AI Isn't Going to Replace Your Sales Team. Here's What It'll Actually Do.

Published 2025-02-25 · Updated 2026-06-23 · By Alan Wizemann

Topics: AI & Machine Learning, Digital Transformation, E-commerce, Growth Leadership

When Proof launched, our sales team was fairly convinced it was there to replace them, and the honest thing to say is that they were not wrong to worry. It helps to think about it from their perspective for a moment. Here was a digital platform available twenty-four hours a day, seven days a week, carrying tens of thousands of SKUs from hundreds of suppliers. It could handle reorders, show inventory, surface product information, and let a customer place an order at two in the morning when the bar has finally closed and there is time to review what is actually needed. A salesperson working a normal day, with a limited and precious amount of face time with any given customer, simply cannot compete with that on volume or availability. So the fear made a great deal of sense, and I never pretended otherwise to them.

But what actually happened turned out to be the opposite of what they feared. Reorders picked up on Proof. After-hours business grew because that is genuinely when many of our customers – bars and restaurants and retailers – have the time to manage their orders. Every transaction that happened on the platform, regardless of where it originated, was credited to the salesperson. That was not an accident or a temporary concession. The salesperson owns the relationship with the customer, and that is how it was set up and how it needed to work in order to build any real trust in the tool. The more interesting development, though, was not the reorders at all – it was the discovery. A salesperson in a thirty-minute in-person visit can cover the main products, talk through a few new releases, and maybe pitch one or two things the customer has not tried before. But they cannot reasonably walk through the full assortment, and they almost never have the time to mention the mixers, the maraschino cherries, or the category adjacencies that could genuinely grow a customer's business (and, not incidentally, increase that customer's order volume with us at the same time). Proof does exactly that. At midnight, while a customer is checking their reorder list, they can stumble onto products they would never have encountered in a sales visit, and that discovery adds up. Customers find things that fit their assortment, salespeople watch their numbers climb on items they never once had time to pitch, and the platform quietly becomes a revenue channel for the relationship rather than a threat to it.

We have been building on that foundation with what we call the "next best action," which is our AI recommendations layer. The concept starts from the fact that we sit in a genuinely unusual position in this industry, right in the middle of the network between thousands of suppliers and hundreds of thousands of customers. We understand our customers' assortments, their ordering patterns, their business types, and their locations. At the same time we understand our suppliers' inventory, their marketing priorities, and what is coming next, which means we hold data at a depth and specificity that most players in this space simply do not have access to. The next best action takes all of that and puts it in the hands of the salesperson before they ever walk in the door – not a generic product push but a recommendation built from what that specific customer has ordered before, combined with what the salesperson already knows about their business and their neighborhood, combined again with what our data shows about buying patterns in that particular market. When the recommendation arrives, it arrives with the reasoning behind it rather than as a number the salesperson is just expected to trust.

What actually makes it work, more than the data or the model, is the feedback loop. This is deliberately not a thumbs-up-or-thumbs-down rating system. When a recommendation falls off the mark, or when the salesperson already sold that product to the customer the day before, they can tell us why, and that input is structured, processable, and routed directly into improving the model. The salespeople are not simply using the tool – they are building it. That is not an accident either, because the entire goal was to make the sales team feel like a part of the technology rather than a subject of it. Their expertise, the neighborhood knowledge, the customer relationships they have built over years, the brand stories they carry around in their heads, all of that is information no model will ever generate on its own, so it has to come from them. By making their feedback part of the process, the model gets better, they feel genuinely heard, and the tool stays grounded in reality instead of drifting toward recommendations that look impressive on a dashboard but make no sense at all on the ground.

There are things the platform does well and there are things it will never replace, and I try to be honest about both. What Proof and all of our AI tools are built to do is free up the salesperson to do the things that only they can do: walk a customer through a new brand story, drop off a sample that lands at exactly the right moment, know that this particular account always picks up on local spirits and bring something from precisely that angle, understand a neighborhood well enough to grasp why one account's assortment looks completely different from another's three blocks away. That is where the relationship actually lives, and that is where our sales team's time should be going. The framing I keep coming back to is a simple one: build AI to take things off their plate, not to take their plate. The fear that AI will replace the sales relationship is pointing at the wrong problem entirely. The real question is not whether AI can handle eighty percent of what a salesperson does on a transactional level – it can – but whether you use that capability to eliminate the salesperson or to multiply what they are able to do with the twenty percent that genuinely matters, the parts that require being human, knowing the customer, and being present in a way a platform never can be.

Every tool we build at Southern Glazer's is pointed at that second answer. That is not because it sounds better in a town hall but because it is also, quite plainly, the right business decision. What makes this company different in the market is the quality of the relationships our sales team has with customers and suppliers, and that is a real competitive advantage rather than a sentimental one. So the last thing we would ever want to do is quietly automate it away. The way we think about it, in the end, is that every hour the platform saves our salespeople is an hour they can spend doing the things only they can do, and that is the trade we are trying to make – and so far, it is working.