Run a portfolio of experiments.
Unlike an actual investment, you can increase sizing *after* the results are in at a similar entry price. Do that ruthlessly.
Suppose hypothetically you have 100 units of mental bandwidth and run 5 experiments with 20 units each.
3 work, 1 has the seeds of greatness.
a) Be told by your manager "Ugh, you've got a 40% failure rate." Many businesses are managed this way. Prefer not to work in them if you work in this fashion.
b) Run 5 new experiments next quarter with 20 units each.
c) Put ~60 units in scaling adoption of #1 result