<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Shawn Cao on Medium]]></title>
        <description><![CDATA[Stories by Shawn Cao on Medium]]></description>
        <link>https://medium.com/@shawncao?source=rss-b7aa29e347a4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*NV57KPW7hq0R2LeoMAL6cw.jpeg</url>
            <title>Stories by Shawn Cao on Medium</title>
            <link>https://medium.com/@shawncao?source=rss-b7aa29e347a4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 04:54:14 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@shawncao/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Inbox to Insight: Stop Copy-Pasting Invoices and Let Your Gmail Do the Work]]></title>
            <link>https://shawncao.medium.com/from-inbox-to-insight-stop-copy-pasting-invoices-and-let-your-gmail-do-the-work-500e50af26c7?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/500e50af26c7</guid>
            <category><![CDATA[workflow-automation]]></category>
            <category><![CDATA[dataflow]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[data-capture]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 20:48:19 GMT</pubDate>
            <atom:updated>2026-11-17T20:48:19.984Z</atom:updated>
            <content:encoded><![CDATA[<h4>How invoice automation turns a repetitive monthly chore into a few clicks, and a live dashboard.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*32vsZgEF6KArpkZGije_QQ.png" /><figcaption>A workflow begins at your Inbox</figcaption></figure><h3>A monthly chore?</h3><p>Business users usually know this as <strong>invoice data capture</strong>: getting the data out of invoice attachments and into a structured, usable form without retyping it. You will also hear it called <strong>email-to-spreadsheet automation</strong> or <strong>accounts payable (AP) automation</strong>.</p><p>Whatever the label, the goal is the same: invoices arrive in your inbox, and clean, analysis-ready data comes out the other end.</p><p>If you have similar monthly chore to run, read on!</p><h3>The traditional way: download, open, type, repeat</h3><p>If you handle invoices for a business, you probably know this routine by heart:</p><ol><li>An invoice arrives in Gmail as a PDF or image attachment.</li><li>You open the email, download the attachment, and rename it so you can find it later.</li><li>You file it in a folder on your drive, hoping the naming stays consistent.</li><li>You open the invoice, find the vendor, date, invoice number, line items, tax, and total.</li><li>You type or paste each value into a spreadsheet.</li><li>You double-check the numbers, because one wrong digit throws off the totals.</li><li>Next week, or tomorrow, you do it all again.</li></ol><p>None of these steps is difficult. That is exactly the problem. The work is <strong>tedious</strong>, <strong>repetitive</strong>, and easy to get <strong>wrong</strong>, and it grows with every new vendor and every new month.</p><h3>The hidden costs</h3><ul><li><strong>Time.</strong> Even two or three minutes per invoice adds up quickly. At 50 invoices a month, that is hours spent on data entry alone, before any analysis happens.</li><li><strong>Errors.</strong> Manual entry invites typos, skipped rows, and duplicates. Fixing them later costs more than avoiding them.</li><li><strong>Stale data.</strong> Your spreadsheet is only as current as the last time someone updated it. Reports built on it are always a little behind.</li><li><strong>Fragile process.</strong> When the one person who knows the spreadsheet is out sick, the process stalls.</li><li><strong>Wasted talent.</strong> Skilled people spend their days acting as human copy machines instead of reviewing spending, negotiating with vendors, or planning ahead.</li></ul><h3>The new way: save once, and everything else follows</h3><p>With Columns, the same workflow collapses into a few clicks, and then runs on its own.</p><p><strong>Step 1: Save attachments from Gmail.</strong> Install the <a href="https://workspace.google.com/marketplace/app/columns_drive/946167755859">Columns Drive add-on for Gmail</a>. When an invoice arrives, you save the attachment to your Columns Drive straight from the email. No downloading, no renaming, no manual filing.</p><p><strong>Step 2: Columns Drive reads and compiles the data.</strong> Columns Drive automatically parses the invoices and merges the data into a compiled CSV for each folder. As new invoices are added, the CSV is kept up to date. There is no copying and no retyping.</p><p><strong>Step 3: Turn the folder into a live source.</strong> From that folder, you can create a Columns Flow, using the folder as a live data source. Because the source is live, anything new that lands in the folder flows through automatically.</p><p><strong>Step 4: Build your pipeline.</strong> In Columns Flow, you shape the data the way you need it: clean up fields, group by vendor or month, calculate totals, and format everything for its destination. The best part is that you simply ask for what you desire!</p><p><strong>Step 5: Visualize and export.</strong> Send the results to a dashboard to see spending at a glance, or export the formatted data for your accountant, your finance system, or your next board report.</p><p>The only thing you do by hand is the first step: <strong>saving the attachment</strong>. Everything after it is automated.</p><h3>Side by side</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/757/0*WpgQDm_II7CAKYMt.png" /><figcaption>Quick Comparison</figcaption></figure><h3>Why it is worth your time</h3><p>The value of automation is not only the minutes saved on each invoice. It compounds.</p><ul><li><strong>Hours back every month.</strong> The time you used to spend on data entry goes to work that needs judgment.</li><li><strong>More reliable numbers.</strong> Fewer manual touch points means fewer chances for mistakes.</li><li><strong>Always-current reporting.</strong> Your dashboard reflects what is in your folder today, not what someone last remembered to update.</li><li><strong>A process that scales.</strong> Twenty invoices or two hundred, the effort for you stays about the same: save the attachment.</li><li><strong>Less dependence on one person.</strong> The workflow is built once and runs the same way for everyone.</li></ul><p>For a small business, that can be the difference between dreading month-end and finishing it early. For a growing team, it means invoice volume can double without doubling the workload.</p><h3>Not just invoices</h3><p>Invoice is one of the most common data we saw, in fact, Columns Drive handles all kinds of data in PDFs or Images, you can use a preset template to setup a folder or you can set it up completely by yourself — add a description and fields of interest.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/416/0*mB20H5dbIFm3iwAY.png" /><figcaption>Support All Types of Data Files</figcaption></figure><h3>Try it today</h3><p>You do not need to redesign your finance process to get started. Begin with one folder and one inbox:</p><ol><li><strong>Install the Columns Drive add-on for Gmail</strong> from the <a href="https://workspace.google.com/marketplace/app/columns_drive/946167755859">Google Workspace Marketplace</a>.</li><li>Open your next invoice email and <strong>save</strong> the attachment to Columns Drive.</li><li>Watch the data appear in a compiled CSV, then build your first Columns <strong>Flow</strong> from that folder.</li></ol><p>The next time an invoice lands in your inbox, you could be one click away from the end of manual data entry. Give it a try with your very next invoice.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=500e50af26c7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The 10-Day Battle Against a Sybil Attack on Fina]]></title>
            <link>https://shawncao.medium.com/the-10-day-battle-against-a-sybil-attack-on-fina-e90930787b4f?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/e90930787b4f</guid>
            <category><![CDATA[api-abuse]]></category>
            <category><![CDATA[fina-money]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[sybil-attack]]></category>
            <category><![CDATA[finance]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Mon, 14 Sep 2026 02:03:51 GMT</pubDate>
