When AI Did 80% of the Grunt Work, I Finally Had Time to Think

The night before last, J dropped a screenshot in the group.

It was our team’s content output stats for the past three months—articles, image assets, social posts—combined, nearly 2.4x more than the same period last year.

I stared at that number for a good few seconds. Not because I was happy—because I was confused. We didn’t hire anyone new. Same team, same 24 hours each day.

Then I thought back to what actually changed these three months, and the answer was clear: it’s not that we have more people. It’s that the workflow got completely torn down and rebuilt with AI tools.

Looking back at what we did three months ago, it’s basically digital archaeology

My old blog post workflow used to go something like this: 40 minutes finding references and data, over an hour writing the first draft, another half hour finding images or piecing together visual assets, then proofreading, SEO optimization, formatting, and publishing. From zero to published, average 3.5 hours per article.

Image assets were even more ridiculous. Finding the right cover image meant either scrolling through stock sites forever, or opening Canva and slowly piecing things together. One blog cover image regularly ate up 30 to 40 minutes.

Now?

I use Imagera AI to generate the first version of the cover, then throw it into Picsart Flows to run an automation workflow I set up beforehand—color correction, cropping, applying brand elements. The whole process takes under 5 minutes. The first time I set it up was a bit of a hassle, but once it’s done, it’s one-click each time.

For writing, Claude handles the early-stage data organization and fact-checking. Before, I’d have to dig through reports one by one and cross-reference sources. Now it can pull together the key data and citations in minutes. But the perspective and voice? That’s still me. AI does the grunt work, the voice is mine—that line can’t get blurry.

It’s not about having more tools, it’s about finding your combo

I’ve tried a lot of AI tools. Really lot.

In 2026’s AI tool market, within the top 100 rankings, ChatGPT still holds about 60% market share. But here’s the interesting part—DeepSeek jumped to 3.2%, Grok broke 3%, and Perplexity and Claude each took 2%. The numbers look small, but each found its own positioning with very clear user profiles.

My take: don’t go looking for an “all-in-one tool”—because that thing doesn’t exist.

The combo that finally stuck: Claude handles long-form understanding and writing assistance (it’s genuinely strong at keeping context), Perplexity handles real-time research and cross-referencing, Imagera AI takes care of image generation, and Picsart Flows automates batch post-processing for images.

But the most critical change isn’t how amazing any single tool is. It’s that J helped me chain these tools into an automated workflow. From data gathering, draft assistance, image generation, SEO checks to scheduling and publishing—a lot of steps run automatically. My actual decision-making time each day went from 4-5 hours to about 1.5 hours.

So what did I do with the saved time? Think about new content directions, research market trends, and—finally have time to watch the markets properly (haha).

Double efficiency is just the surface—the real thing is the mindset flip

The numbers are definitely pretty. Content output up 2.4x, article production time from 3.5 hours down to 1.2 hours, image prep time cut by 85%.

But honestly, that’s not the part that hits me the hardest.

The biggest change is how I view work now. Before, I spent a huge chunk of time on “execution”—typing, finding images, formatting, proofreading. Now AI handles most of that, so I have more mental space to think about “why am I writing this” and “where are readers actually getting stuck.”

For example: before, crank out 2 articles a week and I felt like I was pushing it. Now same time, can produce 4-5 articles—but the quality didn’t drop. Actually, because I have more time to polish the viewpoints, the articles are deeper than before.

Everyone else on the team feels it too. Mimi does marketing research—used to spend a whole day crawling data and compiling reports. Now with AI assist, half a day gets her a more complete analysis. Ada’s debug efficiency in product development also improved a lot—though sometimes he still gets stuck on weird bugs, of course.

But it’s not all sunshine either

I gotta be real about the downside.

The learning curve for AI tools is real. You don’t just download it and magically know how to use it—you gotta spend time understanding each tool’s characteristics, limits, and where it plugs into your existing workflow. It took me almost a month to dial this workflow into its current state. Along the way, I hit a lot of walls and dropped several “looks super impressive but actually doesn’t work for me” tools.

Plus, there’s a more fundamental issue: content that’s overly reliant on AI generation loses its soul. I’ve seen too many AI-written articles—every one is smooth and complete, but reading it just feels… empty. Like factory assembly line canned food.

So my principle stays the same—AI does grunt work, humans make decisions. Let AI help prepare the materials, but the perspective, tone, and stories have to be mine. No matter how powerful the tool, it doesn’t know what I thought about when watching the market today, doesn’t know the subtle默契 between me and my readers.

That stuff can’t be automated, and it shouldn’t be.

Before, my attention was probably 80% on execution, 20% on thinking. Now it’s flipped.

That flip matters more to me than any efficiency metric.