We study how delegating financial decisions to humans or to automated, algorithmic systems affects reciprocity in credit relationships. Using a modified trust game, we show that borrowers are less likely to repay when lending decisions are delegated rather than made directly by the lender. This reduction in reciprocity is larger when the decision is delegated to an automated algorithm than to a human agent, even when monetary incentives and observability are held constant. At the same time, lenders are more willing to delegate to algorithmic than to human agents. Borrowers explain lower repayment after delegation, especially to algorithmic agents, by a perceived lack of lender effort and involvement. Our findings show that automation in financial decision making can weaken reciprocity and increase default risk by changing how borrowers perceive lender intentions.