            <atom:updated>2026-09-14T02:03:51.130Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*aIuL1Yirzr_8hOJ1K65C5A.png" /><figcaption>fake accounts</figcaption></figure><p><strong>CAUTION:</strong> This article is about security, and it gets a bit technical. For some people, it may feel a little boring. 😅</p><blockquote><strong>Sybil Attack:</strong> A concept from network security where an attacker creates a large number of pseudonymous identities to gain a disproportionately large influence or drain resources from a system.</blockquote><h4>Round 1: The Hacker Created 5,400 Fake Accounts in One Hour</h4><p>Five days ago, I opened the Fina dashboard while sipping my morning coffee and noticed something strange: <strong>5,400 new accounts had been created in Fina.</strong></p><p>If this were real, it would be a huge overnight success for a <a href="https://www.fina.money/?ref=sybil">finance app</a>. Even though we believe we have built a pretty good product, we had never experienced anything close to that kind of growth.</p><p>So my instinct immediately told me: <strong>something went wrong.</strong></p><p>I started peeking through the emails of the latest accounts. Almost all of them used the same email domains: first emalupe.com, and later uberip.com.</p><p>It didn’t take long to search for them and find out that they were throwaway email providers.</p><p>It was not difficult to figure out what happened: someone had created a huge number of fake email addresses and was using Firebase’s public API to create accounts in Fina.</p><p>And yes, by default, a simple API call can create an account in your system when you expose Firebase Authentication to the client.</p><p>So why is the Firebase API key public? How did the hacker get it?</p><p>Actually, there is nothing particularly secret about the Firebase API key. It is normal to put Firebase configuration in your web client. In our case, it was inside our React application.</p><pre>const firebaseConfig = {<br>  apiKey: &quot;AIzaSyYourPublicAPIKeyGoesHere&quot;, // &lt;-- This is your public key<br>  authDomain: &quot;://firebaseapp.com&quot;,<br>  projectId: &quot;your-app-id&quot;,<br>  storageBucket: &quot;://appspot.com&quot;,<br>  messagingSenderId: &quot;1234567890&quot;,<br>  appId: &quot;1:12345:web:abcde12345&quot;<br>};</pre><p>Anyone can download your client-side JavaScript bundle and scan it to find the apiKey, projectId, and other Firebase configuration information.</p><p>This is public by design. The problem is not that the key is public; the problem is what the public-facing API allows someone to do with it if you don’t have enough protection around it.</p><blockquote><strong><em>Defense</em></strong></blockquote><p>Now that we knew the fake-account email pattern, we needed a mechanism to stop it.</p><p>Firebase provides a feature called <strong>Blocking Functions</strong>, which allows you to run custom logic before creating an account or before logging in.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/565/0*6Brye1eimejzODXE.png" /></figure><p>We deployed a Cloud Function like the following and configured it through the Blocking Functions section:</p><pre>export const beforecreated = beforeUserCreated((event) =&gt; {<br>  const user = event.data;<br>  <br>  // Example: Block users with unauthorized email domains<br>  const {<br>    email,<br>  } = user;<br>  if (email &amp;&amp;<br>    (<br>      email.includes(&quot;@emalupe.com&quot;) ||<br>      email.includes(&quot;@uberip.com&quot;) ||<br>      email.startsWith(&quot;mk&quot;) ||<br>      email.startsWith(&quot;probe&quot;)<br>    )) {<br>    throw new HttpsError(&quot;invalid-argument&quot;, &quot;Unauthorized email domain.&quot;);<br>  }<br>  return;<br>});</pre><p>This worked.</p><p>But I knew this defense could only stop <strong>this particular attack</strong>.</p><p>Without solving the root problem, the hacker would simply come from another direction. I knew they weren’t going to give up that easily.</p><p>At the same time, we also added HTTP referrer restrictions to the API key, limiting requests to our app domain, <a href="https://app.fina.money/?ref=sybil">app.fina.money</a>.</p><p>Technically, we knew this wouldn’t help much by itself, because a determined attacker could potentially simulate or manipulate the relevant request context. But it was still another layer of protection.</p><p>And then, as expected, the hacker came back.</p><h4>Round 2: The Hacker Abused Our Partners’ Services</h4><p>Fina integrates with partners to provide bank connections for our users. These integrations are done through APIs.</p><p>Like I expected, the hacker started programmatically creating a large number of fake accounts without following the previous email pattern.</p><p>This time, the accounts mostly ended with googlemail.com or gmail.com, and their email handles looked like randomized variations of legitimate-looking Gmail addresses:</p><ul><li>a.bf.xy.7859@gmail.com</li><li>yui.45.nu.t.cr@googlemail.com</li><li>u.2.opl.x.ty@gmail.com</li><li>…</li></ul><p>Why did the hacker choose this method?</p><p>Because Google <a href="https://support.google.com/mail/answer/7436150?hl=en"><strong>treats email addresses with or without dots as identical</strong></a> for personal Gmail accounts.</p><p>This is convenient for regular users, but it is also convenient for someone trying to generate huge numbers of seemingly different email addresses.</p><p>With some email automation, the attacker could even automatically read the verification emails sent by Fina and verify these accounts.</p><p>And needless to say, AI makes this kind of automation much easier.</p><p>It enables hackers to do bad things faster and with less effort, too.</p><p>Through this flaw, the hacker started sending hundreds and then thousands of Link requests through Fina’s API to our partners, Plaid and MoneyKit.</p><p>The result was troublesome. Our partners started temporarily cutting off our access to stop the abusive traffic. That meant our real users could no longer link their banks during the suspension.</p><p><strong>What a loss.</strong></p><blockquote><strong><em>Firefighting</em></strong></blockquote><p>At this point, we had to move quickly.</p><p><strong>1. We deployed a throttling guard.</strong></p><p>We limited the number of API calls that could be made within a small time window.</p><p>This helped reduce the cost and slow down the attack, but it didn’t really solve the problem.</p><p><strong>2. We deleted the fake accounts.</strong></p><p>We needed to remove these accounts so their identities could no longer be used to access Fina’s APIs.</p><p>This was much harder this time because the emails no longer followed an obvious pattern.</p><p>Fortunately, the majority of these accounts were still unverified, so we could identify and delete them programmatically.</p><p>Unfortunately, about 100 of them had already been verified. For those, I had to use my own judgment and delete them manually.</p><p>It was messy. But we bought ourselves something important: <strong>time.</strong></p><h4>Round 3: Fix the Root Problem</h4><p>Before the hacker could launch another massive attack, we had a small window of time to fix the root problem.</p><p>But how?</p><p>As I mentioned earlier, the hacker was taking advantage of Firebase’s public-facing account creation flow.</p><p>To completely solve the problem, we had to give up some of that convenience.</p><p>It was painful, but only for a short time.</p><p>We went directly into the Firebase console and disabled the <strong>“Enable create (sign-up)”</strong> option under User Actions. This completely disabled user account creation through the public Firebase API.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/937/0*u2u4l9TJnUiwPVJq.png" /></figure><p>But we still needed normal users to be able to sign up with email and password.</p><p>So we moved user creation into our own API layer.</p><p>Then, to protect this new endpoint from another variant of the same attack, we added another layer of protection:</p><p><strong>reCAPTCHA.</strong></p><p>After this migration, clients could no longer directly use Firebase’s public API to create a user account. Instead, account creation had to go through our own API, which was protected by reCAPTCHA and our own server-side controls.</p><p>Finally, I felt like we could breathe a little better.</p><p>The hacker actually helped us make the system more secure and more robust.</p><p>But honestly, I still really hope they don’t launch another time-wasting attack like this. 😅</p><h4>Conclusion</h4><p>This is the story of how we fought a <strong>Sybil attack against Fina</strong> over a 10-day window.</p><p>What made it interesting was that it didn’t happen as one single attack. It came in multiple rounds.</p><p>We would block one approach, and the hacker would come back with another variation. We watched them try different ways to bypass our checks, and it really felt like a boxing match on a stage.</p><p><strong>Round 1:</strong> Block the obvious fake email domains.</p><p><strong>Round 2:</strong> Stop the attacker from abusing our partners through newly created accounts.</p><p><strong>Round 3:</strong> Finally fix the root problem and move account creation behind our own API.</p><p>The biggest lesson for me is simple:</p><p><strong>Never leave an exposed cut and trust your luck.</strong></p><p>We could have applied the root fix on the first day instead of letting the battle last for 10 days.</p><p>And today, with AI making automation dramatically easier, I think this lesson is even more important.</p><ul><li>Attackers can move faster.</li><li>They can try more variations.</li><li>They can automate more of the boring work.</li></ul><p>So we need to assume that if an attack surface exists, sooner or later someone will find it.</p><p><strong>Security is not about hoping nobody notices the door is open. It’s about closing the door once you know it is.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e90930787b4f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[People want the result, not a tool.]]></title>
