Current implementations of digital controllers assume that sensing, control and actuation are performed in a periodic fashion. In classic control schemes, where sensors and controllers are directly connected, periodicity does not provide particular drawbacks, but, in the case of wireless sensor networks, such a choice may be questionable. One of the driving constraints in the design of wireless sensor networks is represented by its energy efficiency, and it has been shown that the main cause of energy consumption is due to the radio activities of the sensor nodes. By using periodic implementations, the sensor nodes are enforced to keep on transmitting measurements to the controller even if it is not really needed, thus wasting energy. To cope with these problems, self-triggered control was recently introduced. This technique aims at reducing the conservativeness of periodic implementations providing an adaption of the inter-sampling intervals based on the current output of the system. Existing work on self-triggered control considers linear systems controlled by state feedback controllers under the assumption of small time-delays. In this paper the problem of designing a self-triggered control scheme that applies to first-order processes with large dead-times controlled by PI controllers is addressed. Moreover, the proposed self-triggered scheme is robust with respect to set-point changes and external disturbances, which are typical in process industry. The results are validated by simulations.