The development of X-breed, a knowledge-based decision support system for agricultural advisers consulting on crossbreeding beef cattle, benefited from users' involvement in four types of activities: establishing users' attitudes and needs; iterative modification and expansion of the system after observing user response at various stages of development; field evaluation of the system in the users' work environment; and direct input of knowledge for use in the knowledge base. Our survey showed that the main constraint to computer usage by agricultural advisers was lack of time followed by lack of good programs. Not a single respondent rated decision support systems as ‘not useful’, however advisers wanted systems to be more ‘user friendly’, more reliable/accurate, and require less time to master. The growth of the system and the time spent on user consultation plotted against calender time show that development activity was driven by consultation with users. Even after rigorous technical testing, field evaluation raised substantive technical issues. It also increased the users' acceptance and knowledge of X-Breed. A paradox of acceptability was demonstrated: while the purpose of a knowledge-based system is to make available expertise that is beyond that of the user, confidence in and agreement with the system are a prerequisite for user acceptance of the system. The paradox arises when users disagree with a recommendation that is acceptable to domain experts. Our experience suggests that, in addition to potential users contributing to the knowledge base of the KBS, their involvement produced a valuable, lasting outcome, namely enhancing their knowledge of DSSs and of the biology which underlies the system. It appeared that individual learning, exposure to the DSS at several stages of its development, group discussion and consultation, and gathering of attitudes on an individual basis allowed for interaction among potential users and between users and the knowledge engineer while ensuring that each user was personally ‘heard’ during the development of a DSS.