This AI knows what your customers want before they do

Can an AI know what a customer is going to buy before they do? Partly, yes. It doesn't read minds, but it analyzes behavior patterns—what they look at, what they compare, what they've bought before—and uses that to anticipate which product has the best shot at adding value. SalesPilot, the AI engine behind CrossUp, does exactly that: it picks the best recommendation for every visit and shows it at the exact moment of purchase. In this article, we'll break down how this machine learning technology works and why automating your recommendations changes the numbers for your store.

What is SalesPilot?

SalesPilot is the artificial intelligence engine that powers CrossUp's recommendations—the upsell app for Tiendanube, Nuvemshop, and VTEX. It works like an expert salesperson who never gets tired: it watches every customer interaction and recommends products in real time, automatically. Thanks to its machine learning system, it doesn't just automate the recommendation process—it personalizes it based on each user's preferences and context.

Put simply: it does at scale, for every single visit, what a great in-store salesperson would do face to face with one customer.

How does machine learning work in SalesPilot?

At the heart of SalesPilot is a machine learning algorithm that analyzes data in real time to generate accurate recommendations. It rests on three pillars:

• Real-time analysis: it examines every click, search, and purchase to identify behavior patterns.
• Personalized recommendations: it suggests products based on past purchases, current browsing, and the real affinities within your catalog.
• Continuous optimization: it adjusts suggestions based on how customers respond, so it recalibrates itself whenever stock, trends, or seasonality shift.

That last point is the real edge over fixed rules: you never have to rewrite anything when your store changes. The model learns from the new data.

What your store gains by automating recommendations

Automating recommendations with AI pays off on both sides of the sale.

For e-commerce owners

• More revenue: it raises your average order value (AOV) by suggesting the right add-on or the right upgrade. Across the stores we measured, average order value grows +9.2% on average (results vary by store, catalog, and traffic).
• Zero manual setup: there's no need to enter recommendations one by one; the engine builds them on its own.
• Adaptability: it adjusts in real time to changes in stock, trends, and season.

For your customers

• A better shopping experience: they get useful suggestions effortlessly.
• More satisfaction: they find what they need fast and think more highly of your store.

Does the AI replace your business judgment?

No—and this matters. SalesPilot does the heavy lifting of deciding what to recommend on every visit, but you stay in control: you can exclude products, prioritize new arrivals or clearance items, set price limits, and define the copy, the colors, and the moments where the recommendation appears. The AI handles the scale; your knowledge of the business sets the boundaries. We compare both approaches—and why combining them wins—in AI-powered cross-selling vs. manual rules.

And that work doesn't happen in just one place: the recommendation kicks in at each of the 6 conversion moments in e-commerce, from the product page all the way to the thank-you page.

In summary

A recommendation AI like SalesPilot analyzes your customers' behavior in real time to show the right product at the right moment, and it recalibrates itself when your store changes. The result: a higher average order value from the traffic you already have and a better shopping experience—without losing control over what gets recommended. It's letting technology work for you, with your judgment calling the shots.