How Bulk Scales Merchandising Capacity With an Always-on AI Agent

A large part of any merchandising team’s day-to-day is, by definition, spent on optimization and problem-solving merchandising challenges, and Bulk is no exception: To move quickly, their team needs to understand how results align with the algorithm’s logic so they can decide next steps with it in mind. 
This is where Constructor’s Merchant Intelligence Agent (MIA) came in - with MIA enabled, their merchandising team was able to:

“[MIA] saves you a lot of time. You can’t have eyes everywhere. So, it’s basically picking up on potentially human mistakes or issues that you generally wouldn’t have been able to find yourself, or it would have taken a lot of time to.”

— Dua B., Senior Merchandising Manager at Bulk

With MIA, the merchandising team gained an agent that has full knowledge of all merchandising rules and live performance of their 16 online stores in real time. It could flag potential issues, explain why it’s not as expected, and suggest what to do next, all in one conversation.

This comes in handy when flagging synonym issues and facilitating other cross-country website optimizations. One time, MIA suggested connecting "creatina" and "creatine" as a synonym opportunity for Italy - this went unnoticed as no one on the team spoke Italian at the time. Dua could simply review MIA’s suggestion, decide whether it made sense, and action it.

A good example of this was when the team noticed a sudden revenue drop in France and Italy with no obvious reason. Instead of spending time checking for possible causes across the two countries, Dua just asked MIA to investigate - and it quickly identified the cause: a bestselling product had run out of stock for a few days. The team could then go straight to procurement to sort out the restocking issue.

“I've set it up to flag any product that's been out of stock for more than two days. It means I catch the issue early instead of finding out after the fact.”

The same process also helped with a more complex challenge: Bulk’s team had noticed an add-to-cart rate that looked too high to be accurate. A quick check of the usual suspects revealed no reasons. But rather than escalating the issue for a deeper investigation, Dua asked MIA: Through reasoning and conversation, the agent flagged that a share of traffic was missed in the analysis, causing add-to-carts not to be matched to a click source, which was inflating the results. Following MIA’s suggestion to recalculate the rate with the missing traffic accounted for, the figure was immediately more aligned with the team’s expectations for add-to-cart.

“Normally, something like this could mean an escalation that takes up to multiple days to resolve,” she said. “With MIA, I got to the root cause in an hour, without the back-and-forth.” 

Getting the reasoning behind a situation — along with options to move forward — allows fast-paced ecommerce teams to focus on making the call rather than on investigative and analytical work.

MIA has also helped make collaboration between Bulk’s teams easier and more efficient.

In one case, one of the company’s developers asked her a technical question about data that is being passed from their website to Constructor. A simple question for Constructor’s support team, but an unusual one for Dua. She would have had to ask via email and wait for a response.  Instead, she started by asking MIA, and the response actually helped her understand what was happening.

“The good thing with MIA, as opposed to a human being, is that you can literally say, ‘Explain to me in layman’s terms because I am not a developer,’” she said.

That helped her respond to the developer with more confidence and better understand the broader Constructor set-up, not just the merchandising side. 

“It also helped me be more hands-on and understand a little bit better what’s going on, not just from a merchandising perspective, but also the broader spectrum of Constructor,” she said. “It cuts a lot of middleman processes, in the best way possible.”

For ecommerce leaders, this is one of the bigger operational benefits of MIA: it can help cross-functional teams get unstuck faster without turning every question into an escalation.