Getting large numbers of product images online can be a tedious job, and if you’re using Shopify’s built-in tools, you’ve probably run into the pain of having to manually assign images to each product.
At StudioForty9, we’ve developed a bulk image uploader that automatically assigns uploaded images to the right product, but we recently discovered that it wasn’t enough for one of our clients. An internal requirement meant they still had to manually adjust the order of images on their product pages.
So we built them an AI tool to handle the reordering of product images.
What used to take up to three hours every day now takes a maximum of 30 minutes.
That’s up to 12.5 hours saved every week with one simple tool.
The problem our bulk uploader didn’t solve
During a regular client call, Kasia Ambroziak, Ecommerce Production & System Executive at Shaws, mentioned that despite using our bulk image uploader, she still had to manually reorder images on a lot of product pages.
Shaws is a family-owned Irish department store retailer with stores nationwide and an established omnichannel ecommerce operation, so the team deals with a constant flow of new product imagery.
StudioForty9’s bulk uploader had already made that job much faster. It allows you to upload images in batches and automatically assigns them to the right products by checking the filename for a unique product identifier. Shaws said this alone was a “game changer” because it’s such a huge improvement over Shopify’s native image-uploading process, which requires a significant amount of manual intervention.
However, Shaws also has its own preferences for how product imagery should be ordered.
The bulk uploader can only go so far. It honours any ordering number included in the filename, but supplier numbering doesn’t always reflect Shaws’ preferences, so Kasia still had to check and reorder images manually.
That final reordering step was mind-numbingly tedious. Kasia had got used to doing it, but it was so repetitive and boring that sometimes after about 20 products she would lose track and have to go back and double-check the work she’d already done.
Could AI handle the human step?
When Kasia mentioned this manual reordering step, Marko Čepo’s ears immediately pricked up. As one of our core back-end developers on the integrations team, he’s always on the lookout for ways to automate tedious work. In his own work, any time he finds himself doing the same task for a third time, he starts assessing whether it can be automated.
These days, Marko regularly spots even more automation opportunities because AI can often handle steps that previously required human judgement. Looking at images to determine which product view is being shown is exactly the kind of task that couldn’t traditionally be automated, but AI’s computer vision capabilities have changed that.
“How long are you spending reordering the images?” he asked Kasia. Until that point, he had no visibility of the additional work happening after the bulk uploader had done its job.
Kasia explained that it could take up to three hours a day.
“You’re joking!” Marko exclaimed.
Kasia wasn’t joking. “In fact,” she said, “it’s not funny at all when you have something like 900 images to check.”
“Should we look into whether we could automate this for you using AI?” Marko asked.
The answer was a simple, resounding “YES!”
Marko started testing his theory immediately. He was confident AI could do the job, but which model would be required, and could it be done cost-effectively?
Could a low-cost AI model do the job?
Marko got to work building a tool that would use AI to examine the images uploaded by Shaws, analyse each one, and determine which view of the product was shown. That, in turn, would allow the tool to automatically assign each image to the correct position on the product page, following Shaws’ internal ordering rules.
First, Marko looked through Shaws’ image library and put together a test set containing some of the most difficult images to classify. “I chose the images that even a human might struggle with,” he said, “the ones you need to really lean in and analyse to be sure you’re right.”
Marko’s plan was to test the image analysis with smaller AI models, then move to larger, more capable models until he found one that would accurately label the images. The fear was that a larger, more expensive model might be required, which could make the solution too costly to run.
“I knew the task could be done with AI, but what surprised me was that even with difficult images, the smaller models gave me 100% accuracy in testing. In the end, we settled on using Haiku 4.5, which is a very fast and relatively cheap model.”
The tests had been a success, and Marko had identified an affordable AI model that could do the job accurately. He integrated the new functionality into the existing bulk image uploader, keeping the experience seamless for Shaws while dramatically reducing the amount of manual checking required.
From three hours of manual checking to 30 minutes
Once the new image-ordering feature was in place, the amount of manual work dropped dramatically. On busy days, the three hours Kasia had previously spent checking and reordering images was reduced to 30 minutes of spot-checking, cutting the time involved by as much as 83%. That’s up to 12.5 hours per week saved on a mind-numbingly boring task.
The bulk image uploader subscription is €25 a month and processes around 3,500 images a month for Shaws. In its first two months, the AI reordering tool automatically corrected the position of almost 90% of those images.
Kasia also noted an unexpected benefit. Because the tool reliably puts a suitable front image first when one has been supplied, she can now spot problems with supplier imagery at a glance. If the main image is still a back view or close-up detail, it’s an immediate signal that the image set is incomplete, and that the product needs attention before it can go live.
For Kasia, the automation adds another layer of support to a repetitive daily process, reducing how much the team needs to check and making it less likely that something gets missed. She describes the tool as “a small assistant that is working with you”, taking care of the most boring part of the work while leaving the team to focus on the exceptions that still need attention.
Which manual tasks are worth another look?
Something like reordering product images isn’t a complex task. But when somebody has to repeat it across hundreds or thousands of products, it takes up a significant chunk of their time. When thinking about where AI could be applied, it’s easy to overlook these simple but time-consuming tasks.
That’s especially true when a process has always relied on a team member to manually check something before it can continue. You get so used to a human being needed that it’s easy to miss the fact that AI has changed what can now be automated.
The challenge is identifying which processes AI can now improve. That’s where having people who understand both the technology and your operations becomes invaluable.
As Kasia puts it, “Marko is very good with not only answering my question, but actually asking me more questions.” She says that helps her step back from the immediate task and see opportunities to improve the wider process.
The best AI opportunities are specific to your business
AI is part of everyday work for most people at this point. You might turn to ChatGPT or Claude to research, analyse data, or help develop content. If you’re on Shopify, you might already rely on Sidekick, which simplifies tasks that once required significant technical expertise.
These are the highly visible applications of AI. Assistants you can talk to, ask questions of and bring into your everyday work. They’re enormously useful. But everyone has access to them, and as adoption grows, working with these tools is quickly becoming a baseline expectation rather than an advantage.
Some of the more interesting opportunities lie in the less visible applications, where AI does the heavy lifting in the background. The cumulative advantage of finding a number of even the simplest applications can be huge.
On top of that, the more specific the problem is to your business, the more likely AI automation is to create a genuine advantage, because it’s harder for competitors to copy something uniquely built around the way your business actually works.
So if you’re an ecommerce retailer trying to figure out where AI could give you a competitive advantage, reach out to us. We’d love to have a conversation.
