It really depends on what your objectives are. For design optimization, the currently available automated optimization methods (Adjoint and other topology methods) are not as good as a well experienced Aerodynamicist. There has been a lot of effort in improving these tools and creating more complex design optimization tools (such as neural net optimizers), however these tools do not "understand" the fluid flow and simply try and chase whatever metric is being targeted. A good aerodynamicist will be able to make informed decisions based on their interpretation of results in order to iteratively improve the design.
Simulating a large number of different designs and trying to compare the results can be an inefficient use of resources. This method can be useful when comparing a range of options on a similar concept (think simulating 5 options of winglet height to find the optimal). If changing multiple things, it is impossible to work out which change was the most beneficial - the general best practice is to just change one thing at a time.
Also, running simulations across a comprehensive range of roll, pitch, yaw & wing configurations is extremely expensive. A good Aerodynamicist will be able to optimize using simulations at one or a small number of operating conditions and be able to understand the impact of their changes across the entire operating envelope. An automated optimization method would have to simulate many more operating conditions in order to properly evaluate and compare design iterations.
Once a design starts to become more mature, simulating more operating conditions is needed to understand performance across the whole envelope. If the design performs poorly at certain conditions, or if any of the performance characteristics transition too abruptly, the design can then be improved to minimize this unwanted behaviour.
Understanding the fluid flow, and knowing how to optimize the design to meet requirements are the hardest skills to develop. This is what sets good and great Aerodynamicists apart.