How We Taught AI to See What We See, And Then Some

Our 1,500-hour time saver just got a major intelligence upgrade. Here’s how we’re closing the loop on image selection with a self-learning AI.

Last month, we shared how we helped a client eliminate over 1,500 hours of tedious work through automation: replacing a manual daily review process with an AI-powered system that outperformed the human it freed up. (Missed it? Catch up here.)

That alone was a big win. But we weren’t done.

This client reviews hundreds of images every single day — over 700, to be exact — and needs to identify the best ones for public-facing use. It’s a job that used to rely on instinct, intuition, and a very tired set of human eyes. Our initial AI solution sped things up tremendously, but we noticed some edge cases where the AI struggled to pick up on subtle visual traits that humans naturally recognized.

So, we gave it a vision upgrade.

Using a technique called CLIP image embeddings, we essentially gave the AI a new language for understanding images. It breaks each image down into 512 distinct points of comparison. Think of it as a detailed fingerprint, unique to each photo. We trained the system on a handpicked dataset of over thousands of past “great” images, helping it understand what “good” really looks like in this specific context.

You could say we added another dimension... or 512 of them.

Intergalactic precision, now in production.

The result? A smarter system that doesn’t just guess what looks right, it knows. And now, it’s outperforming even the best human reviewer with unprecedented accuracy.

But here’s where it gets really exciting.

These selected images go out on high-reach, monetized social channels with over 320 posts per day. We’re able to track how much each image earns, then feed that revenue data back into the system. That means the AI doesn’t just learn from human preference, it learns from real-world performance.

This is a closed-loop, self-improving AI. It selects, learns, adapts, and optimizes itself without additional human input. The more it runs, the better it gets. It’s no longer just about saving time — it’s about unlocking a level of precision and consistency that wasn’t possible before.

What’s next?

We’re continuing to refine this loop, exploring how visual trends shift over time and adapting dynamically. As the system evolves, it’s not just supporting growth, it’s driving it.

Stay tuned.

‍