Abstract Self‐service technologies (SSTs) have been widely adopted in industries that require the delivery of physical products by services, in which consumers evaluate a product for both the product value and the service value. A typical service delivery system usually involves sales agents and/or self‐service technologies, for example, online services and kiosks, to serve consumers by a coproduction process. In other words, both the sales agent (or self‐service machine) and the consumer should exert effort, with corresponding service costs. By modeling the coproduction output with a Cobb–Douglas production function, we establish a principal–agent model to study the value of self‐service technologies in designing a service delivery system wherein the sales agent's service cost is private information. We first characterize the main trade‐off between the sales agent and the self‐service machine when the firm provides only one service channel. Then, we analyze the value of the self‐service machine when the firm can provide both service channels. We find that, interestingly, the firm may possibly provide both service channels, that is, services offered by the sales agent and the self‐service machine, when the level of information uncertainty is high, and the self‐service machine's service cost is intermediate. Moreover, when both service channels are offered, only the efficient sales agent will be contracted, and the inefficient sales agents are screened out of the market by the self‐service machine; that is, the self‐service machine can help the firm eliminate the information rent. We also investigate how the firm's service weight in the coproduction process and information uncertainty influence the consumer surplus, firm's choices, contract parameters, and resulting profits. Our results are shown to be robust when our model is extended to consider a single‐contract strategy, contracting on effort, a continuous sales agent type, and a general coproduction function.