Instead of simply trying to "innovate", a good first step is often to instead focus on increasing the speed at which you can get feedback on iterations and experiments.
- SpaceX did it with rockets.
- Machine Learning is doing it for Nuclear fusion.
- Anduril is doing it for defense.
Maybe innovation is just the lag result of increasing the speed of feedback loops.