What If I had AI in 2018: Rent the Runway Fulfillment Center Optimization

will become our digital assistants, helping us navigate the complexities of the modern world. They will make our lives easier and more efficient.” Inspiring and completely unbiased statement from someone who already invested billions on this new technology.

The hype is real for AI agents, and billions are pouring in to build models that will make us more productive and more creative. Hard to disagree when I happily enjoy my morning coffee while Cursor is coding my unit tests. Yet, asking people in my network how they use AI in their day-to-day, their answers often mention anecdotal use cases, anywhere from “I use it to tell bedtime stories to my son” (I guess that would not even be a use case if you had more imagination) to “I use it to optimize my schedule” (Motion AI, please stop targeting me for the love of god).

As a Data Scientist, my mind goes back and forth between two conclusions. The FOMO part of me that does not want to be late to the Robot revolution party, and the cynical one that thinks that there is still a long way to go before artificial intelligence actually becomes intelligent. To find out which side of my schizophrenic personality I should bet on, I am going to use a simple yet powerful framework: reviewing all the projects I have worked on since the beginning of my career and assessing how 2025 state-of-the-art AI models could have helped.

Today, we go back to 2018. I am a candid summer intern at one of the most disruptive startups in America: Rent the Runway.

What the Project was about

The Rent the Runway fulfillment center in Secaucus, NJ, used to be the biggest dry cleaning facility in the United States.

In the Summer 2018, as an Operations Analyst intern, I was given a pretty hard problem to think about: everyday, the fulfillment center was receiving thousands of units back from all around the country. All the items had to be first inspected, then would go through a thorough cleaning process, before being dried or receiving some special treatments. This could be:

• Spotting if the garment was stained during the rental
• Pressing if it was too wrinkled and had to be ironed
• Repairing if it had been damaged

Most of these tasks were done manually by different departments, and required specialized workers to be available as soon as the first batch of units were reaching their department. Being able to predict days ahead what volume of units would have to be processed (and when) was crucial for the fulfillment center planning squad, in order to make sure that every operations team would be staffed appropriately.

The complexity of the flow made it even trickier. It was not only about predicting the inbound volume, but also assessing what part of this inbound volume would require special treatments, where and when bottlenecks could appear, and understanding how the work done at one department would impact the other departments.

The 2018 Solution

At this point you may wonder: given the complexity and the stakes of the project, why was it in the hands of a young inexperienced intern? To be fair, during my 10-week summer internship, I only scratched the surface and wrote an insanely complicated Pyomo script that was later refined by a more senior Data Scientist, who spent two years on this project alone.

But as you can imagine, the solution was this huge optimization model taking as an input the inbound volume forecast for every day of the week, the average UPH (units per hour, i.e the number of units that can be processed in an hour) at each department, and some assumptions on the proportions of units that would require specific treatments. The main constraints were on the timing and regularity of the shifts, and the number of full time contracts. The model would then output an optimized labor planning for the week.

How AI could have helped

Let’s re-clarify things first: you will not see words like “AI-enthusiast” or “LLM believer” in my LinkedIn bio. I am pretty skeptical that AI will magically solve all our problems, but I am interested in seeing if with today’s technology, another approach would be possible.

Because our approach was, you could say, pretty old school, and required months and months of refinements and testing.

The main limit is the static aspect of the solution. If something unexpected happens during the week (e.g a snow storm that paralyzes the logistics in some parts of the country, delaying some of the inbound volume), a lot of assumptions of the model have to be changed, and its results are becoming obsolete.

This is a solution that requires data scientists to go deep into the weeds, instead of relying on an out-of-the-box framework, to rely on a lot of assumptions and to spend time maintaining and updating these assumptions.

Could AI come up with a completely different approach? No.

For this particular problem, you clearly need an optimization model, and I am yet to read about an LLM being able to handle a model with such complexity. One could propose a framework with an AI agent acting as a General Manager, and relying on sub-agents to handle the planning of each department. But that framework would still require agents to have tools that allow them to solve a complex optimization model, and the sub-agents would need to communicate as the situation of one department can affect all the others.

Could AI significantly enhance the “human-generated” solution? Possible.

It is at this point pretty obvious to me that LLMs would not make the problem trivial, but they could help improve the solution in multiple areas:

• First of all, they could help with reporting and decision making. The output of the optimization model might have a business sense, but making a decision out of it might be hard for someone with no strong understanding of linear programming. An LLM could help interpret the results and suggest concrete business decisions.
• Secondly, an LLM could help react faster to certain unexpected situations. It could for example summarize information on events that could have an impact on the Operations, such as bad weather in some parts of the country or other issues with suppliers, and as such, recommend when to rerun the planning model. That is assuming it has access to good quality data about these external events.
• Finally, it is possible AI could have also helped with making real time adjustments to the planning. For instance, it is typically predictable based on the garment characteristics whether they would require special care (e.g a cotton shirt will always have to be ironed manually). Having a VLM scanning every garment at the receiving station could help downstream departments understand how much volume they should expect hours in advance.

Could AI enable Data Scientists to maintain and update the model? Yes!

It is really hard to deny that with tools like Copilot or Cursor coding and maintaining this model would have been easier. I would not have blindly asked Claude to code every constraint of the Linear Program from scratch, but with AI code editors being smarter than ever, modifying and testing specific constraints (and catching human errors!) would be easier.

My conclusion is that an LLM in 2018 would not have trivialized the project, although it could have enhanced the final solution. But it is not impossible to believe that a few years (months?) from now, agents with enhanced reasoning capabilities will be sophisticated enough to start cracking these types of problems. In the meantime, while AI could speed up model iterations and adjustments, the human judgment at the core remains irreplaceable. This serves as a valuable reminder that being a Data Scientist isn’t just about solving mathematical or computer science problems—it’s about designing practical solutions that meet evolving, often ambiguous and not so well defined real-world constraints.

Article 100% human generated