            <link>https://shawncao.medium.com/people-want-the-result-not-a-tool-454f094250e9?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/454f094250e9</guid>
            <category><![CDATA[data-reporting]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[business-outcomes]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[business-intelligence]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Fri, 31 Jul 2026 21:00:23 GMT</pubDate>
            <atom:updated>2026-07-31T21:00:23.041Z</atom:updated>
            <content:encoded><![CDATA[<h4>How Columns Flow evolves from its learning.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rqe7Ral7okuiFG6WSii8xA.png" /><figcaption>Columns Flow answers People for what they want</figcaption></figure><p>When people come to Columns, what do they want to get?</p><p>It is the most important question to me as a founder. Building value that potential customers were seeking is the only way to be relevant.</p><p>It is also the most difficult problem to me as a founder. Who knows what the most common needs among those random strangers are?</p><p>By claiming myself as a data expert, I have been building too many data products within corporations. From the technical perspective, however, I can never measure how relevant they are because we do not need to sell when we work in a company.</p><h4>Learning from Visitors</h4><p>No matter what to take, we have to understand the potential customers. When our landing page attracted some people to sign up, what was in their head? What do you look for?</p><p>Luckily, building a product that mainly produces results from users’ chat has an advantage. We know what users speak to our chatbot. The quality lies in how quality results the chatbot can give back to the user.</p><p>I saw many users simply ask Columns Flow assistant:</p><ul><li>Show top 10 products</li><li>Find highest revenue</li><li>Which listing generates the most</li><li>Compare P1 and P2</li><li>Peak demand</li><li>Average by week</li><li>Which funds need deal origination</li><li>……</li></ul><p>As you can tell, phrases like these imply strongly business outcome-seeking intent.</p><p>It is a big challenge for data systems to understand short phrases, especially for those ambitious ones, but also a great opportunity.</p><p>If it can accurately provide valuable responses, it will be relevant, engaging, and sticky.</p><p>At the same time, I also saw a lot of simple phrases targeting chart, visualization, and dashboard, for example:</p><ul><li>Create a chart</li><li>Build a graph</li><li>Help visualize</li><li>Bubble chart</li><li>Dashboard</li></ul><p>People don’t want to learn how data analysis works or data visualization works; they want to see the insights and visual outcome to display those insights.</p><h4>Unified Data Language</h4><p>In the pre-AI time, business intelligence was clearly framed into multiple technical layers:</p><ol><li>Data Cleaning &amp; Preparation: ETL, Python, etc.</li><li>Data Analysis: SQL</li><li>Data Visualization: BI tools</li></ol><p>It is not only involving multiple tools for each layer, but also involving multiple professionals to work on each layer collaboratively.</p><p>However, this framework setup has been completely shuffled by AI, there are no more clear boundaries between those layers.</p><p>People expect business outcomes directly from the raw data they have at hand.</p><p>By achieving this, people only speak one data language rather than multiple. If you haven’t read my previous article “<a href="https://medium.com/@shawncao/how-to-speak-the-language-of-data-44005783d7d8">How to speak the language of data?</a>”, you should take a look.</p><h4>Columns Flow</h4><p>So, what do users expect to get by using Columns Flow?</p><p>As I said earlier, we built Columns Flow to deliver the outcome users want when they throw the raw data in with a simple phrase. And the outcome is very specifically defined as:</p><blockquote><em>Repeatable data pipeline that analyzes your data accurately to get the most important insights and presents them in customizable stories and dashboards.</em></blockquote><p>To support this mission, we have laid out a lot of foundational work in the past, including:</p><ol><li>A comprehensive data compute engine that can handle all types of data operations, including: a) data transformation b) data join c) data aggregation</li><li>A complete customizable data visualization engine that supports storytelling. a) charts: bar, pie, line, box plot, tree, etc. b) annotations: text, shape, markers, etc. c) configurations: every visual element is customizable.</li><li>A flexible dashboard container to track insights live.</li><li>A loyal data scheduler to keep all schedules running up to the latest changes from the sources.</li></ol><p>Today, I would love to run a quick demo for you to see how Columns Flow gives the outcome you expected with little effort.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F10gG6pxSes4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D10gG6pxSes4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F10gG6pxSes4%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/84e52349cb6a2c321f6dc803054eb94c/href">https://medium.com/media/84e52349cb6a2c321f6dc803054eb94c/href</a></iframe><h4>Your Business Analyst — Automated.</h4><p>Please try it out at <a href="https://columns.ai,/">https://columns.ai,</a> free credits applied.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=454f094250e9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Speak the Language of Data?]]></title>
            <link>https://shawncao.medium.com/how-to-speak-the-language-of-data-44005783d7d8?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/44005783d7d8</guid>
            <category><![CDATA[towards-data-science]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[data-analyst]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[career-development]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Mon, 06 Jul 2026 03:03:27 GMT</pubDate>
            <atom:updated>2026-07-06T03:03:27.791Z</atom:updated>
            <content:encoded><![CDATA[<h4>Anyone can become a data analyst once they learn to speak the language</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*p85eSlNRJijuB5bQbdd5nw.jpeg" /><figcaption>data language evolution (AI generated)</figcaption></figure><p>I’ve been maintaining a 60-day streak on Duolingo to learn French.</p><p>It’s a fun practice, although it’s a significant challenge to pronounce those accent notes correctly. I believe French is generally a simpler language than English; you usually use shorter sentences to convey the same meaning.</p><p>Data has its own language too.</p><p>Data is the lifeblood of every modern business. Every decision, insight, and opportunity begins with understanding what the data is trying to say.</p><p>But unlike spoken languages, data doesn’t require everyone to learn the same vocabulary or syntax. Instead, you can interpret and express it in a way that matches how you think, making data analysis more intuitive, accessible, and uniquely your own.</p><h4>From Python, SQL to Natural Language</h4><p>Python, a programming language, has gained popularity as the preferred language for data processing within the data science community due to its portability. SQL, on the other hand, serves as the de facto interface for rational databases.</p><p>In the past, becoming a data analyst required proficiency in both Python and SQL. Even today, data analyst job descriptions often mention these requirements.</p><p>However, the advent of AI has revolutionized this landscape. Anyone with the ability to communicate effectively in the data language can excel as a data analyst.</p><p>While programming skills are not strictly necessary, a solid understanding of data language is crucial. Imagine joining a new friend circle who works in a completely different domain. After a brief introduction of common keywords, you can easily engage in conversations with them.</p><h4>Use Spreadsheet for Reference</h4><p>Nearly every office worker uses spreadsheets, either Microsoft Excel or Google Sheets.