One of the questions I get often is, "Which is the best AI model to use?". The answer ----> it is less about which model to use (they are rapidly changing and building out capabilities daily), and more about how to build out your workflows so that if you need to change AI models tomorrow, it is like flipping a switch. Why is this important? There are many reasons. Token usage: when you get into big projects, it costs more tokens/money. If you are in an organization, once you have AI working, your next evaluation is how much is this costing? Also, what happens if a model goes away and you are deeply embedded? And last, you will get different results with different models- learning how to test between models is the new AB test that can have an affect on the quality of your work. Not being too reliant on any one model is not only advantageous, it is truly best practices in building AI native systems. Do you feel stuck within one particular platform and not sure how to build out a portable workflow? If you are still using AI to chat, but haven't built a recurring task or workflow, comment below and I can host a free class. 👇

Hi everyone, I am taking a poll for topics this week. Let me know what you would like to learn. Feel free to add anything else in the comments. Your feedback helps me make relevant content 💫

This week in presenting to various clients and groups, a recurring theme was surfacing. People were mentioning poor outcomes with the models. Here are a few tips! Optimizing Your AI Workflow: 3 Rules for Better Model Performance 🧠 If you're experiencing inconsistent answers or hallucinations in your chats and projects, the issue often comes down to scope and context. To get reliable, high-quality outputs, keep these three core principles in mind: 1. Narrow Your Scope Keep chats and projects focused on a single topic. A tightly defined scope prevents the model from wandering and ensures its outputs stay razor-sharp and relevant to your immediate goal. 2. Refine Your Prompting "Improve your prompt, improve your generation." Sloppy, unclear prompts lead to confusing answers. Take the time to be clear, direct, and concise with your instructions—feed the model precision to get precision back. 3. Curate Your Reference Materials When asking an AI to write or analyze content, provide only the specific reference materials it needs for the task. - Avoid connecting models to your entire drive or workspace. - Supplying irrelevant files forces the model to sift through unnecessary data, increasing the likelihood of distractions, confusion, and hallucinations. Try these best practices and let me know if you see an improvement! 👇

Bookkeeping can be the most bothersome, yet important task for a business, especially one with inventory. And depending on how complex your bookkeeping software, it can also create problems if the right person isn't in the driver's seat. This happened to my fashion business - not only did I not have great bookkeepers, I was lured into an Oracle product that was way too complex (and costly) for my business size. Enter AI. Here is how I am automating: 1. We connected Claude to Oracle's Netsuite 2. We created a user account for Claude in Netsuite with strict permissions- read only. 3. We created a "Bookkeeping Project" in Claude, giving it history of the issues, instructions to audit our bookkeeping for the last three years and to report back with findings. I gave it extra instructions to ensure we followed federal and state accounting law and guidelines for the respective years. 4. We backed up our data (taking a snapshot of where things stood prior to AI) 5. We unleashed Claude into Netsuite to do an audit. 6. Within MINUTES, it found duplicate bookings, reverse journal entries that shouldn't be there, poor categorizations, called out things that needed to be fixed, and it even checked with me to ensure I had paid the appropriate taxes and was doing inventory counts. Mind blowing 🤯 We are now working with Claude to make the corrections- we gave Claude further permissions within Netsuite, "bookkeeping", to only make the changes we approve, nothing more. My team and I are the humans in the loop to ensure the work is done properly. Once this is behind us, Claude will have contextual knowledge as to how to handle bookkeeping moving forward. That knowledge lives in the project, but is portable if we need to change models. Do you have a bookkeeping nightmare to solve? Or a Claude / AI in finance workflow to share? We would love to hear 🤗