I think it was the 99% Invisible podcast that did an interesting report about whole networks of people who were paid in Mexico to spread election messages; and these networks are often reused by gangs or other organizations to promote stuff or harass opponents. So in this case, they're not bots. They're actual humans, because it's cheaper to do that than pay people to write scripts.
After Andrea is attacked by a stranger in Mexico City, she just wants to figure out who the guy was. Investigating this question drops her right into the middle of one of Mexico’s biggest conspiracies.
Additionally, you can train a human to become better at astroturfing over time, and train your network as a whole to be resilient to partial takedowns.
These things are much easier to orchestrate than bots, especially if your specialization is not CS but scamming / manipulating opinions, and in many cases, especially with scale, human labor can actually be more expensive.
Of course, subsidizing the labor to impoverished countries definitely helps; these networks tend to lack the sophistication used in more sensitive political or industrial topics.
Sometimes for national or global campaigns you'll find a mix of both cheap labor, a small team of experienced astroturfers with a good grip on their persona acquisition and management, and bot networks, each targeting a different set of demographics. If your target demo is sufficiently ignorant or radicalized in their beliefs, it doesn't take much to convince them and bots will do just fine.