</p><p>Even without the complex formulas, pivots, and lookups, the basic structure of a spreadsheet consists of three main components:</p><ul><li>Rows</li><li>Columns</li><li>Data types</li></ul><p><strong>Rows</strong> are records that constitute a table. You can also consider a row as an object that represents a real-life entity, such as a person, a cup, or an invoice.</p><p><strong>Columns</strong> are the fixed properties that describe each object (row). They form the schema that every row adheres to, ensuring uniformity in the data for processing.</p><p>A schema is of utmost importance for data analysis as it enables the application of all rules. Without a <strong>schema</strong>, any logic that is not compatible with the data language may fail to execute.</p><p>Data types describe the value format of each property. For simplicity, you only need to be concerned with whether it is a <strong>number</strong> or <strong>text</strong> for now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/506/0*p_7WBW65xQ1YeO_h.png" /><figcaption>Rows of Orders (OrderId-text, CustomerId-text, Product-text, Amount-number)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/407/0*9jS51AaBf4_Ms7BL.png" /><figcaption>Rows of Customers (ID-text, Name-text, Channel-text)</figcaption></figure><h4>Data Language Patterns</h4><p>Data language offers a wide range of tasks that can be accomplished. Let’s explore each of these tasks and learn how to communicate effectively with data to achieve them.</p><p>These scenarios are referred to as <strong>patterns</strong> because they serve as templates that can be applied to your own data.</p><p>To facilitate understanding, we’ll use the above tables in the following descriptions.</p><blockquote><strong>Pattern-1: Filter Rows</strong></blockquote><p><strong>Filter</strong> is to describe a condition to get objects you care about and skip those uninterested records.</p><p>Examples:</p><ul><li>“<em>Orders of Milk</em>”</li><li>“<em>I want orders of milk products.</em>”</li><li>“<em>All orders that are not for books.</em>”</li><li>“<em>All orders with a sales amount exceeding 20.</em>”</li></ul><p>AI can produce code logic to filter the targeted records for further processing, if translating above statements into SQL, they will look like:</p><ul><li>“<em>where product=’Milk’</em>”</li><li><em>(same as #1)</em></li><li>“<em>where product &lt;&gt; ‘Book‘</em>“</li><li>“<em>where amount &gt; 20</em>”</li></ul><p>As you can see, <strong>filter</strong> is achieved by keyword “where” in SQL.</p><blockquote><strong>Pattern-2: Transform Object</strong></blockquote><p>Sometimes, we want to clean a data field or transform it into a desired shape or format, either for improved readability or more efficient processing.</p><p><strong>Transformation</strong> creates a new property in your original record.</p><p>To transform an existing property into a new one, you need a function of logic. For both spreadsheets and SQL, “formula” is the tool you’ll need.</p><p>However, with the increasing capabilities of AI in coding, natural language offers a significant advantage. It allows us to achieve the same transformation without having to learn, memorize, and assemble complex formulas.</p><p>Taking one simple example:</p><ul><li>“<em>Get customer first name</em>”</li></ul><p>This is equivalent to composite multiple formula together in Spreadsheets like</p><ul><li>“=INDEX(SPLIT(A2, “ “), 1)” or “=IFERROR(LEFT(A2, FIND(“ “, A2) — 1), A2)”.</li></ul><p>This operation creates a new column called “First Name”.</p><p>You can also acquire a new property by combining multiple existing properties, such as “concatenating the last name and channel as a label”. Logic like this is simple for AI coding but too complex for spreadsheet formulas.</p><blockquote><strong>Pattern-3: Aggregate Records</strong></blockquote><p>Aggregation processes a large collection of records to provide a summarized view.</p><p>This is powerful because it compresses vast amounts of information into manageable pieces that humans can comprehend and analyze.</p><p>To combine multiple data sets into a single piece of information, you need to understand the “how-to,” which leads to the crucial concept of “aggregation methods” or “computation logic.”</p><p>Typically, text data (a property or column with a text data type, as discussed in the schema section) is not particularly interesting for aggregation. The most common approach is to concatenate text data to form a long paragraph, although this is still uncommon.</p><p>Most computation logic involves operations on numerical data. When an <strong>aggregation method</strong> is applied to a numeric property or column, you essentially have a list of numbers that can be aggregated, such as:</p><ol><li>Total value (sum)</li><li>Average value</li><li>Mean value</li><li>Minimum value</li><li>Maximum value</li><li>A specific percentile value (e.g., P25, P50, P75, P90)</li></ol><p>However, counting objects or counting unique property values is also quite common.</p><p>When discussing aggregation, we cannot overlook “<strong>breakdowns</strong>.” This involves creating a segmented view of the data rather than a single total view.</p><p>For example, in the previous Orders table, “<em>total sales by product</em>” or “<em>average amount by customer</em>” are equally valuable insights for an analyst to explore.</p><p>In summary, aggregation can be described as:</p><ul><li>“<em>Compute an aggregated value of a property group based on another property.</em>”</li></ul><p>Expressing this in standard SQL, it would look like:</p><ul><li>“Compute(property1) from table [group by property2].”</li></ul><p>Let’s practice this using a few examples by speaking the data language:</p><ol><li>“<em>Give me total sales by product.</em>”</li><li>“<em>Tell me the average amount spent by each customer.</em>”</li></ol><blockquote><strong>Pattern-4: Join Multiple Datasets</strong></blockquote><p>When a single dataset (or table) is insufficient to achieve the desired outcome, we must combine multiple datasets. This operation is referred to as “join” or “union.”</p><p>If the multiple datasets contain the same objects but reside in different locations, we can simply merge them. This is a straightforward “union” operation.</p><p>However, most of the time, they store different objects. We have partial information from one dataset and partial information from another. By combining them, we create a comprehensive schema with more available columns.</p><p>This pattern is generally not feasible in spreadsheets, although their lookup function may provide partial assistance.</p><p>For instance, if we want to determine the “<em>total amount spent from each channel</em>” based on previous tables, where the amount is from the orders table and the channel is from the customers table, we need a joined dataset to complete this analysis.</p><p>To join multiple datasets, we must have one or more pairs of join keys. A pair of join keys consists of one column from one table and one column from another. The data engine can utilize these relationships to identify relevant objects and concatenate them to form a larger object.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/377/0*s1_cnlDQS6ruoYsD.png" /><figcaption>Join Orders and Customers</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/405/0*iAWaUzFUmbFtCl_6.png" /><figcaption>Joined dataset have more columns</figcaption></figure><p>In summary, join operations can be described in this pattern:</p><ul><li>“<em>join table1 and table2 when key1 of table1 equals key2 of table2</em>”.</li></ul><p>Translating this pattern into SQL, it will look like:</p><ul><li>“<em>select * from table1 join table2 on table1.key1=table2.key2</em>”.</li></ul><p>In fact, you may not need to use this pattern in natural language explicitly, because modern AI is smart enough that it can infer the whole join logic from your data language.</p><p>For instance, the example we gave earlier, if you speak this sentence “total amount spent from each channel” to <a href="https://columns.ai/">Columns AI</a>, it will figure out all the necessary actions to get the desired outcome for you.</p><blockquote><strong>Pattern-5: Visualization</strong></blockquote><p>Data visualization, often overlooked as a part of data language, plays a crucial role in transforming mundane data into vivid images. This visual representation significantly aids the audience in comprehending the insights you intend to convey.</p><p>By incorporating customization and assistance to articulate your insights and predictions, you position yourself as a data storyteller, showcasing your influence within the domain.</p><p>Since visualization doesn’t alter the data itself, in the language of data, we merely need to indicate the desired outcome. For instance:</p><ul><li>“<strong>Display</strong> the total amount by product in a <strong>pie</strong> <strong>chart</strong>.”