Progress in linear programming and integer programming solvers and software tools over the last ten years or so has provided unprecedented power at the fingertips of practictioners tackling industry problems. However recent experience with industry problems highlights the challenges that remain: it does not take much for an industry problem to push the boundaries of what is solvable with current technology, particularly if one wishes to manage uncertainty in the data. This talk will discuss cases of industry problems, highlight the challenges they present, and investigate the solution approaches used. Included will be recent work on the use of stochastic programming for mine production scheduling in the presence of uncertainty in the geological estimates. This represents a particularly difficult type of stochastic programming in which information learned about the uncertainty depends on decisions made. However special mathematical structure in the problem allows progress to be made.