This all makes sense but it’s a secondary explanation about ‘economic impact’. The hardest part of economic impact right now is the painful bottleneck of juggling all necessary context, tools, and goal specs for your problem. It’s way too hard to assign problems to the models.
The models are already more than good enough for so many things I’d like them to do for me. I just don’t have the time or interest in sitting down and painstakingly collecting every piece of information they need to get a complete view of what I’m about to ask them to do.