You <- Manager/PMs <- VP <- SVP <- Bezos or C-level.
Usually a manager runs a 2-10 person team who they advocate for, and a few PMs interact with the teams that work on the projects they oversee. The PMs are sort of like sales in a traditional company, customer-facing roles who prioritize and occasionally promise features.
There are usually enough VPs that you can meet with one if you have an idea that your peers like, and the support of a VP seems to be enough to let a small team work on a tangent for a year or two.
It sounds a lot like how Google develops new products, because it's surprisingly easy to turn a working proof-of-concept into a public-facing product or service. So why don't they suffer from the product-killing disease that plagues Google?
I don't know, but I have a couple ideas:
* Amazon throws people and money at things that work. Products like Lambda and Alexa used to be small upstarts with modest expectations, but when they sold well, the company invested heavily in them.
* Amazon promotes people based on peer feedback and manager recommendations. It's an easy system to game, and there's a bit of graft as teams defend likeable underperformers since they stopped leaning so heavily on "rank-and-yank" performance reviews. But it also means that you can get promoted for maintenance, like "X prepared our retail service for holiday traffic, and we didn't lose millions in sales when it didn't crash."
I'm not sure how applicable their management structure is to most organizations, though. They can tolerate massive loss-leaders because they have a couple of money printers, and they re-invest most of that money in the company rather than sitting on it or paying it out in dividends. People tend to like working there if they can cope with a bit of stress, and I saw as many people leave the company as I saw take internal transfers during my time there, so "tribal knowledge" also fades more slowly despite the churn.