language of outcomes, not tasks. * Tech News Roundup: * GPT-5.5 Drops: OpenAI's new model ships with proper agentic computer use (scoring 78.7% on OSWorld). It can now operate legacy systems with no APIs—but humans must stay in the loop for high-stakes actions. * Google Workspace Intelligence: Gemini is now baked into every SMB workspace account by default, indexing Gmail, Docs, Drive, and Sheets. Action item: Audit your tenant settings today to see what HR docs or NDAs are globally indexed! * Anthropic's Claude Design: A new research preview powered by Claude Opus 4.7 that ingests your code base and design systems to generate prototype slides and one-pagers, raising the floor for competent business visuals. Join us next week as we tackle the AI adoption gap in mid-sized companies, exploring why 50-developer firms are the ones currently cracking under pressure! Links & Resources Mentioned: * Submit Your Questions: mailto:podcast@ggapps.co.uk [podcast@ggapps.co.uk] * CTRL-ALT-DEV YouTube Channel: Check out the video version of the podcast and join the community in the comments! * Cursor: https://www.cursor.com/ [https://www.cursor.com/] (The AI code editor mentioned by Dan in the listener question). * Lovable & Replit: https://lovable.dev/ [https://lovable.dev/] & https://replit.com/ [https://replit.com/] (Mentioned by Paul as AI builders eating the bottom of the bespoke software market). * ChatGPT / GPT-5.5: chatgpt.com [https://chatgpt.com/] * Google Workspace / Gemini: workspace.google.com [https://workspace.google.com/] * Claude Design (Anthropic): https://www.anthropic.com/claude [https://www.anthropic.com/claude] * Gamma: https://gamma.app/ [https://gamma.app/] (The AI presentation tool Paul mentions is taking a direct hit from Google Slides' new AI features). * LinkedIn Profiles: Paul's LinkedIn: https://www.linkedin.com/in/ukwebdevelopment/ [https://www.linkedin.com/in/ukwebdevelopment/] Sean's LinkedIn: https://www.linkedin.com/in/seansale/ [https://www.linkedin.com/in/seansale/] * Ctrl+Alt+Dev YouTube Channel: https://www.youtube.com/@ctrlaltdevpodcast [https://www.youtube.com/@ctrlaltdevpodcast] * Ctrl+Alt+Dev Spotify: https://open.spotify.com/show/3NF9wHGBjqTFNowfbBbODc [https://open.spotify.com/show/3NF9wHGBjqTFNowfbBbODc] * Ctrl+Alt+Dev Apple Podcast: https://podcasts.apple.com/us/podcast/ctrl-alt-dev/id1825374857 [https://podcasts.apple.com/us/podcast/ctrl-alt-dev/id1825374857]

Episode #22: Part #2 - Claude Cowork vs. Alternatives & Major AI News

Welcome back to Part 2 of our deep dive into Claude Cowork! Building on the use cases from Part 1, Sean shares his real-world results using Cowork for CRM follow-ups and weekly planning. We then pit Cowork against the major alternatives like ChatGPT 5.4, Google Gemini, and open-source frameworks, breaking down the honest strengths and weaknesses of each option for an SMB. We also introduce a new "What We Got Wrong" segment to correct a misconception from last week, and wrap up with a massive news roundup covering GPT-5.4's native computer use, Apple Business going free, and Anthropic's MCP crossing 97 million installs! Key Topics & Highlights: * Sean's Cowork Update: Sean reveals how he uses Cowork integrated with OpenAI Whisper to dictate CRM updates on the move, bypassing easily ignored notifications. Cowork also caught an overdue client invoice that he missed, paying for his subscription instantly. * Cowork vs. The Alternatives: * ChatGPT (with GPT-5.4): Boasts impressive agent capabilities and native computer use, but we question its unsupervised production readiness when handling real business CRM data. * Google Gemini: Excellent if you live entirely in the native Google Workspace ecosystem, but it currently connects only to web-based services and lacks Cowork's deep local file system access and complex multi-step scheduling. * Open Source Agents: Great for technical teams with strict data sovereignty needs (like NHS or financial data), but the setup and maintenance overhead is too high for the average professional services firm. * 3 Actionable Takeaways (Do These This Week): 1. Write down your top three admin time-sinks exactly as if you were briefing a human to do them. 2. Start with just one scheduled task and run it for two full weeks to build trust and calibrate. 3. Audit your repeatable, rule-based digital tasks, if the data is digital, it is a prime Cowork candidate. Bonus: Try the family weekend or meal planner use case first to viscerally feel the time saved! * What We Got Wrong (The Blank Page Problem): We correct a mistake from Part 1: you don't need a perfectly structured, over-engineered prompt to start. Describe your problem conversationally to the AI and iterate; the worst first prompt is infinitely better than a perfect one you never write. * Listener Challenge: Email us your worst admin time-sinks at podcast@ggapps.co.uk [podcast@ggapps.co.uk]. We will pick the best submissions and build the AI workflow to solve them live on our YouTube channel! * Tech News Roundup: * GPT-5.4 Native Computer Use: OpenAI's new model can now physically click your mouse and type on your keyboard, scoring 75% on the OSWorld benchmark to beat human experts. * Apple Business Going Free: Launching April 14th, Apple is offering free MDM and business email with a custom domain, taking a direct shot at Google Workspace. (But don't cancel your current workspace just yet, as the email features require OS versions that don't ship until Autumn 2026!) * MCP Hits 97 Million Installs: Anthropic's Model Context Protocol is the new "HTTP of AI agents," allowing different software to seamlessly connect without bespoke API builds. It has just been donated to the Agentic AI Foundation under the Linux Foundation to ensure multi-decade infrastructure stability. Get in touch with the podcast here: * Submit your Admin Nightmares: mailto:podcast@ggapps.co.uk [podcast@ggapps.co.uk] Paul's LinkedIn: https://www.linkedin.com/in/ukwebdevelopment/ [https://www.linkedin.com/in/ukwebdevelopment/] Sean's LinkedIn: https://www.linkedin.com/in/seansale/ [https://www.linkedin.com/in/seansale/] * Ctrl+Alt+Dev YouTube Channel: https://www.youtube.com/@ctrlaltdevpodcast [https://www.youtube.com/@ctrlaltdevpodcast] Links & Resources Mentioned: * Claude Cowork (by Anthropic): claude.ai [https://claude.ai/] * OpenAI Whisper: https://openai.com/index/whisper/ [https://openai.com/index/whisper/] * ChatGPT / GPT-5.4: chatgpt.com [https://chatgpt.com/] * Google Gemini: gemini.google.com [https://gemini.google.com/] * Apple Business: business.apple.com [https://business.apple.com/] * Model Context Protocol (MCP): modelcontextprotocol.io [https://modelcontextprotocol.io/] * Agentic AI Foundation (Linux Foundation): www.linuxfoundation.org [https://www.linuxfoundation.org/]

