Improved methods for providing recoverable workflow of complex and/or risk-prone workflows in a computer system or computer network include segmenting such workflows into sub-processes, associating a checkpoint with the end of each sub-process, and determining whether processing of each sub-process was successful. If so, a set of parameters is stored at each checkpoint before processing the next sub-process, but otherwise processing of the failed sub-process is retried, using the parameter set stored at the previous checkpoint. In one embodiment, the methods are implemented by a certified messaging capability that guarantees delivery of any message. In the event a hardware or software failure occurs during a sub-process execution, the certified messaging capability returns a failure indication, so that execution of the failed sub-process can be reinitiated. Improved computers and computer networks for providing recoverable workflow are also described.