Issue #5: Why Most Businesses Don't Need New Software
![]() | |||||||||||||||||
Issue #5 · November 3, 2026 Hey {{ first_name | default: "friend" }}! Issue #5 of Stop Being the System. Last time we talked about why a single source of truth starts with agreeing on what the words mean. This time we're going to address why fixing the process usually beats buying another tool. | |||||||||||||||||
| |||||||||||||||||
Fix things in this order Most owners start at the end. They buy a tool or build an AI workflow and hope it fixes the mess. I fix things in a different order: process, then technology, then automation.
Automation amplifies whatever is already there. Automate a broken process and you get a broken process, faster. AI didn't change the order. It made skipping the first step cheaper and more tempting. I've done it myself. For years I kept internal commitments and client requests in one system, and kept switching tools to fix it. The fix wasn't a new tool. It was one decision: a single triage step, with "issues" as a flag on a task instead of a separate system.
| |||||||||||||||||
Fast models and thinking models Some AI models answer right away. Others work through the problem first. That's the difference between a simple model and a reasoning model. A simple model takes your question and gives you an answer. It's fast and cheap, and it's plenty for a lot of everyday work: summarizing, sorting, pulling facts out of a document, rewriting. A reasoning model works through its steps before it answers. It's still not thinking the way you do, as we covered last week. It's predicting with more steps. That makes it slower and more expensive, but better at problems with several moving parts, like math, planning or tangled logic. Models come in sizes too. Claude has Haiku, Sonnet, Opus, and now Fable, from smaller and faster to bigger and more capable. ChatGPT and Gemini have their own versions. The mistake is reaching for the biggest one every time. You pay more and wait longer for work a smaller model does just as well. Start small. Move up when the answer isn't good enough. | |||||||||||||||||
Put a step between the PDF and the order Every customer sends a purchase order a little differently. Some come as emails with a random PDF attached. The SKU might be missing a dash, or have a different number on the end. You'll never get every customer onto one process. A company I work with had been dropping those PDFs into Claude and having it create the orders directly in their system. Too many things were going wrong along the way, so we adjusted. We didn't try to go from PDF to order in one jump. We put a step in the middle. The PDF shows on one side. On the other is a grid of products with the SKU, quantity and price. Someone scans down the list and says, "Yup, that looks right." Where a SKU doesn't match exactly, it gets a big red "Review this line." It still takes its best guess, even when the SKU is missing a dash. The model matters here too. I use Claude for most of my work, but this job needed a lot of product data passed in alongside the PDF. We used Gemini because it could read the PDF, pull out the details, and handle all of that data cheaply. This week's experiment: take one messy document you handle by hand, like an invoice or a form. Ask AI to put the details in a table and flag anything it's unsure about. Don't let it enter anything. Just look at what it flags. | |||||||||||||||||
Rework
Tools are the easy thing to buy and the wrong place to start. Get the process right first, then pick the tool. | |||||||||||||||||
Fix the funnel before filling it A client recently came to me wanting to use AI to improve their marketing so they could get more leads. We did, but not the way they expected. On our first call, I went through their website and pointed out the gaps that would make someone hesitate to work with them. Then I showed them how AI could help us fix those. They already had the technology. Their process needed work first. AI made some of that work easier, and it also let them make decisions on their own without having to be experts. We'll still build new systems that use AI to bring in more leads. But the first step was making sure we weren't losing people out of a hole in the funnel. | |||||||||||||||||
Thanks for reading. Hit reply and tell me which section was most useful. I read every reply.
| |||||||||||||||||
|


Responses