The Category Manager Who Can’t Retire

Every organization eventually faces the same question: What happens when the people who know how to do something exceptionally well leave?

That question followed me through Murano, the island off Venice where glassmakers have worked the same furnaces for centuries. The craftsmanship is extraordinary, but what fascinated me wasn't just the glass. It was how the institution endured. You aren't born knowing when molten glass is ready to be worked, or how to move with gravity to coax it into a new form. A thermometer won't tell you. You learn by standing next to someone who already knows. The few who truly master the craft earn the title “maestro,” not because they memorized a process but because they cultivated judgment. The remarkable part isn't that Murano produces legendary glassmakers. It's that it kept producing them for centuries.

I returned home thinking about category managers. Over the past few months, I've sat with grocery leaders in the U.S. and Europe, and two words kept surfacing: natural attrition. Our most experienced merchants are reaching the age where they want to retire, but they’re finding they can't. There's no one standing behind them ready to take over.

Replacing Years of Pattern Recognition

The immediate problem is the pipeline. Few college graduates know category management is even a career. Retailers have invested heavily in digital commerce, analytics, and technology over the past decade, but far less in deliberately developing the next generation of merchants. Career paths are less obvious, mentorship has become less structured, and people move roles before they've had time to accumulate real judgment.

The pipeline is a symptom. Category management is difficult because it isn't primarily about learning processes. It's about developing judgment, and judgment doesn't scale with a training manual.

An experienced category manager told me she spent much of her day mentoring new merchants who were smart, capable, and still couldn't see why certain decisions were obvious. The gap wasn't intelligence. It was experience. Every explanation rested on years of situations they simply had not encountered yet.

We're under relentless pressure to do more with fewer people. Even when we replace a retiring merchant, we don't replace the judgment they've accumulated. It's the quiet pattern recognition that lets you know when to trust the forecast, when to override it because something has gone viral overnight, when the stores won't execute a recommendation, or when a number that looks wrong is actually exactly right.

That's the gap we need to close: the process can be taught in months, but the judgment built through thousands of good and bad decisions takes decades.

Knowing Which Fires to Let Burn

Merchandising is simultaneously an art and a science. The science is optimization, statistical models, and the endless sea of numbers. The art is knowing which numbers matter. In a 200-store chain, thousands of things are "wrong" at any given moment. The science tells us about every one of them in intricate detail. The art decides which three deserve our attention this week, and which fires we let burn because they don't matter to who we are.

When a veteran category manager retires, the art leaves with them. It leaves at exactly the moment the job has never been more demanding. The fundamentals haven't changed, but the number of variables has. Customer behavior shifts constantly. Viral trends emerge before the forecast can catch up. Every decision arrives with more data, more recommendations, and more competing priorities than the last. The number of variables keeps expanding, while the number of people capable of seeing how they fit together does not.

Complexity naturally pushes us toward more optimization: more recommendations, more dashboards, more ways to rank every decision. That works well when the objective is simple, whether it's impressions, clickthrough rate, or conversion. Merchandising asks a different question.

A merchant isn't simply deciding which item is most likely to sell. They're deciding what kind of retailer they want to be. When do we favor local choice over operational simplicity? When do we sacrifice short-term productivity to strengthen a destination category? When do we protect customer trust instead of maximizing this week's margin?

Those aren't optimization problems because the objective isn't fixed. The science tells us how to optimize. The art decides what's worth optimizing.

We've historically developed judgment through apprenticeship. Assistant merchants sit through line reviews, build assortments, defend recommendations, watch an experienced merchant pull their assumptions apart, and then see what happens in the real world.

Murano understood this centuries ago. Apprentices stood beside a maestro until they learned to see what the maestro saw. Retail has worked much the same way, minus the furnace. It only works if experienced and developing merchants spend enough time making decisions together.

The problem is capacity. Apprenticeship scales one relationship at a time. A single experienced merchant can only mentor so many developing merchants, and only while they're still there. As that chain becomes thinner, so does our ability to transfer judgment.

This brings us to artificial intelligence.

Expertise That Survives the Expert

The first question we tend to ask of AI is straightforward: Can it make the recommendation?

Increasingly, yes. It can summarize performance, surface exceptions, test scenarios, forecast outcomes, and recommend actions. These capabilities continue to improve.

The recommendation is only part of the problem. The harder question is how the recommendation was reached in the first place. What did the merchant notice that the model didn't? Which assumption caused them to reject an otherwise attractive recommendation? What did they know about the customer, the supplier, the stores, or the competitive landscape that never appeared in the data?

A grocer in the Midwest put it to me like this: Their veteran category managers had no interest in documenting the obvious. Everyone already knows you don't put an entire category on promotion at once. What they wanted to preserve were the judgment calls that made them “them.” Why they trusted one supplier over another. Why they protected a low-productivity SKU. Why one exception was worth making and another wasn't. They weren't asking us to build a better recommendation engine. They were asking us to teach the system how they think.

There are two ways to approach this. One is to build yet another knowledge repository. Interview the experts. Store the documents. Capture the decisions. Eventually it becomes a data closet that people intend to visit but never do, or an AI model that confidently repeats yesterday's answers because it never learned what changed.

The other is to treat AI as a new form of apprenticeship. Imagine your first day as a category manager beginning with 30 years of experience instead of zero. Similar situations are no longer trapped in someone's memory. A new merchant can explore how prior merchants approached the challenge, where they disagreed, which assumptions survived contact with reality, and which ones didn't. It's the equivalent of compressing 25 years of apprenticeship into the first few years of a career. Like listening to an audiobook at 2× speed, you don't skip the story. You simply reach it sooner.

The final years of a veteran merchant's career can look different too. Instead of asking them to stay because the organization can't function without them, we use the time we have with them differently. We preserve not just their decisions, but the reasoning behind them. Their experience no longer lives only in the categories they personally managed. It becomes the starting point for everyone who follows.

Keep the Judgment Along With the Position

The glassmakers of Murano understood something that feels surprisingly modern. Expertise isn't the goal. Expertise that survives the expert is. The first iteration of AI has made remarkable progress helping us make better recommendations. The next iteration should help us build better merchants by making every veteran merchant's experience available to the people who follow. Retail has always been a people business. If we get this right, the next merchant won't have to spend 25 years rediscovering what the last one already knew.

John Lin is senior vice president, pre-sales and solution architecture, retail, SymphonyAI, a global leader in vertical AI.