We propose a human-machine hybrid approach to automating decision making in high human-interaction environments and apply it in the business-to-business (B2B) retail context. Using sales transactions data from a B2B aluminum retailer, we create an automated version of each salesperson, which learns and automatically reapplies the salesperson’s pricing policy. In a field experiment with the B2B retailer, we provide salespeople with their own model’s price recommendations in real time. We find that, despite the loss of private salesperson information, reducing intertemporal behavioral biases by providing the model’s price to the salesperson increases profits for treated quotes by 11% relative to a control condition. Using counterfactual analyses, we show that although the model’s pricing leads to higher profitability in most cases, salespeople generate higher profits when pricing out-of-the-ordinary or complex quotes. Accordingly, we propose a machine learning hybrid pricing strategy with two levels of automation. First, a random forest model automatically allocates quotes to either the model or the salesperson based on its prediction of whose pricing would generate higher profits. Then, if the quote is allocated to the model, the model determines the price. The hybrid strategy generates profits significantly higher than either the model or the salespeople.