Why Everyone Is Switching to Claude (and What You're Missing)

Right now, all anyone talks about is Claude as the best AI ecosystem. I was skeptical about Claude's advantage vs the other ecosystems (models).

Over the last few years, every 3 months, we get new features on Gemini, OpenAI, Claude or Grok. And for me, history kept repeating itself:

Even though I had used Claude some time ago, I mainly use ChatGPT, I had decided to bank on the most widely used tool on the market.

And then we decided to invest time in Claude to see the differences. And that's exactly what this edition is about: is the hype real? What extra can you get out of Claude?

A reminder of the existing limits

Just before I share Claude's big pluses, you need to keep the tool's limits in mind:

• no image / video generation
• no voice dictation to prompt out loud (fairly frustrating)
• no advanced search, aka deep research (only classic web search)
• usage limits (even when paying), daily and weekly

Now that that's said, I'll share Claude's main advantages, we'll break this into 4 main parts:

1. Claude chat: Claude's conversational AI model
2. Claude Cowork via the Claude Desktop app, which carries out tasks on your computer
3. Claude for Excel and PowerPoint
4. Claude Code: Claude's AI dev agent

In each section, I illustrate the main strengths of these 4 tools.

1/ Claude chat

The chat space doesn't change much, I'll focus on the big differences you don't find elsewhere and what makes Claude so well liked.

I recommend regularly enabling extended thinking and choosing the Opus 4.6 model, which is the most powerful and will give you better results

Personalization

You can go further than the other solutions in personalization. There are 2 main options:

1. The capabilities (skills) = a complete file that shares the context on a specific task. A kind of SOP for the AI to follow. Example: “writing guide / branding / …”Pro tip: you can ask Claude to create skills for you once you've shared enough context on a topic. This will let Claude easily redo the same task later.In the chat, Claude spells out its use of skills as you can see in this example for generating a document:

Skills are a file in markdown format with a summary that Claude relies on to decide whether to activate it or not, then the full content below.

Note that some “turnkey” skills exist (PPT, Excel, PDF, ..) and can be imported directly.

There's a composability logic = you can combine several skills to create complex workflows:

1. The connectors = integrations between Claude and your tools. It gives Claude tools at its disposal to answer your requests, like retrieving context from your tools (rather than doing it by hand). You can go further, all the way to editing information in the tool in question. On that last point, I advise against it given the possible hallucination risks, unless you create, for example, skills to correctly fill in a deal in your CRM. As you can imagine, there's a strong incentive to document as thoroughly as possible how you operate so that Claude follows the same logic across the skills.
   There are 50+ connectors on Claude with the option to add your own connector (=MCP).
   To sum up, by adding a connector, you give Claude the ability to carry out all the possible actions (via API) on the tool in question. Under the hood, it's MCP (to dig into the MCP topic, I invite you to take a look at this newsletter edition)

An example here with the Fireflies integration and the possible actions for Claude (you control the type of access per action):

Projects in Claude, aka your closed environment

I'll go quickly over this section because there isn't much of interest. Basically, it's comparable to the Project mode on ChatGPT, a closed environment where you can share knowledge and it will affect all your chats in that space.

You can also define custom instructions (a bit like a persistent system prompt) that apply to all your conversations in the project = tone, role, response format, etc.

If you share too much knowledge, then Claude uses RAG in the background (rather than injecting everything into the context). All your "project" conversations are kept and reused within the project itself.

Note: each project is completely isolated from the others, so no cross-referencing between your different spaces.

But it's not a game changer (unlike the skills and connectors above, or the artifact)

Artifacts: Claude's real visual plus

I find it's one of the features that makes a big difference compared with the alternatives.

When you ask Claude to create something visual, a dashboard, a diagram, a landing page, a document, it directly generates HTML/CSS in a panel next to the chat. And the result is often impressive.

Where ChatGPT will output raw text or a fairly basic canvas, Claude spontaneously produces interactive visual renders (comparable to a kind of Google AI Studio). The render is clean, sometimes interactive (buttons, filters, navigation…), and above all, it's usable immediately.

I use it a lot to generate sales proposals with dedicated skills that frame the format, the branding, the structure. All I have to do then is export to Google Drive. In a few minutes, I have a client deliverable ready to send.

Pro tip: artifacts can be downloaded directly from the interface. You can grab them as HTML or copy them to embed them wherever you want.

You can also generate interactive charts (underrated and still little used). Claude generates manipulable visualizations: you can test scenarios, vary assumptions, and see the impact. Example here:

Pro tip: spell out that you want an interactive chart and not to code, to avoid ending up with code

Security and memory: two underrated points

On the security side, one thing I appreciate about Claude is that it explicitly asks whether we allow our data to train the models. On ChatGPT, it's enabled by default and many users don't even know the option exists.

