The efficiency gains of AI agents are plentiful—especially considering the unit economics—but also with significant 'trade-offs'. Previous advancements welcomed new jobs e.g. the advent of the telecoms enabled call centres, the rise of cloud computing welcomed IT support. Now both stand to be replaced. Unlike these examples, the rise of (semi-)autonomous agents is unlikely to create any new roles. Perhaps it's too early to tell.
That said, there could—and should—be systems in place given the rapid rate of change being witnessed. In theory, the rise of agents capable of executing repetitive tasks—which humans have dreaded doing for years—should be the wake up-call policy-wise. In other words: innovative policy approaches are necessary to address novel technological advancements. For decades, we've dreamt of a world where work is voluntary, or simply no longer required, due to technological innovations. What was once a mere dream seems increasingly possible from a technical standpoint, but without considering underlying incentives.
Governments could still manage multilateral incentives in a few different ways. A few come to mind: policies optimising carbon reduction in data centres, job guarantee schemes limiting layoffs YoY, etc. These aren't perfect, but are some ways to mitigate the impact of change and ensure these advancements remain sustainable whilst being unprecedented. Better late than never.
The choice for governments is not between innovation and regulation, but ensuring decisions regarding technology in the public and private spheres address societal needs at the forefront, rather than as an afterthought. It's wholly possible for AI to serve the common good, but this is ultimately a matter of (seemingly) conflicting incentives.