- Just a Guy With His AI
- Posts
- Professional Services and AI
Professional Services and AI
AI is going to change how you communicate with your contractor, plumber, and more.
I’ll share which AI tools I’m exploring, some experiments I’m conducting, and insightful information about what I’m observing in the world every week.
🔧 Three Tools I’m Testing
🖊️ Rytr.me - Another content writing tool. The quality initially seems solid, but compared to Writesonic, it’s similar but offers fewer features right now.
👶 Pika Baby Me - Pika continues to add fun transformations for me to provide endless enjoyment to my friends and me. I use this to see the continued innovation in image generation.
🧠 NotebookLM Mind Maps - Google continues to update NotebookLM with solid features for learning and understanding content. The latest is the ability to have it create a mind map of the information, and then you can interactively dive into those. I find myself jumping into NotebookLM weekly.
🧪 AI Experiment of The Week
This week, I’m starting to build an MCP server to test integrations with the Losant platform. My long-term goal is to enable Cursor or Claude to use Losant data and create resources inside the platform. I’m also trying to understand the underlying hype and usefulness of MCP servers.
I’m using Cloudflare’s MCP Worker framework to build and deploy the server for testing. I’m not sure if this is the best long-term solution, though.

The Most Basic MCP Server
I broke out Cursor to do some vibe coding and create my MCP server. I’m still making decisions about how to integrate Losant’s authentication structure for application data. I understand you can build stateful and stateless (a new capability) MCP servers. I believe I can better authorize Losant resources using the stateful version.
This is just the beginning for me on MCP and this specific implementation. More to come.
📰 Article of The Week
Chain of Experts Increases Efficiency and Accuracy - Chain of Experts (COE is a modification of the Mixture of Experts (MOE) model processing method. Experts are trained pieces of the model on a specific subject (finance, health, etc…). MOE is a method that processes data in parallel with their Experts. This produces excellent results, as seen in DeepSeek-V3, and GPT-4o is believed to use this method. The modification of this method for COE is to provide the data sequentially instead of in parallel.
Experiments show that CoE significantly outperforms previous MoE models in multiple aspects:
- Performance: CoE with 2x iterations reduces Math validation loss from 1.20 to 1.12
- Scaling: 2x iterations matches performance of 3x expert selections, outperforming layer scaling
- Efficiency: 17.6-42% lower memory usage with equivalent performance
- Flexibility: 823x increase in expert combinations, improving utilization, communication, and specialization
These advantages constitute a "free lunch" effect, enabling efficient scaling of LLMs.
🌎 Where the World is Going
There will be a revolution in the service industries, but it isn’t about robots or self-driving cars—it's about solving the oldest problem in the business world: communication. Whenever I call a plumber who never confirms appointments or a contractor who forgets to send an estimate, I'm reminded that the fundamental pain point for small service businesses hasn't changed in decades. These aren't technology companies, yet poor communication costs them thousands in lost business and customer goodwill every year.
What fascinates me is how AI-powered digital workers will address this gap. Imagine a small HVAC company where an AI assistant handles all client scheduling, sends automated appointment reminders, follows up on quotes, and even drafts initial responses to customer inquiries—perfectly matching the owner's communication style. Technology exists today, but adoption in these traditional industries is moving at a glacial pace. The same electrician who still uses paper invoices isn't rushing to implement conversational AI, yet these are precisely the businesses with the most to gain.
The actual transformation will come from this unique human-AI partnership. The master carpenter keeps doing what she does best—crafting beautiful cabinets—while her digital counterpart manages the communication logistics that previously fell through the cracks. This isn't about replacing the human element in service work; it's about amplifying it by removing the administrative burdens that prevent craftspeople from focusing on their craft. What's most encouraging is that this technology could actually preserve small service businesses rather than threatening them, making them competitive against larger operations with dedicated office staff. I believe it can bring a golden age for small service companies.
👨💻 About Me
![]() Just a Guy with An Ostrich | My name is Charlie Key. I love technology, building awesome stuff, and learning. Over the past twenty-plus years, I’ve built several software companies. I’ve written this newsletter to help inspire and teach folks about AI. I hope you enjoy it. ➡️ Learn More About The Guy ⬅️ |