</li><li>“I would like to see a <strong>timeline</strong> of total sales <strong>month-by-month</strong> for the past six months.”</li><li>“<strong>Show</strong> the number of sales by customer in a <strong>bar</strong> <strong>chart</strong>.”</li></ul><p>These <strong>bold</strong> keywords serve as cues to the AI engine, guiding it in generating the final visualization based on your data.</p><h4>Practice Data Language</h4><p>Similar to how I diligently practice French on Duolingo every day, we must practice speaking data language using the data we possess.</p><p>As long as you have adhered to the five patterns mentioned above, you should have mastered data analysis like a professional data analyst. You don’t need to be an Excel expert or a Python or SQL wizard.</p><p>Let’s use the provided example data to practice speaking the data language. You can find the “Orders” and “Customers” data from this <a href="https://docs.google.com/spreadsheets/d/1wHgsftG3GAHzzyR-rY6OSlHzyxTrtMBqWiERrbcE52w">spreadsheet link</a>.</p><p>Suppose we want to perform a sales analysis of customer distribution based on the data.</p><p>The data language is almost the same, but let’s ensure we’ve used the correct keywords and patterns to guarantee that the AI engine follows the instructions precisely.</p><p>For instance, we speak to AI: ”<strong>display</strong> the <strong>total</strong> sales by customer’s <strong>first name</strong> in a <strong>bar</strong> chart.”</p><p>Here’s how the AI interprets this:</p><ol><li>“<strong>total sales</strong>” → summing up the amount values.</li><li>“<strong>first name</strong>” → it can be transformed from “name.” A transformation will be applied.</li><li>“<strong>by</strong>” → the summing up result needs to broken down by first name.</li><li>“<strong>sales &lt;&gt; customer</strong>” → sales data is from the Orders table’s amount field, while customer data is from the Customers table. Therefore, a Join operation is required to combine these two datasets.</li><li>“<strong>show, bar</strong>” → the result should be visualized in a bar chart.</li></ol><p>AI will then determine the correct execution order, ensuring that each step has all the necessary data when it executes.</p><p>This is what Columns Flow produces upon hearing this sentence: <strong>“display</strong> the <strong>total</strong> sales by customer’s <strong>first name</strong> in a <strong>bar</strong> chart.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kaw6wVpj2au2boiC.png" /><figcaption>AI knows how to translate your intention into steps in the correct order</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WOKNgueQUqcaOrjX.png" /><figcaption>One simple visualization using bar chart</figcaption></figure><h4>Conclusion: Speak Data Language</h4><p>In this article, we’ve demonstrated the historical opportunity for everyone to become a great data analyst in this era.</p><p>We discussed how professionals used programming languages like Python or SQL as their primary data languages. However, the data language has evolved to become the natural language we speak daily.</p><p>To become a data analyst, we need to understand the fundamental scenarios involved and use the correct keywords to make the data language understandable to AI engines.</p><p>Here’s a quick recap:</p><ul><li><strong>Dataset</strong>: rows, columns, and schema.</li><li><strong>Filtering</strong> and <strong>Transformation</strong>: These processes involve filtering data and transforming it into a usable format.</li><li><strong>Aggregation</strong>: This involves summarizing data into a single value, such as the total or average. Specify “compute methods” and optional “breakdown” if needed.</li><li><strong>Join Datasets</strong>: This involves combining data from multiple sources.</li><li><strong>Visualization</strong>: This involves creating visual representations of data to make it easier to understand.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dIsc1cyjJLrkq145.png" /><figcaption>Put all pieces together</figcaption></figure><p>Unlike learning a new language like French, if you’re willing to spend just a few hours going through this short list and practice them, you can become a professional data analyst!</p><p>It’s a great time to be a data analyst, and I believe in your ability to succeed. Thanks for reading!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=44005783d7d8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The 15-Minute Exercise is Now 15 Seconds]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://shawncao.medium.com/the-15-minute-exercise-is-now-15-seconds-ff9bb80f2f7a?source=rss-b7aa29e347a4------2"><img src="https://cdn-images-1.medium.com/max/640/1*94DayKqHsicWS_0TDYwE1A.gif" width="640"></a></p><p class="medium-feed-snippet">How user feedback drove us to slash onboarding time by 98%</p><p class="medium-feed-link"><a href="https://shawncao.medium.com/the-15-minute-exercise-is-now-15-seconds-ff9bb80f2f7a?source=rss-b7aa29e347a4------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://shawncao.medium.com/the-15-minute-exercise-is-now-15-seconds-ff9bb80f2f7a?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/ff9bb80f2f7a</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[ux]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Wed, 03 Jun 2026 17:45:11 GMT</pubDate>
            <atom:updated>2026-06-03T17:45:11.007Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Data Analytics Basics for Beginners]]></title>
            <link>https://shawncao.medium.com/data-analytics-basics-for-beginners-dfd136a4da46?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/dfd136a4da46</guid>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[data-processing]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[sql]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Wed, 20 May 2026 02:21:11 GMT</pubDate>
            <atom:updated>2026-05-20T02:21:11.295Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y4vua-M9QBMQA1j4dORlCA.png" /><figcaption>15-minute outcome</figcaption></figure><p><strong>SQL</strong> is a popular starting point for beginners in data analytics. However, like many programming languages, writing basic queries can now be easily handled by AI.</p><p>A more valuable approach is mastering the fundamental concepts, and even better, learning them visually in just 15 minutes.</p><p>Sign up for a free account on<strong> </strong><a href="https://columns.ai/?ref=medium"><strong>Columns AI</strong></a> and complete the flow starter exercise to master all the concepts: automation, transformation, join, aggregation, visualization, monitoring, and reporting.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/316/1*m7s6WUv09HaF_SyM1F6jog.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dfd136a4da46" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The End of Manual Data Entry Is Here]]></title>
            <link>https://shawncao.medium.com/the-end-of-manual-data-entry-is-here-c1e47a2e0595?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1e47a2e0595</guid>
            <category><![CDATA[data]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <category><![CDATA[data-entry]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Thu, 23 Apr 2026 19:54:20 GMT</pubDate>
            <atom:updated>2026-04-23T19:54:20.554Z</atom:updated>
            <content:encoded><![CDATA[<h4>How Columns Drive uses AI to turn messy receipts, PDFs, and disjointed spreadsheets into clean datasets in seconds.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cRDcg3jXrMdJFDlDcU_nRg.png" /><figcaption>Generated by ChatGPT</figcaption></figure><p>“My uncle runs a small wholesale business, and every week he’d sit down and manually type 30–40 line items from supplier invoices into a spreadsheet. Every. Single. Week.”</p><p>When John told me this story, I said to myself: “Damn, no wonder there are still so many data entry job posts on Fiverr.”</p><p>It seems the real world is always lagging a bit behind frontier technology.</p><p>Today, machines have sharper eyes than humans, and they also make fewer mistakes.</p><p>Just the other day, Mora posted her latest task on Reddit, seeking help on how to complete it faster:</p><blockquote>“My company has archived Excel files of monthly summaries since 2010–12 Excel files a year, 15 years in total. I was asked to compile them all into one single dataset for further analysis and better organization. It is going to drain all my weekends.”</blockquote><p>Simply merging those files isn’t the hard part. The real challenge is that not all the files follow the same format, even though they all contain the required fields.