A Little Stack History...
Science Fiction has been dealing with the idea of robots taking over the world for a very long time. Based on a question I asked on our Sister Stack, Science Fiction & Fantasy, humanity has been toying with the idea since 1909 when E.M. Forster wrote The Machine Stops.
And it doesn't help that you've asked a question that's very common on this Stack. When presented with the proverbial godlike character (androids with no inferior comparison to humans and at least one superior comparison qualify...) and asked, "how do I kill this character?" (asking how to stop them from taking over the workforce is not particularly different), there is only one answer:
Introduce an imperfection.
You'd be surprised how often authors don't want to introduce an imperfection. I guess it's a bit like being offended because someone thought your baby was ugly. Often, they want someone to think of a way to circumvent their perfection. Which is, by definition, impossible (thank goodness you're not doing that...). So long as the character humanity is compared to is in no way inferior and is in at least one way superior, the character will always win (unless dumb luck is introduced, like forgetting the cliff edge was right over ther...).
So let's introduce some imperfections.
BTW, all of these and a fair amount more could be found by researching the history of androids in science fiction. Yup, a bit of work, but you're actually expected to do some research before asking here.
Imperfection: intellectually inferior
Define a reality in your world, in some way no A.I. can be wholly superior to humanity. Maybe your A.I. lacks creativity, making it a terrible artist, musician, or even problem solver. Maybe it lacks psychological insight, making it a lousy counselor, pastor, or even friend. Manning the suicide hotline it would not do. Computers are naturally good at mathematics and comparisons... but not judgement. This was a core premise in Will Smith's I, Robot.
Imperfection: economically inferior
Occasionally addressed in science fiction is the idea that humans are honking cheap. Oh, it may take a while to train one to some useful purpose, but when all is said and done, they're cheap. The average cost of raising a child in the U.S. today is about $300,000.
To make our comparison, let's include the U.S. Army's estimate of \$50,000–\$7$,000 to train a soldier. Now let's compare this to the average cost of a tomahawk missile (\$1.87 Million) or a B-21 stealth bomber (\$737 Million. Why won't androids take over the workforce? Because it won't make economic sense for a \$2.5 Million dollar android to flip burgers (just to make a point...).
A corollary of that idea is that humans are insanely cheap to repair, even though they're less durable than androids. Androids may be really hard to damage... but heaven help the insurance company that's expected to foot the repair bill. This means you might not find androids involved with hard rock mining, either.
Imperfection: limited agility
Sure, your android may have the ability to compute to fifteen digits exactly how much force and motion will be required to dodge the oncoming bullet. But does it have the time to actually move out of the way?1 To be quite honest, robotics today has come a long way, and maybe someday robotics will completely mimic the human body. But I wonder. The human body is breathtaking in its grace and ability when properly trained. It may simply be true that an android will never make a good lumberjack.
Imperfection: they can't replicate themselves
I believe it was an episode of Star Trek: Voyager that introduced the idea that the androids could completely replicate themselves... except for their power source. As a means of guaranteeing they couldn't just take over society, they were denied information about how to replicate the power source. Want to guarantee they can't take over the workforce (or the world)? Make sure there's some critical part of themselves that they absolutely cannot and do not know. Make sure they're programmed and wired to reject the knowledge so they can't use a Tor node to perform a dark-web data search. If you need another android, you gotta have one or more humans involved somewhere.
Imperfection: inferior lifespan
An inferior lifespan might be a subset of economically inferior, but let's run with ut because it'll take us to a rabbit hole that we need to enter.
In Ridley Scott's 1982 blockbuster Bladerunner, audiences were introduced (not for the first time) to the idea of an entirely biological, entirely human android. One that was economically practical for use in hazardous situations. How did they keep said creature from taking over the workforce? They programmed in a limited lifespan. For the most part, this kept the replicants from filling every job — they just don't last long enough for it to make sense.
There's an INCREDIBLY IMPORTANT point to make here
But that didn't keep the replicants, let's call them "biological androids," from wanting to be human.
And if you're thinking "Pinocchio!" you're not far from the truth. Every android story is inevitably a version of Pinocchio — because if it's not, then your question doesn't make any sense. Give your android equal-or-better human attributes and one of the first questions they'll ask is, "why am I not treated equal or better than any human?"
Or, more precisely, the moral of the story. In many ways a question like this is Too Story-Based, but they tend (when written well) to fall into our "a finite list of things question is on-topic" policy. But it's very much worth remembering (for your future questions) that you are expected to avoid asking us to help tell your story. You included a lot of data that could help us — but did not explain any of your story goals or purposes. Frankly, a lot (most) of your question could be deleted without affecting the question at all. Think of it this way... what reasons would stop your androids from taking over the world? Well... that depends on your story....
Remember that you need to remove all story dependencies before you ask a question or risk having it closed as Too Story-Based. Tell us where you want to go and we're very good at getting you there. The more you expect us to tell you where to go, the quicker your question will be closed (see the Help Center for more).
1 OK, OK... a human couldn't dodge the bullet, either. But I hope you get my point. Robotics today still can't replicate every movement that a human can do... but they're getting closer. And yet it's not enough to replicate the movements. To do so in a timely manner, in an effective manner, based on the "free will and choice" of the robot. Go watch some YouTube videos of dancer Ann Reinking. If you still disagree with me then you have no eye for what that lady could do. Land'o'goshen.... We're kinda a long way away from that. Let's say it can't be done. That's a pretty good imperfection.