AI Is Turning HTML Into a Living Document for Work Teams

The New Office: When AI Starts Writing the Files

For decades, the office ran on Word, Excel, and PowerPoint. Those files were static—you opened them, edited them, saved them, and sent them. But AI is changing what a file can be. Instead of text on a page, AI increasingly outputs interfaces, dashboards, and interactive pages. HTML, the language of the web, is becoming the new canvas for work.

Recently, engineers like Thariq from Claude Code argued that HTML is a better output format for AI than plain Markdown. Andrej Karpathy, a well-known AI researcher, also noted a shift: humans are better at speaking requests, while AI is better at showing complex information visually. That means the output of AI isn’t just paragraphs—it’s pages, charts, and interactive tools.

But there’s a catch. Even if AI can generate HTML, most office workers can’t edit it, share it, or keep it updated. That’s where WorkBuddy’s latest update comes in. They’ve built a “knowledge base” that turns AI-generated HTML into something that behaves like a living document.

From One‑Off Generation to a Place Where Work Lives

WorkBuddy isn’t a brand‑new tool. It already let people and AI write together in Word, Excel, and PowerPoint—what they call “human‑machine co‑writing.” The new update expands that to HTML and smart Markdown, and adds “light apps” that give HTML the ability to store data and sync across devices.

The idea is simple: when an AI task finishes, the output becomes the starting point for the next task. Instead of a dead‑end file, you get a living artifact. WorkBuddy organizes this into two areas: “My Documents” for personal work (with optional sharing) and “Team Space” for collaboration. In the team space, members can view, edit, comment, and share—and AI agents can join in too.

In practice, it works like this: you finish a task, save the final document—HTML, Markdown, or CSV—into the library. Next time, you point the agent to that file, and it reads the existing content to continue the work. When the team needs to maintain it, you move it to the team space and set permissions. That’s it.

A Real Test: Building a Project Dashboard

To see if this actually works, I tried it myself. I gave WorkBuddy a set of raw project data: five projects (car launch, website redesign, store expansion, etc.), each with owners, departments, status, budgets, and milestone dates.

WorkBuddy turned that into a structured data table, then generated an HTML dashboard that looked like a lightweight project management board. I could filter by status, see budget progress, and check who was responsible for what. Not bad for a first pass.

But the real test was updating it. Instead of regenerating the whole page, I linked the dashboard to a CSV file in the library. Then I typed a prompt: “Change the dashboard to read data dynamically from the linked table.” After that, the board pulled live data from the CSV. When I updated the spreadsheet, the page refreshed and showed the new numbers. No code, no regeneration.

Editing HTML Like a Document, Not a Developer

Here’s what impressed me most: you can edit the HTML page without touching code. I clicked a button, selected the “risks” section, and asked the AI to lighten the background and add a note: “Needs a project meeting by Friday.” The change applied only to that section. The rest of the page stayed intact.

That might not sound revolutionary, but for non‑programmers, it changes everything. Before, editing a web page meant either regenerating the whole thing or diving into source code. Now it’s like editing a document—you select, you type, you’re done.

And when you’re ready to share, you don’t need a server or a domain. WorkBuddy can publish the HTML as a live web page with a link. Your boss can open it on their phone, your client can click it in WeChat, and they can even leave comments by highlighting text—just like commenting on a shared doc.

HTML Becomes a Team Player

This opens up new possibilities for office work. A marketing plan can become an interactive timeline with clickable stages. An operations report can let managers filter data by department. A product proposal can include videos, animations, and a clickable demo. And when something needs a quick change, you just select the text and ask the AI to adjust it.

WorkBuddy also added an AI‑native review mode for smart Markdown. When your team writes weekly reports, PRDs, or proposals, you can focus on the content in Markdown. The AI’s suggestions appear as revision marks, and you accept or reject them one by one. Once the text is final, you can convert it to HTML for a more polished presentation.

That creates a natural split: Markdown for lightweight writing and collaboration, HTML for visual presentation and interactive sharing. Both are editable, both are shareable, and both stay in the library for future use.

Lightweight Apps: Giving HTML a Memory

The next step is making these pages actually store data and sync. WorkBuddy calls these “light apps.” I tested it with a bakery inventory CSV—product names, SKUs, stock levels, safety thresholds, costs, and restock dates. WorkBuddy generated an inventory dashboard that flagged low‑stock items and let me filter by category. The CSV stayed as the data source, so the page wasn’t a one‑time snapshot; it became a mini tool I could keep using.

I also tried a personal use case: I put my daily energy levels, stress events, and recovery activities in a CSV, and WorkBuddy made a “life battery” dashboard. It tracked how my week was going. You could even use it for couple’s chores and points, or a lunch‑picker based on nearby restaurants’ prices and ratings.

These aren’t meant to replace full‑fledged databases. But for individuals and small teams, you often just need data to be saved, accessible on multiple devices, and shareable—while also being something AI can pick up and work with next time.

The Two‑Generation Office

WorkBuddy’s approach doesn’t try to kill Word, Excel, or PowerPoint. Those files still work, and the tool still supports them with co‑writing. What’s changing is that office software now accommodates two generations of files.

On one side, the classic trio built for human hands and eyes. On the other, Markdown, CSV, and HTML—formats that AI can generate, understand, and modify more naturally. The “file” you open each morning is no longer a static object. It’s a working surface that AI can edit, update, and turn into pages, all while you and your team keep collaborating around it.

File formats have always shaped how we produce, modify, and share information. In the office era, that meant documents. In the AI era, it means living web pages. And the participants have expanded from just people to people plus agents.

WorkBuddy’s update is a small but telling sign: the office is getting a new set of tools, and they speak HTML.