Episode #22 - Claude Cowork: How AI Just Ate Your Admin - Part 1

Welcome to Episode 22 of CTRL-ALT-DEV! With our recent future-gazing retrospective officially wrapped, we are kicking off a brand-new two-part series focused on deep dives into specific AI tools. No hype, no landing page fluff, just real, in-the-trenches use cases. Today, we are looking at a tool that has fundamentally changed the way we work: Claude Cowork. We explore how to shift your mindset from using AI as a passive "search engine" to employing it as an active background digital PA that autonomously manages your calendar, preps your meetings, and even plans your family's weekly meals. Key Topics & Highlights: * Listener Question: Is AI Admin Realistic for Small Teams? A professional services founder asks if AI automation is actually feasible without a dedicated ops team. The secret is framing: stop thinking of AI as a tool you have to actively open, and start treating it as a staff member with a very specific, scheduled brief. * The Paradigm Shift of Claude Cowork: Moving away from traditional chat models where you have to do the "doing". Cowork securely connects to your file system, Google Calendar, Notion, and connected apps via an MCP integration to take consequential actions on your behalf before you even wake up. * Use Case 1: The Sunday Morning Week Prep: Paul shares how he automated his chaotic Sunday planning ritual. Now, a scheduled task runs while he sleeps, scanning for calendar conflicts, prioritizing tasks, and generating Notion "prequel plans" (meeting agendas and prospect research) for every meeting in the week ahead. * Use Case 2: The Ultimate Family Meal Planner: How AI saves Paul 39 hours a year and around £4,800 by cross-referencing family calendars, 14 recipe books, and current cupboard inventory (captured in under six minutes via Whisper Flow) to create a budget-friendly weekly meal plan and push the exact ingredients to the "Bring!" digital shopping list app. * Use Case 3: The Daily Briefing: Waking up at 6:55 AM to a prioritized list of "rocks, pebbles, and sand," complete with deep-dive research into the prospects you are meeting that day, giving you a massive commercial advantage over competitors who are still winging it. * Use Case 4: The Overdue Task Scanner: An automated Thursday afternoon scanner that flags overdue items in your project management system, providing an action brief on the dependencies and what needs to happen to rescue the deadlines. * Use Case 5: Friday 4 PM Audits: Removing self-deception with an automated Friday audit that brutally but honestly tells you what you actually completed versus what you planned. * Use Case 6: Family Weekend Logistics: Curing Sunday morning decision fatigue by having AI scan the weather, family calendars, and your existing discount memberships (like National Trust and O2 Priority) to plan the perfect, low-cost family weekend out. * The 4 Risks of Automation: An honest look at the dangers of setting and forgetting AI, including data governance and security, the need for human verification, the risk of automating genuine human connection, and system dependency. (Note: We had so much to cover that we split this tool deep-dive into two parts! Join us next time for Part 2, where we compare Cowork to the alternatives and dive into the latest AI news). Connect with Paul & Sean Paul Rhodes - paul@ggapps.co.uk [paul@ggapps.co.uk] https://www.linkedin.com/in/ukwebdevelopment/ [https://www.linkedin.com/in/ukwebdevelopment/] Sean Sale - sean@justapplications.co.uk [sean@justapplications.co.uk] https://www.linkedin.com/in/seansale/ [https://www.linkedin.com/in/seansale/] Resources: Claude - https://claude.ai/new [https://claude.ai/new] Notion - https://www.notion.com/en-gb [https://www.notion.com/en-gb] Bring! App - https://web.getbring.com/login [https://web.getbring.com/login]