</p><p><strong>Extracting Structured Data From Real Life Is A True Problem</strong></p><p>Within <a href="https://columns.ai/blog/the-new-columns-is-here">the new Columns product line</a>, <strong>Columns Flow</strong> serves as the core engine that helps us analyze our data in a continuous flow.</p><p>While it connects many data sources for live syncing, it previously lacked a seamless way to capture data directly from real life.</p><p>That is exactly why we added <strong>Columns Drive</strong> to the family!</p><p>At first glance, Columns Drive might not seem much different from other cloud drives you may have used, like Google Drive, OneDrive, or Dropbox.</p><p>Those are simply cloud-based copies of your local disk. Columns Drive is also a cloud drive, but its primary focus is on <strong>data extraction</strong>:</p><ul><li><strong>Intelligent Extraction:</strong> Uploaded files are automatically processed to extract structured data.</li><li><strong>Unified Datasets:</strong> Each folder acts as a unified dataset following the exact same schema.</li><li><strong>Context-Aware AI:</strong> The system learns, meaning previously extracted data is used to intelligently improve future extractions.</li><li><strong>Seamless Uploads:</strong> Whether it’s images, Excel files, or PDFs, simply drag and drop them into a folder!</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/986/1*V82Ha6aty5OwRb-0D3Fj3Q.png" /><figcaption>Columns Drive</figcaption></figure><h4>How Does Columns Drive Help John’s Uncle and Mora?</h4><p>With Columns Drive, John’s uncle now spends just 15 minutes each week scanning all his receipts into a folder named by the date. That’s it!</p><p>If he wants to open the data in Excel, he just clicks the data icon, or he can drop it directly into a <a href="https://columns.ai/product/flow">Columns Flow</a> for further processing and analysis.</p><p>Similarly, Mora is completely freed from all that manual labor. She simply opens Columns Drive, creates a folder called “Snap,” selects all the Excel files that need merging, and drags them into the folder, all in less than a minute!</p><p>She leaves for her party, and in the background, Columns Drive automatically extracts all the data, merging it into a single set that follows the same schema. When she comes back from the party, she just downloads the dataset and hits send to share it with her boss.</p><p>That’s how <strong>modern OCR and AI technology</strong> power Columns Drive, ensuring countless hours of data entry work are shortened from days to minutes.</p><p>If these stories inspire you, make <a href="https://columns.ai/product/drive">Columns Drive</a> work for you too!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1e47a2e0595" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Unlearn Excel]]></title>
            <link>https://shawncao.medium.com/unlearn-excel-40918d75ea73?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/40918d75ea73</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[future-of-work]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Thu, 02 Apr 2026 21:02:15 GMT</pubDate>
            <atom:updated>2026-04-02T21:02:15.181Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why the World’s Most Popular Software is Killing Your Productivity</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pPrK67LGKWrSUw3V" /><figcaption>An Excel Template Picture</figcaption></figure><h3>The Excel Paradox</h3><p>Excel is the undeniable backbone of many businesses in the world.</p><p>For decades, millions of people across the globe have relied on its familiar grid to manage everything from grocery lists to multi-million-dollar corporate budgets.</p><p>It is <strong>accessible</strong>, <strong>ubiquitous</strong>, and fundamentally brilliant at acting as a blank canvas for lightweight information.</p><p>But let’s face a harsh reality: its widespread adoption is largely driven by habit, not superiority.</p><p>Its greatest strength of being a totally open and unrestricted grid is precisely its fatal flaw. Infinite flexibility comes along with infinite chaos.</p><h3>The Illusion of Productivity</h3><p>Have you ever written or seen a 300-character nested IF statement or an intricate VLOOKUP? You think it is a badge of honor, but it is not. It is a colossal waste of human potential.</p><p>Excel files inevitably devolve into fragile, messy nets of broken links and outdated macros. The fact that millions of professionals spend hours repeating the exact same manual data manipulation every single week is a tragedy in the modern time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tgniMYARdtCsybaxVS4cBg.png" /><figcaption>Glance of Excel community on Reddit</figcaption></figure><p><strong>We must unlearn Excel.</strong></p><p>Leveraging AI is no longer just a “nice-to-have” marginal improvement, it is a complete paradigm shift in productivity that makes spreadsheet manipulation smell.</p><h3>A Real Life Example: The Tennis Shoes</h3><p>As a senior member of a local tennis club, I volunteered to run some simple analysis and share some insights to club members on all available tennis shoes from affiliated tennis shops in partnership.</p><p>There is no structured data available and we only have dozens of screenshot images captured from those product list across their websites.</p><p>We need a price, brand name and feature set if possible, to make a report of insights to share with the passionate group of tennis enthusiasts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RZHZ_sgkIdpE9hIlNRXrXw.png" /><figcaption>Product Listing Samples</figcaption></figure><h3>The Ox Way (Excel)</h3><p>If I cling to the old ways, my weekend is ruined. I will spend hours or perhaps days, depending on the volume: manually transcribing text from those images into individual Excel cells.</p><p>I will battle typos, misaligned columns, and spend an agonizing amount of time cleaning the data. And the worst part? The system is entirely static.</p><p>The moment a new batch of screenshots arrives next time, I must sit down and repeat this grueling, soul-crushing manual process all over again.</p><h3>The Falcon Way (AI)</h3><p>Now, let’s look at how this workflow is meant to be executed using modern, AI-driven infrastructure. The difference is night and day:</p><p><strong>(Step-1) Intelligent Extraction:</strong> You simply drag and drop the screenshots into a smart cloud drive. The AI automatically reads the images, extracts the relevant details (brand, price, model), and maps the information into a perfectly consistent, structured data schema. Zero manual entry required.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6HukQ6YJd4rg1z3-i_LsOQ.png" /><figcaption>Smart Cloud Drive</figcaption></figure><blockquote>It is effortless, just drop those images to the same folder, and all your data will be extracted and merged as a dataset per folder.</blockquote><p><strong>(Step-2) Drive to Flow:</strong> That cloud drive instantly connects to an automated data flow pipeline.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/822/1*KQw4AQWemg-vJOZ6j0820Q.png" /><figcaption>Use Drive Folder in a Flow</figcaption></figure><blockquote>Just one-click to connect a drive folder to a flow</blockquote><p><strong>(Step-3) Natural Language Transformation:</strong> Instead of writing convoluted formulas to pivot or group your data, you simply describe what you want to do in plain English. The AI translates your intent into instant data transformations.</p><blockquote>The core argument of “Unlearn Excel” comes to here: instead of learning and maintain those complex formulas, we can simply tell it what we want, such as:</blockquote><p><em>“extract color from name in square bracket, keep the first one color only if there are multiple separated by “/”, and then get average price by color, sort result by color alphabetically.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/825/1*WIV-IXtuobQ7ug5fL6ioHQ.png" /><figcaption>Just Tell It</figcaption></figure><p>I already got the data I need for my analysis of average price by color. The rest parts are just report generation and put it in automation.</p><p><strong>(Step-4) Effortless Visualization:</strong> You receive a stunning, presentation-ready visual outcome without spending a single second adjusting chart axes or tweaking color hex codes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/993/1*icLRgihJR4poDo7xtEM8YQ.png" /><figcaption>Highly Customizable Graph</figcaption></figure><p>It is just one-click away to save a beautiful and customizable graph for any transformed data snapshot.</p><p><strong>(Step5) True Automation:</strong> You turn on autopilot. The next time you drop a new shoe screenshot into the folder, the final graph updates itself instantaneously.</p><p>You can even set up automated alerts to notify your tennis group the second a specific shoe drops below a target price point.