When is it wise to use artificial intelligence for platform operations considering consumer returns?

We consider a supply chain comprising a manufacturer and a platform, where the former sells its products through an offline channel and the latter. The platform has two operational modes, namely the marketplace and reselling modes. Platform power refers to the platform's ability to increase the market size, while network effects concern the platform's effect on the offline channel. Consumers may return dissatisfied products brought from the platform within a return window. Without artificial intelligence (AI), the optimal production quantities in both the marketplace and reselling modes increase (decrease) with the platform power under low (high) network effects. The optimal profit of the manufacturer in the marketplace mode increases with the platform power and the direction may be positive in the reselling mode. With AI, the optimal profits of the manufacturer in the two modes always increase with the platform power. Comparing the cases with and without AI, we find that using AI in the marketplace mode increases (decreases) the manufacturer's profit under low (high) network effects. And using AI in the reselling mode increases (decreases) the platform's profit under low (high) network effects. With and without AI, the marketplace mode cannot coordinate the supply chain, while the reselling mode can. We also extend our work by considering several cases to check the robustness of the results.