Article Read Time
Every week I talk to someone who tells me the same thing.
They have Claude Pro. They use it regularly. But they can’t shake the feeling that they’re not really getting what everyone else seems to be getting out of it. So they do what makes sense: they look up better prompts. They read articles about how to ask better questions. They save prompt templates. They try the suggested frameworks.
And it helps a little. But not nearly as much as they expected.
Here’s what I want to tell them, and what I want to tell you: the problem is almost certainly not your prompts. The problem is that Claude doesn’t know who you are, what tools you use, or how you work. And until that changes, no prompt in the world is going to fix it.
The setup problem nobody talks about
Most of the conversation around AI productivity is about prompting. How to ask better questions. How to give Claude more context in a single message. How to be specific enough to get a good answer.
And prompting does matter. But prompting is the last step in a chain, not the first. Before any prompt can do its best work, three things need to be true.
Claude needs to be connected to your real world. Right now, unless you’ve set up connectors, Claude has no idea what’s on your calendar, what’s sitting in your inbox, or what tasks are waiting in your project management system. It’s answering you blind, with only what you type into the chat box as context. That’s a significant limitation, and no amount of clever prompting gets around it.
Your projects need proper instructions. If you’ve set up Claude projects but never written instructions for them, Claude is going into every conversation with no idea who you are in that context. It doesn’t know what you do, how you communicate, what your clients are like, or what you’re trying to accomplish. Again: not a prompting problem. A setup problem.
Claude doesn’t know your voice. When you ask Claude to write something and it comes out sounding stiff or generic, the instinct is to prompt it differently. Tell it to be more casual. Tell it to sound like you. The problem is that telling it once in a single chat doesn’t stick. The next chat starts from zero. Until you install a tone of voice skill, Claude will always default to a generic output that doesn’t sound like you. That’s not a prompting failure. It’s a missing skill.
What changes when the setup is right
I want to be specific here because I think this is where people underestimate how different the experience becomes.
When Claude is connected to your tools, you stop copying and pasting your tasks and emails into the chat. You ask Claude to look at what’s in your ClickUp and tell you what actually needs your attention this week. You ask it to scan your inbox for anything that needs a response. You ask it what’s on your calendar tomorrow and what you should prep for. Claude knows. It just goes and looks.
When your projects have proper instructions, you stop reintroducing yourself every time. Claude already knows what the project is for, who you are in that context, what you need from it, and how it should behave. Every chat inside that project starts from a place of understanding rather than a blank slate.
When your tone of voice is installed as a skill, you stop editing Claude’s outputs to make them sound like you. You type “use my voice” and Claude adjusts. The difference between a generic AI output and something that actually sounds like you is not prompting. It’s training.
None of this is complicated once it’s set up. But none of it happens by itself, and the documentation for how to actually do it is scattered, technical, and not designed for someone who just wants their AI to work properly.
What I’ve seen in practice
The clearest example I can give you is from the first person I ran my Life in Claude session with.
She is a CEO, author, and founder. She runs four professional identities simultaneously. She had been using Claude Pro for months. She had projects set up. She used it regularly. She is one of the most organized and capable people I know.
She had never installed the desktop app. She had no connectors set up. She had zero instructions in any of her projects.
Which meant every time she opened Claude, it had no idea who she was, what she was working on, or how she communicated. She was essentially prompting a very smart stranger every single time and hoping for good results.
In three hours we fixed all of it. Nine connectors installed. Project architecture rebuilt for all her roles. First custom skill built and tested. By the time we closed the call, Claude knew her world.
She called me the next day to book ongoing sessions.
The prompts didn’t change. The setup did. That’s the whole point.
The honest question to ask yourself
If you’re reading this and recognizing your own experience in it, the question worth sitting with is this: how much time have you spent trying to prompt your way out of a setup problem?
Because there’s a simpler explanation for why you’re not getting the results you expected from Claude. It doesn’t know you yet. And until it does, you’re always going to feel like you’re leaving something on the table.
The good news is that fixing this is not as hard as it sounds. It’s a few hours of proper configuration, with the right person guiding you through it, and then it just works.
If you want to understand exactly what that configuration involves, the pillar post in this series breaks down every component of a proper Claude setup and what each one actually does.
And if you’re ready to just have it done properly and stop figuring it out alone, that’s exactly what the Life in Claude workshop is for.
Frequently Asked Questions
Regular Claude chats have no persistent memory. Each chat starts fresh with no context from previous conversations. Claude Projects solve this by storing your files, instructions, and conversation history in a saved workspace, so Claude has consistent context across every chat within that project.
Yes, but only up to a point. Prompting improves your results within a single conversation. But it cannot replace the context that comes from proper setup: connected tools, project instructions, and installed skills. Think of prompting as the steering, and setup as the engine. You can steer beautifully but if the engine isn’t built right, you’re not going anywhere fast.
A tone of voice skill is a file that teaches Claude how you communicate: your vocabulary, your sentence style, what you never say, what you always say, and how your writing feels. When the skill is installed and triggered, Claude applies those rules to any output it generates, so the result sounds like you rather than like a generic AI assistant.
A core configuration covering connectors, project instructions, and two to three workflow skills takes approximately three hours with an expert guiding you through it. Doing it alone from scratch, without knowing what goes where, typically takes much longer and often results in an incomplete setup.
Gina Dunn is the founder of OG Solutions and a brand strategist with 25 years of experience helping female founders find and sustain their authentic brand voice. She works with founders, consultants, and creative professionals globally from her base in the Netherlands. Her Mirror, Not Mask framework is built on one conviction: strategy isn’t invention. It’s remembrance.