That's the issue indeed. The entry threshold and the amount of minimum needed features for agtech products are very high. Moreover, the seasonality of ag business makes the situation for agtech startups even more difficult (you can usually sign new test clients only in between ag seasons)
This is very needed for the industry.
I'm a co-founder of agtech startup https://geopard.tech, we act as a platform and infrastructure for ag businesses (provide analytics and APIs) in the precision agriculture niche. When we created the company, it was the idea - to support agtech companies to launch their software faster/cheaper (our engine analyses yield, soil, topography, satellite, ground sensors, drone data and provides analytics on top of it). Before GeoPard we had another agtech company acquired by ag giant Bayer in 2015, then inside Bayer, we built Xarvio digital farming system. So, this is very needed, I know what I talk about. It takes usually minimum few years to launch solid agtech product.
The biggest issue I see right now is where to get valid data. Model is nothing without a huge amount of validating datasets, which only have ag giants. They will not share the data so easy since all of them build their digital platforms and understand the value of data.