Four ways people come to this. The build changes depending on whether it’s just you or a team, and whether you’re starting clean or working with what you’ve already got.
Just me. Solo or running a small company, with nothing much in place yet.
One system where your data lives. Notion, Supabase, Obsidian, depending on what you’re doing. Then Claude or GPT plugged into it, so you can ask for something instead of going and finding it. Most days you won’t open the system at all. When you do, it’s organized.
Plus: a starting set of agents for the things you do every week.
Just me. Solo or running a small company, with some things working but hitting a wall.
You’ve got a couple of spreadsheets Claude can read and some things dialed in. What’s capping it now is your own process. This is skills instead of one-off prompts, scheduled runs that pull from the tools you already use, and a look at where your data lives.
Plus: if you don’t do something the same way every time, it can’t be automated reliably. Some of this is fixing the process first.
A team that knows it should be using AI, and isn’t sure it is.
Most teams are already using AI. Everyone’s just using it differently, and nobody’s sure it’s being used well. One customer care team was pasting Claude’s answers straight into client emails, em dashes and all, which is worse than not using it.
Plus: training on what to use it for and what to leave alone, a shared library of skills so everyone works the same way, and an internal champion who keeps it moving after the build is done.
A team past the basics, running into the limits of what it built.
You already know the difference between an automation, a skill, and a connector. Custom connectors into the tools your data sits in, Claude Code running in your own repo with hooks for your workflows.
Plus: technical training for your internal champion, and a cleanup pass on what’s already there.
Stakt Consulting
© 2026 Stakt Consulting. All rights reserved.