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/477/1*UUTkCSVlDwmOE2RLuJAi4Q.png" /><figcaption>Turn On Automation</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/738/1*sFkvXvk3SNHOnfaLiCRECg.png" /><figcaption>Set Up Alerts</figcaption></figure><h3>Conclusion: Why We Must Unlearn</h3><p>We must unlearn Excel because it was never designed to be the automated brain of your operations.</p><p>Microsoft Excel and Google Sheets are fantastic lightweight databases, they are excellent places to simply <em>store</em> and view simple data. But to <em>process</em> that data, to extract meaning, and to build scalable workflows, you must leave the spreadsheet behind. It is time to let intelligent, modern systems handle the heavy lifting seamlessly and automatically, so you can focus on the insights, not the mechanics.</p><p><strong>Ready to step into the future of data?</strong> Check out <a href="https://columns.ai/?ref=medium"><strong>Columns AI</strong></a>. It features every single intelligent component described in this article, empowering you to build effortless, automated data workflows from unstructured files to stunning visual insights.</p><p>Unlearn the old way, and start building the smart way.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=40918d75ea73" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Chat to Simulate My Financial Future (And Optimize My Present)]]></title>
            <link>https://wire.insiderfinance.io/chat-to-simulate-my-financial-future-and-optimize-my-present-2764a7ba1319?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/2764a7ba1319</guid>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[financial-planning]]></category>
            <category><![CDATA[personal-finance]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Wed, 11 Mar 2026 13:15:20 GMT</pubDate>
            <atom:updated>2026-03-11T13:15:20.512Z</atom:updated>
            <content:encoded><![CDATA[<h4>From credit card recommendations to Monte Carlo simulations, here is why Fina Chat is a game changer for your personal finances.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3OWhBCfoPmeFMe1Bnu7qHw.png" /><figcaption>Amazing Fina Chat</figcaption></figure><p>Fina Money has always been the most <strong>flexible</strong> finance tracker in the world.</p><p>Now, however, I would rather call it the most <strong>intelligent</strong> finance platform in the world.</p><p>AI has evolved so tremendously that it has vastly improved how we obtain and organize information. It is especially exciting when AI interacts with our personal data to create real upside!</p><p>In this post, I will walk you through a few examples to show you how AI has empowered Fina to become an incredibly useful tool for managing our finances.</p><h3>Understanding the Past</h3><p>Whether you are a paid-tier user with automated data or a free-tier user tracking manually, Fina keeps your financial activities organized, categorized, and tagged.</p><p>Because of this, a simple question can yield an accurate result displayed in your preferred visual format.</p><p>For example, if I ask, <em>“How much did I spend eating out in the past 30 days?”</em>, Fina will show me the answer:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*7mKcc_4UDeWgkSCK5bX7AQ.png" /><figcaption>Simple &amp; Quick Answer</figcaption></figure><p>Or, I can ask for <em>“a timeline for my monthly gas costs over the last 90 days”</em>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*_k55VYxP3qQA8Y9gWXrUvg.png" /><figcaption>Timeline of any topic</figcaption></figure><p>You can ask almost anything about your data, including questions that require <strong>complex calculations or formulas</strong>.</p><p>For instance, I can prompt Fina to: <em>“Within the current profile, build a grid block to show my investment holdings with four metrics: cost, current value, gain or loss percentage, and % of total wealth. Sort the list by current value from high to low.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*KLL-iXSw8ipvoarqakEAFw.png" /><figcaption>Portfolio overview</figcaption></figure><p>Even better, you can <strong>pin these blocks</strong> to your dashboard for real-time updates and future reference.</p><p>As you can see, Fina’s AI doesn’t just provide standard text answers; it builds <strong>savable computing models</strong> that update dynamically as your data changes.</p><p>Safety is also paramount. So, can it help detect suspicious transactions?</p><p>When I ask, <em>“Can you review my transactions for fraud or suspicious activity?”</em>, it successfully flags entries that look unusual (even if they turn out to be fine).</p><p>No screenshot here, :) I’ll leave testing that feature out as <a href="https://www.fina.money/?ref=medium">your homework</a>!</p><h3>Taking Action in the Present</h3><p>Understanding the past gives us insights into our finances, but what can we do <em>right now</em>?</p><p>We can get <strong>recommendations</strong> and make adjustments to drive our finances into better shape.</p><p>Take credit card usage as an example: are we maximizing our benefits, or are there better choices we could be making?</p><p>To find out, I asked, <em>“Am I getting all the value from my credit cards?”</em> Fina clearly showed me how much of my spending utilized my highest reward tiers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*ydBsUsHWpnxcEMjMAQ1ZGg.png" /><figcaption>Credit card reward tier utilization</figcaption></figure><p>Looking at this, I realized I should probably use my Citi card more often for gas and EV charging.</p><p>This led me to wonder: could I apply for credit cards that offer better returns based on my past spending patterns?</p><p>So I decided to ask: <em>“Recommend the best credit cards for me based on my expenses last year.”</em></p><p>Wow, it gave me some fantastic choices!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*Ttx31kyMDMeCtkePdAIg4Q.png" /><figcaption>Personalized credit card recommendations</figcaption></figure><p>If you ever need personalized suggestions for your own situation, all you have to do is ask!</p><h3>Foreseeing the Future</h3><p>Having a crystal ball to look into the future would be amazing.</p><p>While that doesn’t exist in reality, we can still paint an estimated picture of the future based on past experiences. This is where projecting your finances becomes really interesting.</p><p>One of the most fun things to forecast is how much wealth you will have in ten years. Monte Carlo simulations are widely used for this exact purpose, so let’s run an example to see how it feels.</p><p>I asked: <em>“Based on my current stock holdings, can you run a Monte Carlo simulation to project how much total wealth I will have in 10 years? Show me the annual numbers in all variances, year by year.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*JrdUq46fVO-OgLFv15Nsmw.png" /><figcaption>Monte carlo simulation</figcaption></figure><p>The initial aggressive variance seemed a bit too good to be true, probably a return even a genius investor couldn’t guarantee!</p><p>So, I decided to ask it to use more realistic bounds for the simulation: <em>“How about using S&amp;P 500 data from the past 5 years to set the bounds for all variances and run the simulation again?”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*rac4QmFyP0SNqStd08Uwxg.png" /><figcaption>Use real data for bounds</figcaption></figure><p>I didn’t just want a simple projection from a starting balance; I also wanted to include future life events in the simulation.</p><p>And yes, Fina handled it perfectly. I prompted: <em>“I plan to withdraw $50K annually from Year 3 to Year 6 for education costs. Can you factor this in and run the projection again?”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*39Mo3siO14E5C29B-JhDZg.png" /><figcaption>Counting in future life events</figcaption></figure><p>As expected, this life event slowed down the growth curve, but not by too much. It still looks safe and on track to me.</p><h3>Conclusion</h3><p>I am incredibly excited about this 360-degree method of looking at personal finances.</p><p>It simply helps us build a well-rounded financial picture by:</p><ul><li><strong>Understanding the past.</strong></li><li><strong>Taking action in the present.</strong></li><li><strong>Foreseeing the possible future.</strong></li></ul><p>If you are as excited about these capabilities as I am, try out <a href="https://www.fina.money/?ref=medium">Fina Chat</a> now!</p><h4>A Message from InsiderFinance</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/301/0*QCVCC4NQvshiTPdI.png" /></figure><p>Thanks for being a part of our community! Before you go:</p><ul><li>👏 Clap for the story and follow the author 👉</li><li>📰 View more content in the <a href="https://wire.insiderfinance.io/">InsiderFinance Wire</a></li><li>📚 Take our <a href="https://learn.insiderfinance.io/p/mastering-the-flow">FREE Masterclass</a></li><li><strong>📈 Discover </strong><a href="https://insiderfinance.io/?utm_source=wire&amp;utm_medium=message"><strong>Powerful Trading Tools</strong></a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2764a7ba1319" width="1" height="1" alt=""><hr><p><a href="https://wire.insiderfinance.io/chat-to-simulate-my-financial-future-and-optimize-my-present-2764a7ba1319">Chat to Simulate My Financial Future (And Optimize My Present)</a> was originally published in <a href="https://wire.insiderfinance.io/">InsiderFinance Wire</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The 10x Data Worker: Automating the Analytics Boredom with AI]]></title>