On the memory side, it's disabled by default, and I encourage you to enable it. Very useful for Claude to pick up context about you over the course of conversations, your preferences, your way of working, your ongoing projects.

And if you're coming from ChatGPT, Anthropic has even provided an import tool. In a few minutes, Claude knows you as if you'd been using it for months.

2/ Claude Cowork

I'll keep it fairly concise on Claude Cowork, we could easily do a whole edition on it.

For me, Claude Cowork is probably one of the most underrated features of the Claude ecosystem. However, it happens on Claude's desktop app.

The idea is simple: Claude no longer works only from a prompt in a chat. It works directly from your files, your folders, your context, and can produce real deliverables.

You give it access to a folder on your computer (here COWORK on the screenshot), it reads the documents, understands the context, asks you questions if needed, then produces a directly usable result. In this folder, you share with it all the instructions in markdown format (.md), comparable to the skills we saw above.

An example here where Claude Cowork creates a media kit for newsletter sponsoring, with slide creation, retrieval of info from Substack (open rate, number of subscribers, etc)

It's a bit as if it had your eyes to gather context (if you have windows open, it can grab the content from them) and your trackpad to carry out actions.

It's comparable to Claude Code (see below) but for non-technical profiles.

Concretely, you can use it to:

• write a deliverable from a brief and source documents
• analyze a set of files or notes
• produce a document, a spreadsheet or a working document
• work from your templates, your tone, your past examples

Here, it generates a complete business plan for an app idea with market analysis, technical feasibility, etc. The document is directly exportable and editable on Google Drive for example (via the connector).

Take this example where Cowork retrieves all the podcast episodes to analyze them and extract the most relevant takeaways for a specific audience (product builders)

The real paradigm shift here is that the subject is no longer really the prompt. It's the context. With Cowork, the better you structure your files, the more relevant Claude becomes. It takes a bit of time to set up.

I made this LinkedIn post which gives you all the necessary steps on Cowork to do your setup: (you can save the post)

We move from a logic of prompting well to a logic of organizing / documenting your working environment well so that AI works with us.

My advice: create a dedicated folder with a few key files = about-me.md for your context, your roles, your goals + voice.md for your tone, your style, your examples + rules.md for your constraints.And ask it to ask you questions before executing, it changes a lot. (Rather than heading in the wrong direction with rough assumptions, you clarify what it's missing)

It's typically the kind of tool that becomes very powerful as soon as you often work on deliverables, research, analysis, or document production.

I'll keep it brief here, Claude Cowork almost deserves to be treated as a tool in its own right.

But above all, remember that it's not just another chat, it's a workspace where Claude can collaborate with you.

3/ Claude for Excel and PowerPoint

Claude has developed an extension for Excel and PowerPoint and it's quite impressive. You've probably tried to use AI to manipulate Excel / CSV files, it was very complicated. Lots of hallucinations.

Claude, directly in Excel, is super powerful. I've shared this cheat sheet with prompts and the possibilities.

I'll slip in a concrete example of a financial model on Apple generated in a single prompt by Claude:

The real gain is at least 1 to 3 hours a week for a regular user (without pushing it). But it doesn't replace the verification work that's still necessary.

For finance teams, it's a no-brainer. Less time wasted, more decisions.

And once you've worked on your models, you can easily bridge to PowerPoint to switch to slides.

4/ Claude Code

To avoid going too long on Claude Code, which could deserve a dedicated edition, I'll just share the broad outlines.

Claude Code is Claude's AI agent capable of creating, editing and running code to build tools for you. It can be scary for the uninitiated, but there's now a web version, more accessible. (rather than a terminal or an IDE). And in some cases, you don't even need to go through GitHub (particularly if you work locally with the Remote Control mode). And you don't even need to log into a GitHub account.

The real shift is that you don't need to know how to code. You describe what you want in natural language, Claude writes the code, creates the files, and can even deploy the result.

Concretely, you can:

• create a website or a landing page
• generate a dashboard or an internal tool
• prototype a product in a few minutes
• or simply brief a developer better with something concrete

Here's an example to generate task automation scenarios with n8n.

In a few prompts, Claude is able to structure the entire logic of a workflow, with the right nodes, the right conditions, and a directly usable foundation.

💡 Pro tip: think like a project manager → describe the expected result, not the technical steps.

The real value here isn't replacing a developer but drastically reducing the time between an idea and something concrete.