Original video in Chinese.
Key Takeaway
- Claude’s Analysis tool can help users analyze data tables and present them visually.
- Built on Claude’s powerful coding capabilities and Artifacts, this tool can handle CSV files and visualize data.
- AI applications in data analysis, like AI coding, will empower more non-professionals.
- Claude’s data analysis capabilities give it a clear advantage in “serious work” scenarios, with use cases in marketing, sales, IT, and more.
- The article emphasizes that “serious work” is the most valuable scenario for AI application deployment.
AI isn’t just here to take programmers’ jobs anymore — now it’s coming for data analysts too. Claude’s Analysis tool, which launched a few days ago, is especially useful. If you have a data table and want AI to help you analyze it and visualize the results, definitely try this feature.
Right now, this analysis tool is still in preview mode and isn’t enabled by default, just like Artifacts was at the beginning. So you need to go into Feature Preview, check it, and then you’re good to go.
Analyzing data tables has always been something many people desperately need, but it’s also a weakness of large language models. Top products like Claude can force their way through it, but the results are mostly summarizing and fairly high-level; they still can’t do more detailed, precise analysis. This new data analysis tool fills that gap. It is built on two foundations:
First, coding ability. Claude’s coding ability is currently recognized worldwide as the strongest, no exception. So it can directly use JavaScript to read, parse, and restructure user-uploaded CSV files, just like a human data analyst would handle them. And if any errors come up during processing, it can fix them itself.
Second, Artifacts. For Claude and all chatbots, Artifacts is a very important innovation. It opens a dedicated window separate from the main conversation window. All content generated in response to the user’s request is displayed in that dedicated window. To ensure safety, it also uses sandbox-like technology to create a safe playground.
So Claude’s stunning performance in data analysis all comes from its far-ahead foundational capabilities.
Hello everyone, welcome to my channel. Humbly speaking, I’m one of the few bloggers in China who can explain the Why and How of AI clearly. Remember to follow. As long as one video really clicks with you, it’s totally worth it. If you want to connect with me, come to the newtype community. More than 600 friends have already paid to join!
Back to today’s topic: Claude Analysis tool.
I mentioned in the community before that the value of AI is not just reducing costs and giving bosses an excuse to squeeze employees harder; more importantly, it brings new capabilities. Many people have already enjoyed the value of AI coding. Next, data analysis will also be lowered to more people through AI, just like coding. From this perspective, Claude is truly doing great merit.
To demonstrate the new data analysis feature, I downloaded a set of Bitcoin historical data, including price, trading volume, and so on. The format seems to have a small issue, but I didn’t inspect it closely. Anyway, just drag it directly into the chat box and let AI handle it.
After receiving the request, Claude first processes the data, then starts the Artifacts feature and uses code to create a visual display.
You may have only used Excel for data analysis before. But Excel can only handle some simple tasks. For more professional work, you’d use Python, whose pandas and numpy are both very popular data analysis libraries, each with its own strengths. If statistics are involved, there’s also R. Its learning curve is much steeper than Python’s. Of course, no matter which one it is, for us, mastering it is not something that happens in a day or two. That’s where AI comes in.
Back to Claude. You can see that it has already finished the visual presentation and reconstructed a price chart from the data. Not only that, it also proactively made some preliminary conclusions, finding that Bitcoin is highly volatile, is currently in an upward trend, and that the price is in a historically high range.
Take a look: after pausing briefly, Claude also gives three suggested questions. This design is clearly meant to guide ordinary users to continue analyzing. Otherwise, many people would look at all this and be totally confused. So let’s have AI keep going and create an interactive price chart.
Probably because this feature is still in preview, there will be some errors. But that’s fine — let Claude fix them itself. Or if there’s something you’re not satisfied with, you can say it directly. For example, I wanted it to stretch the time period as long as possible. Then Claude will go back, review the data, adjust the code, and ultimately display an analysis over a longer time span.
These errors from Claude are all minor issues. At this stage, I think there are two things that most need improvement: first, the rate limit, which is currently too low and gets used up in no time; I hope the official team can raise it soon. Second, the suggested questions I mentioned earlier are too weak and don’t really guide the user.
Through this demonstration just now, you can see that Claude’s understanding range has expanded from text to data. In terms of text, I’ve done comparisons before, and Claude’s logic is significantly stronger than ChatGPT’s. Now it has also added precise data analysis. In this way, Claude firmly occupies the serious work scenario.
If you do marketing, you can upload customer interaction data; if you do sales, you can upload the performance of each sales region; if you work in IT, you can upload server logs. Claude can handle all of these and help you with all of them. That’s worth far more than helping a white-collar worker write a little Xiaohongshu copy or come up with a brand slogan.
I believe serious work is definitely the most valuable scenario for AI deployment, and it’s also the one most AI companies are ignoring right now.
OK, that’s it for this episode. If you want to connect with me, come to the newtype community. See you next time!