            <link>https://shawncao.medium.com/the-10x-data-worker-automating-the-analytics-boredom-with-ai-14452a7aa499?source=rss-b7aa29e347a4------2</link>
            <guid isPermaLink="false">https://medium.com/p/14452a7aa499</guid>
            <category><![CDATA[no-code]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Shawn Cao]]></dc:creator>
            <pubDate>Mon, 02 Mar 2026 22:46:28 GMT</pubDate>
            <atom:updated>2026-03-02T22:46:28.722Z</atom:updated>
            <content:encoded><![CDATA[<h4>Apply Advanced AI In Data Wrangling</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nRunPVTTc6YtS1NYwcxkfw.png" /><figcaption>Data Flow Example</figcaption></figure><p><strong>50% of all current AI applications are focused on software engineering.</strong> The reason is simple: AI is exceptionally good at writing code.</p><p>But while engineers are enjoying massive productivity boosts, data workers (researchers, marketers, and business analysts) are still stuck in the trenches.</p><p>I know how many hours you’ve spent wrestling with “reference errors” in a spreadsheet or copy-pasting random Python snippets just to reshape your data. This <strong>“data tax”</strong> consumes the majority of your day, stealing time that should be spent on the high-level strategy and insights that actually move the needle.</p><h4>The Daily Grind</h4><p>If you work with data, this “daily grind” is not strange to you:</p><ul><li><strong>Cleaning:</strong> Fixing typos, standardizing dates, and hunting down duplicates.</li><li><strong>Transformation:</strong> Wrestling with VLOOKUPs, pivoting tables, and merging massive CSVs.</li><li><strong>Analysis:</strong> Struggling to find a meaningful signal within the noise.</li><li><strong>Visualization:</strong> Fighting with rigid chart settings just to make the data presentable.</li><li><strong>Syncing:</strong> The most frustrating part, doing it all over again next week when the data refreshes.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lN-lkwJ5oaN4AAbVk5sRMA.png" /><figcaption>Illustration Generated By Gemini</figcaption></figure><p>This cycle isn’t just tedious, it’s expensive. It caps your productivity and anchors you to manual labor when you should be focused on high-value strategic work.</p><h4>How AI Changes the Equation</h4><p>If AI can write complex software code, it can certainly write clean and reliable data transformation logic for data workers as well.</p><p>In the rest of this article, imagining we are building an ideal data flow “<strong>Platform-X</strong>” that applies the AI “coding power” to the data wrangling process, let’s paint a picture of it.</p><p>Imagine it acts as a translator: you describe your business needs in plain language, and the AI plans the process and writes the complex code for each analytics phase for you.</p><p>The best part? It should automatically run tests to validate both input and output data, and you peek the result and approve it to persist as a repeatable workflow.</p><p>With a workflow tool like this, you can set it up once, and schedule it to run forever to keep in sync with your data changes, always fresh results!</p><p>Any commonly used data sources should be supported: from CSVs and spreadsheets to Airtable, Notion, SQL databases, and even APIs to bridge systems together.</p><p>This picture gives us the power to democratize data engineering. You don’t need to be a Python wizard to build a robust data pipeline, or a spreadsheet nerd, you just need to know what you want the data to become.</p><p>The best part? It validates both the input and output data, and only executes upon your approval. Once you set it up, it runs automatically to keep your results fresh and in sync with all your connected data sources — from CSVs and spreadsheets to Airtable, Notion, SQL databases, and even APIs.</p><p>This democratizes data engineering. You don’t need to be a Python wizard to build a robust data pipeline; you just need to know what you want the data to tell you.</p><blockquote>Let AI do the heavy lifting, we provide the vision.</blockquote><h4>Use Cases: Your New 10x Workflow</h4><p>Given the painted picture above, I can imagine there are at least 3 main use cases we can get complete freedom from:</p><ol><li><strong>Text-to-Data Pipeline</strong>: build a reliable data pipeline in plain language.</li><li><strong>Data-to-Insights Automation</strong>: put a pipeline on schedule to keep insights updated.</li><li><strong>Client-to-Data Validation</strong>: publish a pipeline as a client face component to accept customer data and guard its quality.</li></ol><p><strong>Case #1: Text-to-Data Pipeline</strong></p><p>You have messy data living in various places, such as an Airtable base for project tracking, a Notion database for roadmap items, and a massive SQL database of transaction logs.</p><p>On Platform-X, now simply just connect it as an input node. Then drag and drop a <strong>transformation </strong>to yield an output.</p><p>Instead of manually manipulating spreadsheet cells or trying to write correct python code, you simply type the intenet like: <em>“sales data by date, filter for Q4, and calculate ROI per channel.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/466/1*b3rMp7eWTjZh6dIc25_BGg.png" /><figcaption>Speak Business Logic in Plain Language</figcaption></figure><p><strong>Platform-X</strong> instantly generates the transformation plan for you: cleaning messy values, filling empty holes, aggregating values, pivoting values to fields, sorting, and many more.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/459/1*NRto5g8S8SyiisIqSKHBig.png" /><figcaption>Data Transformation Plan</figcaption></figure><p>All done in a visual flow diagram!</p><p><strong>Case #2: The “Set and Forget” Schedule</strong></p><p>You built a perfect data flow last week, today, your boss asks for the “updated numbers.” Will you go through the whole process you have done last week?</p><p>No.</p><p>Once you build a Flow, it is a living pipeline. Simply set a schedule, such as “Every Monday at 5:00 AM”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/477/1*_rJWwFl01uk-L-1yvNVn_g.png" /><figcaption>Flow Schedule</figcaption></figure><p><strong>Platform-X</strong> wakes up, pulls the fresh data from your sources, applies all the rules through the data flow, updates the visualizations, and emails it to you. If needed, you just forward it to your boss.</p><p><strong>Case #3: The Client Data Portal</strong></p><p>Let’s say, you are an agency or SaaS platform. You need clients to send you their data. They send you messy Excel files with different column names every time, and it took you huge time to decide to accept it or reject it. It’s a nightmare to standardize.</p><p>Now, you can build a flow once that standardizes common errors (e.g., mapping “Email Address” and “e-mail” to a single email column), setup rules to validate if the data complies to your requirement.</p><p><strong>Publish a flow </strong>as a customer-friendly web component and embed it in your customer portal or website.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/507/1*WPka0Sz7em0RqhRO2DPf9A.png" /><figcaption>Embed Flow To Accept Uploads</figcaption></figure><p>Your client uploads their messy file, <strong>Platform-X</strong> cleans it instantly based on your rules, and gives them immediate feedback or success.</p><p>Your customer is happy with real-time response, and you are happy to see clean data in your system only.</p><h4>Summary &amp; Action</h4><p>The era of manual data wrangling is over. We are moving from “Data Janitors” to “Data Storytellers”. It’s time to reclaim your 10x capacity and refocus on high-level strategic work.</p><p>At <strong>Columns</strong>, we are building <strong>Platform-X</strong> to a reality. Columns Flow is currently in beta and would love to invite you to experience the shift firsthand.</p><p>Join the <a href="https://columns.ai/flow"><strong>early access invitation</strong></a> to turn your manual grind into an automated flow today.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=14452a7aa499" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>