In this one-off interactive, gamified workshop, we’ll simulate real-world work scenarios at your organisation via a board game, helping you identify and eliminate bottlenecks, inefficient processes, and unhelpful feedback loops.
Workshop Details
A few weeks ago, I opened Claude and said "Moin."
Claude said hi back, used my name, and asked if I wanted to pick up where we left off on a client's transformation slides. It knew I'd been working on a trimester-based Release Package framework. It knew the Flight Levels structure I use. It remembered that I prefer direct, no-fluff communication and that my German emails should be signed with "Benjamin," not "Ben."
I didn't tell it any of that. Not in this conversation, anyway.
Compare this to January 2025, when every single Claude conversation started from absolute zero. Open a new chat, explain who I am, what my company does, what tone I use, what formats I prefer — and then, finally, get to the actual work. Twenty minutes of context-setting for every conversation. The AI equivalent of introducing yourself to your coworker every Monday morning.
The customization features Claude has shipped over the past year have fundamentally changed that equation. Most people aren't using even half of them. Here's the playbook — largely based on what I picked up from Anthropic's own learning platform, plus a year of daily use.
The Three Layers of Making Claude Know You
Think of Claude's customization as three concentric circles. Each layer does a different job, and the real leverage comes from stacking them.
| Layer | What It Does | Scope | Analogy |
|---|---|---|---|
| Memory | Global context about you — who you are, what you're working on, how you like things done. Synthesized from chat history, updated daily. | All conversations | Your colleague's brain |
| Projects | Isolated workspaces with their own instructions, documents, and memory. Separate context for separate domains. | Per project | Your desk at work vs. your kitchen table |
| User Preferences | Explicit settings — formatting, tone, feature toggles, writing style. The control panel. | All conversations | Your IDE settings |
Most people only touch one of these. The magic happens when you use all three.
Memory: Build a Relationship, Not a Transaction
Memory is the feature that changed my daily workflow the most. Before memory, every Claude conversation was stateless. You start fresh, provide context, get output, the conversation ends, everything evaporates. It's like hiring a brilliant consultant who gets amnesia every evening.
With memory enabled, Claude maintains a running synthesis of your conversations. It knows my name, my company, my active client engagements, my communication style, my podcast guests, even that I have a thing for Tolkien references and satire. It picked all of that up from months of conversations.
If you're switching from ChatGPT or Gemini, you can export your memories from those platforms and import them directly into Claude. No starting from scratch.
Treat your Claude memory like you treat your CRM. It's only useful if you keep it current. Go to Settings, look at what Claude remembers about you. Delete things that are outdated. Add things that are missing. Tell Claude directly: "Remember that I always want German proposals formatted with formal salutations." The more curated your memory is, the better every conversation gets.
Projects: Context Boundaries That Actually Work
If you're a consultant, you work across multiple clients. Each has their own terminology, processes, sensitive information. You don't want Claude mixing up your manufacturing client's R&D transformation notes with your automotive client's SAP migration context. Projects keep those worlds separate.
Each project gets its own instruction set, its own uploaded documents, and its own memory. When I start a conversation inside a client project, Claude already has the context about their organization, the frameworks we're using, the specific deliverables. I don't need to explain any of it.
What goes where in a project
| Put This In… | Examples | Why |
|---|---|---|
| Project Instructions | Tone, language, template rules, key constraints, stakeholder names | Short, behavioral. Claude reads this every time. |
| Uploaded Documents | Org charts, past deliverables, framework descriptions, brand guides | Detailed reference material. Don't bloat the instructions with it. |
| Project Memory | Patterns Claude learns from conversations within the project | Stays separate from your global memory. Exactly what you want for client work. |
A practical tip: don't try to put everything in the project instructions. "When creating presentations for this client, use their corporate template and write in German" beats a three-page org chart description every time. Upload the heavy reference stuff as documents. Let the instructions stay focused on behavior.
User Preferences: The Settings You're Probably Ignoring
Open Settings in Claude. Seriously, do it right now. There's more in there than most people realize.
| Setting | What It Controls | My Take |
|---|---|---|
| Capabilities | Toggle web search, code execution, file creation, skills, memory | I keep everything on. Claude is my operating environment, not just a chat window. |
| User Preferences | Global tone and formatting instructions | Set this once, stop correcting Claude's tone in every conversation. |
| Style | Customize Claude's writing style with specific instructions | Worth the time if you produce content regularly. |
| Past Chat Search | Claude can reference previous conversations when relevant | Combined with memory, this is what makes it feel like continuity instead of repetition. |
| Incognito Chats | Conversations that don't feed into memory or history | The private browsing window of AI. Use it when you need to. |
Connectors: Claude in Your Actual Ecosystem
This is where customization goes from "nice to have" to "I can't work without this."
Claude supports connectors to external tools — Google Drive, Gmail, Google Calendar, Notion, Figma, Slack, and more through MCP (Model Context Protocol). These aren't one-directional integrations where Claude can just read your files. They're two-way connections where Claude can search, create, update, and organize things in the tools where your work actually lives.
I have Claude connected to my Google Drive, Gmail, Google Calendar, Notion, and Figma. When I ask it to find a client proposal from last month, it searches my Drive. When I ask it to draft an email, it pulls context from previous threads. When I need to check a design reference, it pulls assets from Figma.
The combination of memory, projects, and connectors turns Claude from a chat interface into an actual work environment. It knows who I am, what I'm working on, how I like things done, and it has access to the tools where my work lives.
That's not a chatbot. That's infrastructure.
The Compound Effect
Setting all of this up takes time. Writing good project instructions, curating your memory, connecting your tools — it's not a five-minute thing.
But here's how I think about it: every minute you spend customizing Claude is a minute you'll never spend re-explaining yourself. It compounds. A well-configured Claude setup in March means thousands of saved keystrokes by December.
I spent a decade telling manufacturing companies to invest in their operating systems — the processes, tools, and structures that make work flow. Now I'm taking my own advice.
Start with memory. Then build a project. Each layer makes the next one more powerful. The AI that knows you is categorically different from the AI that doesn't.
