Step-by-step guide: Using AI Agents to track humanoid robot funding from scratch

When I first started getting into AI Agents, I was completely clueless. But after trying ABot-World Studio to scrape funding news for robots like Figure AI, I found it simpler than expected, though full of pitfalls. Last week, Bloomberg published news about Bezos, Arnault, and Premji flocking to invest in humanoid robots. I happened to be playing with this tool, so I ran through the process. Today, I'll break down the steps, from registering an account to outputting a report, detailing every step so you can follow along.

First, the pros and cons of this method. Pros: Automation. No need to manually scan dozens of news sources daily. The Agent can automatically extract key info (funding amount, investors, valuation) and generate analysis according to your template. Cons: Configuring data sources requires patience. If the news source URL is wrong or the RSS format changes, it stops working. Also, the free version of ABot-World Studio has usage limits; running too many times requires payment.

What You Need to Prepare

• A computer with internet access; Chrome or Edge browsers recommended
• An email address to register for an ABot-World Studio account
• A URL for a news source you're interested in (e.g., a specific topic RSS from Bloomberg, or a robotics industry aggregator site you find yourself)

Step 1: Register for ABot-World Studio

Open the ABot-World Studio official website (assuming this is the URL; verify yourself) and click "Sign Up" in the top-right corner. Enter your email, set a password, and you'll receive a verification link via email. Click it to log in. After logging in, you'll see a blank Dashboard with a menu on the left and a canvas in the middle. Don't panic; everything else is just clicking the mouse.

Step 2: Create an Agent

In the top-left corner of the Dashboard, click "Create Agent." A window pops up asking if you want to choose a template or start blank. I recommend beginners choose "Blank Agent" because templates come with many preset rules, but you need to understand how it works first. After selecting blank, give your Agent a name, e.g., "Robo Funding Tracker," and write a random description like "Scrape humanoid robot funding news." Click "Create" to generate an empty canvas with a "Start" node.

Step 3: Configure Data Sources

This is the most error-prone step. On the canvas, drag an "RSS Feed" node from the "Sources" section on the left onto the canvas and connect it to the Start node. Double-click this node to open a configuration box. You need to enter the news source URL.

My example uses Bloomberg's robotics topic RSS, but Bloomberg's official RSS requires a paid subscription. You can use free alternatives, such as "TechCrunch Robotics" RSS (https://techcrunch.com/tag/robotics/feed/) or "The Robot Report" RSS. I used a fake example URL here, but the operation is the same.

Fill in the URL in the "Feed URL" field, e.g., https://example.com/robotics-news.xml. Then click "Test Connection." If it returns a green "Success," it means data can be read. If it errors, 99% of the time it's due to incorrect URL format or the website blocking crawlers. Pitfall 1: Many news sites restrict RSS feeds, requiring a User-Agent header. But the free version of ABot-World Studio doesn't allow changing this, so it's recommended to use well-known RSS aggregators like "Feedly"'s public RSS, or "Google News" RSS (append ?hl=en-US&gl=US&ceid=US:en to the URL, but Google News RSS is unstable).

If the test fails, switch domains. I tried TechCrunch's RSS, and it worked.

Step 4: Extract Key Information

Once the data source is connected, tell the Agent what to extract. Drag an "Extract" node from the "Processors" section on the left and connect it to the RSS node. Double-click the Extract node to enter the configuration interface.

Here you need to write simple extraction rules. ABot-World Studio supports XPath or CSS selectors similar to HTML parsing, but beginners don't need to understand them. It has a "Smart Extraction" mode where you just input the field names you want to extract, such as "Funding Amount," "Investors," "Valuation," and click "Auto-Detect." It scans the first 5 news items and recommends possible extraction paths. However, in practice, auto-detection is often inaccurate, especially when titles and body text are mixed.

Pitfall 2: News titles often contain numbers like "raised $67.5 million," but formats vary—some say "$675 million," others "6.75 yi USD." Extraction rules are best done with regex, but beginners can start with "Contains" keyword matching. For example, to extract "Funding Amount," set the rule to "Keyword contains '$' or 'yi USD'" and associate it with the "Funding Amount" field. Simpler method: In the "Extract" node, select "Template" mode and choose a preset "Funding Round" template. It defaults to grabbing amounts, investors, and rounds. I tried it; for the Figure AI news, it extracted "$675 million" and "Bezos, Nvidia, OpenAI," but missed "Microsoft" because the template only grabs the first three entities. Manual adjustment is needed.

Step 5: Run and View Results

After configuring, click the "Run" button in the top-right corner. The Agent starts executing. The progress bar takes about 30 seconds to 1 minute, depending on the volume of data returned by the news source. After completion, an "Output" panel appears below the canvas, displaying the list of extracted results. Each news item is a row, with fields defined in the Extract node.

If the result is empty, it's likely because there are no new articles in the RSS feed, or your extraction rules don't match. You can click the "Test" button to debug a single item, entering a news URL to see if it extracts correctly. I tried the original news link for Figure AI (found from Bloomberg). The extracted "Funding Amount" was "6.75 yi USD," but the "Investors" field listed "Jeff Bezos’s Explore Investments" and "Nvidia Corp." instead of just "Bezos" and "Nvidia." The fields were too long and required subsequent cleaning.

Pitfall List

• Invalid URLs: Some RSS addresses are temporarily valid and return 404 after a few days. It's recommended to use stable aggregation sources, like "The Robot Report" RSS, which I've used for 3 weeks without interruption.
• Inaccurate Extraction Fields: Both "OpenAI" and "Microsoft" appeared in the news body, but the template only grabbed the first two. Solution: Add the "Include All Entities" option in the Extract node, or manually write regex (?i)(bezos|nvidia|openai|microsoft|amazon).
• Free Quota: The free version allows only 5 runs per day, processing max 10 news items per run. For batch monitoring, you need to upgrade to the paid version ($29/month).
• Chinese News Support: I tried switching TechCrunch to a Chinese "Robotics Industry Network" RSS. Smart Extraction basically failed because Chinese word segmentation and entity recognition are poor. It's recommended to stick to English news.

What to Try Next

You can now automatically scrape humanoid robot funding news. Next, you can connect this Agent to a Telegram bot or Discord channel for daily automatic pushes. Specific method: Add a "Webhook" node after the "Output" node in ABot-World Studio, fill in your Telegram Bot Token and Chat ID, and you can achieve online push notifications. Going further...