Before you roll out a business app to your team, check what each person can see. One owner I know was setting up a prebuilt app for his team. He checked the permissions before sharing it and found that every user could see every financial record. Checking early let him choose a different setup before the rollout. Each person should see only the records their job requires, and the owner should see everything. First, ask who needs to see payments. Second, ask who needs to edit records. Third, ask who only needs to see their own tasks. Many tools put detailed permissions on a higher pricing plan, so check the plan details early, right? If a prebuilt tool covers your permissions, it's usually the faster and cheaper option. At @LowCodeAgency, we set up access by role in custom builds, so each person sees what their job requires. I'm happy to talk it through when you DM me "permissions."

PHI is any data that connects a diagnosis to an identity. A name and a symptom in the same database table. An address linked to a condition a user mentioned in a wellness form. The rule is about the connection, not the category. The federal rule is the floor. Some states build a ceiling above it. Most teams building in California are operating under federal rules alone and have never looked at what the state layer requires.

The destination is an AI companion you can actually work with, not a chatbot, but something you message the way you'd message a capable colleague. It's achievable. The organizations that get there didn't skip the first two steps: context and information infrastructure, before training. Skip them and you get a chatbot.

Quick math for an example company with 20% turnover: one in five people leave every year, and each one has a laptop, a phone, and a key card. That means 20% of your equipment needs to find its way back every year. If 1 in 10 of those items goes unaccounted for, that is 2% of your total equipment, and about 10% after five years. (Round percentages for the example.) 2% never feels big enough to fix the process, so teams live with it for years, like a slow drip in a big house. We built an issuance app for Rising Ground, a New York nonprofit we work with, where every item has an owner, a status, and a history. DM me your turnover rate and I will walk you through your math.

The most expensive software is the kind that doesn't reach full adoption after launch. That's why we do consulting and process work before development. Map the workflow, train the team, then build. The highest adoption projects all start this way.

A hallucination is an AI stating something false with full confidence. It predicts the next likely word. It doesn't look facts up. You can't turn it off. Give it the real source, let it say "I don't know," and check the answer against the source.

Most of our clients come to us at the same point. They've seen AI work for them personally. They know what it needs to do. Now they want to extend it to the whole team and need someone who can build it around how they actually operate. Our job is to take what they know and build the workflows that will stick. They bring the context. We build the system.

the to-do tab for slack is the best thing ever since 2020

There are two kinds of companies when it comes to software investment. Companies that stack tools and companies that build infrastructure. A company that stacks tools ends up with a collection of products, each with their own data, their own logic, their own login. The stack gets heavier. Nobody has a complete picture of anything. A company that builds infrastructure ends up with a foundation. Each layer connects to the one before it. The data compounds. The question that separates them is pretty simple. What am I building toward, and does this layer get me closer to that?

More business owners now arrive at a first call with an app plan they drafted using an AI assistant. One owner I talked with used an AI assistant to outline the data, pick a platform, and estimate a timeline for the app, which is pretty cool, right? An AI assistant can take you a long way on a first draft, and a builder still needs to check the data design, the integrations, and the security before the build starts. If you don't have a plan yet, we build one with you during discovery, which is the planning phase before the build. First, we map how your team works today. Second, we put together wireframes, which are simple layouts of each screen. Third, our tech lead estimates the build, so you know the scope, the timeline, and the cost. If you already have an AI assistant draft, bring it along. Sometimes an off the shelf tool covers what you need, and I'll tell you that on the call. At @LowCodeAgency, I'm happy to help you build the plan, so DM me "plan."

Every inbound question needed a human to respond. Every after-hours question waited until morning. We built an AI agent trained on verified Mexican legal sources, live 24/7, with lead capture built into every conversation. Franklin 90 days in: "You guys have delivered. We're getting the leads."