AI and machine learning has been making inroads into many aspects of commerce, with one of the notable areas being retail. Leaders like eBay are using a combination of AI and humans to achieve the best they can in customer service.
In a simplistic sense, AI basically uses data and algorithms to achieve the intended results. While powerful, modern AI has its limitations. For example, humans are better than machines at empathizing. Customers act differently with humans than they do with machines. Humans are better than machines in establishing the contest of particular purchases - for example, wedding and anniversary gifts call for different sensitivities than many other more routine purchases. Machines on the other hand are better at finding nuances in data being used, in determining product categories for particular market segments, even in determining customer clothing sizes.
What this means is that human and machine activities must be carefully thought out and implemented. Experience has shown that these activities should be as discrete as possible, so they don't overlap.
Effective implementation of AI also requires the means to update that strategy as experience accumulates.
For a particular use-case, check out this link.