Nothing create more jobs than it destroys. Or almost nothing. At least as far as B2B service (as AI is mainly) is concerned.
For a simple reason: a new technology is adopted if it cost less than the old one. And cost less because it destroys more job than it creates.
Did ATM created more jobs than it destroyed? Of course not. Surely it created jobs in ATM manufacturers. But it destroyed more bank tellers jobs. Otherwise, no bank would have adopted ATM.
That is how it works. That is how progress works.
Is that a reason to forbid ATM and force banks to hire army of bank tellers to give the customer the same capacity to get cash 7/7, 24/24? Of course not.
Because, surely, it destroyed jobs, but it also cost less money to customers. And what those customers will do with this money? Spend it on other things. So indirectly, it created jobs that are invisible. You will never hear a farmer say "I've hired people thanks to ATM invention". Because there is no way he can know that. Nevertheless, it is probably true; people have more money so they can spend more on food.
Same goes for AI. Anywhere AI will be adopted, that will be because it destroys more jobs that it creates. For example, people from call centers should be worried. I've worked for a company that replace those people with AI. Not for the whole conversation, sure. But for the first part: we are now able to make conversational agent able to ask "what can I do for you", listen for the answer, as complicated it is, and understand correctly which service the client should be redirected to. And after all, real human doing that part of the job now are already just executing algorithm like a robot, so it is not surprising if they can be replaced by real robots.
But if it cost less to have real (underpaid, probably oversee) people to do the job, AI will never be used in that domain. But it will cost less to use AI (even if that mean handling AI errors, maintaining the computers that host the AI, etc). Because it will destroy jobs.
I know that is harsh for those people who will lost their jobs (I've lost my job myself several times in my life). But, globally, that loss is not a bad news. Because it will indirectly create jobs not only visibly in the AI domain, but also, invisibly, in all domains.
People are richer today than they were in my parents generation
The progress of the wealth is not very fairly distributed, nevertheless, even the poor are less poor than in my parents generation. In my parents generation, in my country, the poor were starving. Now, mainly, they don't. Their problem is more housing. In my parents generation, the poor could not have decent health care. Now (in my country; if you are American, still not in yours) they have.
Why is that possible? Because things cost less. Because you need less workhour to build things, and because people have more money to spend on those things, so the rise of consumption diminished the cost.
We have (my country as much as America) lot of progress to make so than everybody benefit of this progress, and so that nobody is left behind. But nevertheless, it should never be a bad news when less work is needed than before.
Each time something that should be obviously a good news is said to be a bad news, that is just by ignorance. Because the direct consequence seems bad, and the indirect consequence impossible to precisely enumerate.
It is exactly (but from the other side) like an economist would say that a war or a disaster is a good thing for economy. That is never true, obviously. But the direct consequences (economics) seems good: more production, more GDP, so economics growth. And the indirect consequences (a generation of less educated people, effort on reconstruction that could have been used on progress we will never know, etc) impossible to foresee.
When humanity learn to do automatically what needed lot of works before, that is a good news for humanity. That is more work power we can devote for something else. Even if economist are able to enumerate bad consequences and unable to enumerate good consequences, good consequences are always more than bad.