If you mean to actually weld one off parts-- is the sensing technology in place to accomplish that? The speed and feed you go at depends on the ambient conditions and part geometry. Even if so, you have the problem that much welding is and must be done in-situ because the assembly is not transportable. In some cases you could bring in a robot arm, but even where you can that seems like lot of effort vs just sending a welder in.
This is partly because, ultimately, it is more cost efficient to produce multiple parts using similar processes on the same equipment than to produce all parts on dedicated equipment using a broader mix of processes. It is also partly because outsourcing fabrication makes sense for many people, which means the fabricator themselves is receiving a range of orders that has to be executed in rapid succession on a given equipment line.
While this transition presents a challenge for everyone (design, management, HR, etc.) it is also a huge opportunity, particularly for we software people.
Often really substantial experience is required to approach welding problems, including interpreting order intent (quantity and material related concerns such as jig and fixture design, adjacent process speed and cycle time affinities, likelihood of follow-on orders, etc.), interpreting drawings (often made by non-welders and potentially requiring feedback) plus understanding finished workpiece intent and loading, intermediate work holding and handling, welding process selection, shield gas and flow rate selection, manual dexterity and precision, bead material selection, optimal power levels, temperature monitoring, feed rates, environmental safety and pollution concerns, questions of finish including scratch prevention, grinding and polishing, cleaning, and potentially subsequent surface treatment, inspection and testing, etc.
If anyone wants to work on automation for these sorts of fabrication related problems, we will be hiring in a few months.