I'm not sure the numbers pencil out with our current state of automation implementation for say, small towns; pigging as pointed out elsewhere in the thread does this, but the equipment is pretty expensive, and you have to have sewer lines designed to take the pressure (especially when they run over aquifers) when some old infrastructures were not designed with that kind of pressure in mind along the joints as a regular occurrence. Current state of the art from what I can see is automation of this sort of work is more of "let's create a tool that does
X task or at best tasks in a narrow range", and less of creating a Moravec-class android, or
Mote In God's Eye-reminiscent Engineer/Watchmaker Motie-style creatures-but-as-bots.
Developing world humans in their near-innumerable billions will always be economically (though arguably not organizationally nor for innovation-wise) cheaper than narrow automation for much "blue collar" work like this for as long as we leave Moravec's Paradox on the table, and not an evolving/improving science and engineering solution in the market.
Once I get into equipment to automate manual work (especially any commercial grade kit), I quickly discover the cost hurdle to clear is quite high. I either need lots of people I want to free up for another part of my business, or use the automation as close to around the clock as possible, or both before the capex and opex fall below the financial benefits. The ongoing maintenance of the new tool is a part that usually is not accurately accounted for in many people's opex calculations. I suspect this is why well-capitalized large companies are leveraging automation benefits sooner and we're seeing increasingly more emergent oligopolies, and Moravec-class automation when it ever arrives might be very profitable to the suppliers and disruptive to these incumbents.