What If AI For Work Was Just Like TikTok's For You Page?

@jgreze
Jean-Denis Greze đź’ˇ @jgreze
Thursday, March 26, 2026 AI

Tweet

https://t.co/ccZ6T7Odtn

X Article

TikTok never asked you what you wanted to watch. You opened the app and it just knew. It watched what you lingered on, what you skipped, what you rewatched at 2am. No preference forms. No search bar. No curation. It observed, learned, and served. That single insight turned a lip-syncing app into one of the most valuable companies on earth. Now imagine that same idea. But for your job. For your email. Your calendar. Your entire life. That’s TikTok FYP for Work. AI's Dirty Secret Here's the thing nobody building with AI wants to admit: most people have no idea what to do with AI. Because every AI product on the market asks them, asks you, to do the work. You have to realize you have a problem AI can solve. Open a product. Write a perfect prompt. Evaluate the output. Paste it back into whatever you were doing. That's five steps where the human does everything and the AI just... responds. That's not a revolution. Meanwhile, you're drowning. 28% of your workweek goes to email. You spend hours drafting status and project updates, on meetings that should've been a Slack message, copying data from a doc to a spreadsheet, and admin work that makes you fantasize about quitting. AI could handle all of it, today, but it won't, because it's sitting in a tab you forgot you opened, waiting for a prompt you'll never write. The One Thing TikTok Got Really Right TikTok's entire product philosophy fits in one sentence: don't make the user do the work. Not: here's a blank text box, good luck. The chat box isn't bad. Chat is a great interface once you know what to say. But making it the starting point — the first thing a new user sees — is like handing someone the keys to a Ferrari before they've ever seen a car. What TikTok understood is that you earn the right to a relationship with the user by showing value first. Not by asking them to figure out your product. What This Could Actually Feel Like Imagine you open your laptop on Monday morning and your AI has already: Prepped briefing docs for every meeting on your calendar Flagged the 3 emails that actually matter (and drafted replies to the other 40) Booked the plumber you've been putting off for two weeks Found open spots at summer camp for the kids and registered them Written your weekly status update based on what you actually shipped Put draft emails in your CEO's outbox for your top five key prospects Inserted mentions to a product your client shipped yesterday into your QBR deck for your call later that day You didn't ask for any of it. But it's what you would have spent the first 2 hours of your day working on! And here's the magic: now you're curious. Now you lean in and think, "wait, what else can this thing do?" Now you open the chat, voluntarily, and type: "How many of my accounts have tried the new enterprise sharing feature we released last week?" “Prospect users who recently visited our website?” The AI earned those tasks by acting first. Action first. Chat second. That's the unlock. Why This Will Be the Biggest Product Shift of the Decade Every major platform shift follows the same pattern: the technology exists for years before someone figures out the right product. Touchscreens existed before the iPhone. Short-form video UGC existed before TikTok. AI models are incredible right now, and yet almost every product built on top of them is a text box. The company that wins this era won't have the best model. Models are commoditizing. GPT-5, Claude 4, Gemini: they're all converging on "extremely good" (or is it AGI). The company that wins will be the one that stops asking users to come to the AI, and instead brings the AI to the user's work and life. Watches how they work. Learns where they spend time on things that could be automated. And just handles it. TikTok won the attention war by matching content to people better than anyone. The AI winner will match capabilities to people better than anyone. We're Building This. It's Early. It's Working. Obvious disclosure: I'm the CEO of Town, and this is exactly what we're trying to build. I'd be lying if I said we've figured it all out. We haven't. We're early. But here's what I find genuinely exciting: today, roughly 25% of all user interactions on our platform happen without the user ever opening a chat box. No prompt. No typing. Just AI that watched, learned, and acted. Auto-drafted emails, meetings that scheduled themselves, suggestions that surfaced at the right moment, all fully AI-driven. And the user's only job was to say "yes" or "no." Twenty-five percent isn’t AGI. But a few months ago it was zero. We think by end of year, 75% of interactions will be AI-first — the AI acts, the user steers. The chat box will still be there (and it's great for the moments when you want to chat to your AI), but it won't be the main event. It'll be the thing you reach for when the AI has already earned your curiosity. We're nowhere near done. But the trajectory hints at the future: when you stop asking people to figure out AI and instead bring AI to them, they use it. A lot. For things they never would have thought to ask for in the first place. The best AI product isn't the one you chat to. It's the one you don't have to.
Explore Further

This has a linked resource worth reading, a tool worth trying, or an idea worth prototyping

Quick Insight

This is a product thesis arguing that AI tools should act proactively (like TikTok's algorithm) instead of waiting for user prompts. The author (CEO of Town) claims their platform already achieves 25% prompt-less interactions where AI watches user behavior and takes action automatically. It's a compelling vision but heavy on marketing for their specific product.

Actionable Takeaway

Study how your current AI integrations could shift from reactive (user prompts) to proactive (observing patterns and acting first). For your side projects, experiment with AI that watches user behavior in your Chrome extensions or print-on-demand tools and suggests actions rather than waiting for commands.

Related to Your Work

This directly applies to your fintech platform's analytics dashboards and webhook integrations. Instead of users manually querying data or setting up alerts, AI could proactively surface anomalies, suggest optimizations, or auto-generate reports based on merchant behavior patterns. Your event-driven architecture is already positioned for this kind of proactive AI layer.

Source Worth Reading

Yes - the linked article is worth reading. It provides concrete examples of proactive AI actions (auto-drafting emails, scheduling meetings, flagging important messages) and includes actual metrics from their platform. The vision is clearer than most AI product thinking pieces, though filter out the company pitch.

Tags

#ai-agents #proactive-ai #product-strategy #automation #ux-patterns