There’s a version of Claude that writes in your voice, remembers your instructions, searches your content library, and builds working tools inside the conversation. Most marketers are nowhere near that version. These five capabilities are the gap — each one with a concrete prompt to start today. Some of them sound more technical than they are. I’m not technical at all and I use every single one of them. So can you.
1. Build a version of Claude that writes like you
My best advice for training AI on your voice has always been: find five good examples of your writing, feed them to Claude, let it analyze your style. It works — when you have the right examples. Most people don’t. And even when they do, half of them were already edited by AI.
There’s another method: let Claude interview you instead.
You answer 20 questions on your worldview, how you work, what you’d never write, what you’d defend in an argument. Claude compiles the answers into a Voice DNA document that captures your positions, not just your sentence rhythm.
Find the full interview prompt here.
Set aside 30–50 minutes. Answer specifically. The more concrete your answers, the more useful the output.
Do this first. Everything else on this list works better once Claude knows who you are.
2. Save any instruction once and reuse it with Skills
The Voice DNA document you built needs a permanent home — a place where Claude can access it automatically.
That’s what Skills are for. And the same applies to any other instruction you want Claude to follow consistently: a copy checklist, a scoring framework, a briefing format.
What is a Skill
A Skill is a plain text file saved in a specific format called Markdown — with the extension .md instead of .txt or .doc. Claude reads .md files natively, which is why Skills use that format.
You don’t need any special software to create one. Here’s the simplest way:
Open a Google Doc or any text editor
Write your instructions in plain English
When you save the file, name it something like brand-voice.md — the .md at the end is all that matters. Or save it as a .doc in Google Drive and download it as an .md file
That’s it. No coding, no formatting rules to follow. Plain sentences work fine.
How to upload a Skill to Claude
You have two options, and which one you choose depends on how widely you want the Skill to apply.
Option 1: Upload at the account level
This makes the Skill available across all your conversations, in any Project.
Go to Settings → Capabilities → Skills and upload the .md file.
Option 2: Upload inside a specific Project
This keeps the Skill contained to one Project — useful when a Skill is client-specific or task-specific and you don’t want it bleeding into unrelated work.
Open the Project → click Set project instructions → paste or upload your Skill there.
What’s the difference between a Skill and a Project?
A Project holds context — who you are, what you’re working on, which files Claude should know about. Think of it as the briefing room.
A Skill holds behavior — a specific way of doing a task. Think of it as a trained specialist you bring in for a particular job.
You can have the same Skill installed across multiple Projects. Or a Skill that only lives in one. The two work together, not instead of each other.
3. Give Claude access to any database with an MCP connector
Beyond what it’s been trained on, Claude only knows what you tell it in the conversation. You paste an article, it reads that article. You upload a document, it reads that document. It has no access to anything sitting in your Google Drive, your content library, or any external database unless you manually bring it in.
MCP connectors change that. They give Claude a live connection to an external data source — a database, a content archive, a tool like Notion or HubSpot, so it can search and pull from it directly, mid-conversation.
The setup to an external library takes two minutes and requires no technical knowledge.
Here’s a concrete one to try today.
Lenny Rachitsky released 320 podcast transcripts as raw data — conversations with April Dunford, Seth Godin, Brian Chesky, and others across product, growth, and marketing. With an MCP connector (created by Akshay Chintalapati and available for free on GitHub here), you query all 320 interviews in natural language.
To connect it in Claude:
Go to Settings → Connectors → Add custom connector
Name it “Lenny’s Podcast”
Paste this URL: https://lenny-mcp.onrender.com/mcp
Toggle it on
Then open a new conversation and ask: “Search Lenny’s podcast for what guests have said about content strategy for early-stage companies.”
That’s one connector. The same setup works for Notion, Google Drive, HubSpot, and a growing list of other tools. Once you’ve done it once, you’ve understood the pattern.
4. Turn a complex idea into an animated visual
Claude can build animated, interactive diagrams directly inside the chat. Most marketers don’t know this exists.
Process flows, framework maps, funnel visualizations — the kind of thing you’d normally hand to a designer or spend an afternoon in Canva building. The output is an HTML file you iterate on, then export as a GIF ready to post.
The workflow:
Step 1
Finalize your content before you touch the tool.
Categories, hierarchy, order. Every content change forces a visual rebuild. Agree on the structure first.
Step 2
Ask Claude to build the diagram.
Describe your content and let it propose the layout:
“I want to create an animated diagram showing [your topic]. Here are the elements: [list them]. Help me figure out the best layout and visual structure, then build it as an HTML file.”
Step 3
Refine one thing at a time.
Font sizes, spacing, order, hierarchy. One prompt per change. Fixing multiple things at once creates regressions.
Step 4
Export to GIF:
“Convert this HTML file into a GIF by capturing it frame by frame using a headless browser. It should play once and freeze on the final frame.”
The “play once and freeze” instruction matters — a looping diagram is distracting. You want it to animate in, land on the final state, and stay there.
5. Turn a brief into a working tool without leaving the conversation
Most marketers use Artifacts for writing — drafts, outlines, email copy. That’s the least interesting thing you can do with them.
Ask Claude to build something instead of write something. A content calendar with working filters. A budget calculator that updates in real time as you change inputs. A scoring matrix for evaluating campaign ideas. Things you’d normally build in Excel — except you’re still inside the conversation where you just mapped out your strategy. The context is already there. The prompt is the brief. The Artifact is the deliverable.
Try this prompt now:
“Build me a monthly content calendar template. Include columns for date, channel, content format, theme, title, status, and owner. Themes should be limited to 5 max: Product, Culture, Education, Growth, Community. Add a dropdown for status with these options: Idea, In Progress, In Review, Published. Add a filter by channel at the top so I can view one channel at a time — channels: Instagram, LinkedIn, X/Twitter, Blog, Email, YouTube. Make it visually clean and easy to scan.”
Here’s what the result could look like.
If the first version is buggy, type: “What you described is what I want, but what I see is a buggy version of it. Review and fix.” It usually resolves on the second attempt.
From there, iterate the same way you would with a colleague.
“Make the status column color-coded”
“Add a notes field”
“Change the theme options
Each instruction updates the Artifact in real time.
These five features won’t all be relevant to you at the same time. Start with the one that solves your most immediate problem. The Voice DNA interview if you’re tired of generic output. Skills if you’ve already done that but keep re-explaining yourself. An MCP connector if you work with a specific research source regularly. The payoff compounds — each one makes the next more useful.
