One of the key limitations is throughput. I can collect a sample but then I have to identify all of the pollen. If I want to detect a key (but rare) plant I might need to go through millions of grains just to find pollen from that species. What is the sample size I need to reliably estimate abundance of a species that makes up 0.000003% of the individuals (or biomass) in my area? Maybe AI could help analyze samples faster and more reliably.