After scrolling through the 8th image generation tool ranking, I tweaked WorkBuddy's asset library again

I scrolled past that 2026 AI image generation leaderboard, seeing Flux.1 and Midjourney V6 at the top with nice scores. I glanced twice and closed it. It's not that image generation tools are useless; it's that for my work, the real bottleneck has never been generation—it's organization.

Last week I took an e-commerce order from an old client: over 200 product images, mixing white background, lifestyle scenes, details, and model shots in one folder. The client needed different sizes for various platforms. I spent forty minutes flipping through folders, my back hurting worse than after a day of outdoor shoots.

Actually, I encountered this pain point two weeks ago with WorkBuddy. Back then it was wedding footage, hundreds of RAW files needing archival by workflow, and I took many detours. This time, I specifically reorganized the configuration and managed to establish a reusable workflow.

Let's start with the key implementation configuration.

My current approach is creating an Action in WorkBuddy called "E-commerce Delivery Archival," then setting classification rules. To explain the term: an Action is a fixed processing pipeline, like giving it a template. Next time you throw a whole folder in, it runs according to these rules. In "New Rule," I selected classification by content features, instructing it to read filenames and image info using four keyword groups: "White Background," "Scene," "Detail," and "Model." This step reads local file paths, not uploading images to the cloud—I confirmed this because I was previously worried about privacy.

The first test run had issues. It categorized a bunch of white-background images into an auto-generated category called "Clean Background" instead of my preset "White Background." Later, I went into settings and turned off the "Auto-create new categories" option, forcing it to match only within my four preset categories. This toggle is on the classification rules page, scroll down to find the checkbox for "Allow AI to autonomously expand categories," which is on by default. After turning it off, accuracy jumped from about 70% to over 95%.

After running, it marks each image with its assigned category. Then I ask it to generate an Excel list in each subfolder, listing filename, original path, target path, and a remarks column. The remarks column is for the client's proofreading; I asked it to add a judgment on "Suggested Delivery Size." However, it often guesses wrong here, so I stopped letting it guess. Instead, after generating the list, I paste the spec sheet into the last column myself. This manual step takes ten seconds.

For permissions and collaboration, I divide it this way. For clients, I share directories via "Read-only links," sending them the list and filtered thumbnails for confirmation. This entry point is in the Share button on the folder page; select "View Only," don't enable "Editable." Internally, I gave my partner and freelance retoucher "Editable" permissions, but restricted to their respective subfolders. WorkBuddy's collaboration permissions can be set individually per folder. Change this in the member management section of the root directory. Don't use the default "Full Library Editable," otherwise if one person changes classification rules, the archival logic for the entire library gets messed up.

Daily operations happen roughly every three days. I drop new materials into the "To Archive" folder and run the previously created Action. The "To Archive" folder is a named entry point I set under the root directory. It's physically a folder, but WorkBuddy defaults to monitoring this directory; I named it for easy identification. The process takes about two minutes. Afterward, I check if any unmatched files remain in "To Archive." It doesn't touch unmatched files, which is actually a feature: it prefers leaving them alone rather than stuffing them randomly.

The most tangible improvement is that last Wednesday afternoon, the client requested urgent delivery of full detail page assets by 7 PM. I started archiving at 5:30 PM. Including the time to generate the Excel list, it took about twenty minutes total. I used the remaining hour-plus to double-check Dolby colors against the list. Previously, just finding images took at least two hours.

Regarding trends, I'd say since these image generation tools started competing on scores, I feel they're getting further from photographers' real needs. Copyright of generated images, style stability, consistency with live-action shots—no one seriously addresses these practical issues. Meanwhile, local file management, which sounds unsexy, has become indispensable daily. Scores on image generation leaderboards keep refreshing, but those hundreds of originals on my hard drive still need WorkBuddy to classify them properly. This value won't be replaced in the short term.

Maybe it's because I'm getting older. No matter how much I chase visual novelty, grounding myself by archiving negatives feels safer. I already forgot the name of the tool ranked 8th on that list, but I remember the 8 images in the "To Archive" folder that couldn't find a home.