Honestly, I think you would be better off creating an expert ANN for how long individual tasks may take . There will be more data, more often; which will increase the specialization of the system.
There are just a lot of factors in a general project so that it can accurately predict without a huge data set.
If the expert system is successful, you can link the ANN manager from above, who takes the total project time from all tasks (as ANN expert predicts) and some other project factors (for example, total number of employees, time of year, budget, etc.) and trained on this data from previous projects; and predicted from this.
Too much information can be made for very noisy data and insensitive output in extreme cases, especially in this case when they may be insufficient training data.