I share the general skepticism, but I am open to the idea that (much like other electronic commerce), the smart contracts might make 99% of transactions much more efficient; so the expensive manual interventions are only used on a small proportion.
Of course, it is not necessarily obvious that blockchain-powered smart contracts are the only way — or the best way — to achieve that automation and efficiency!
> the smart contracts might make 99% of transactions much more efficient
The entire point of proof-of-work is that it is inefficient, by design. Efficiency is counteracted: Advances in technology that lead to more efficient mining techniques directly mean the difficulty for the next block being adjusted up.
Operating on traditional databases (where automation is regularly implemented, too) is orders of magnitude more efficient. Because like most other things other than proof-of-work, those processes directly benefit from getting more efficient.
PoW wasn't designed to be more efficient, the Nakamoto consensus type implementation exists to improve honesty and security. There is no traditional database that's as secure as a blockchain, it's always a tradeoff between security/immutability and efficiency. But since efficiency does matter many 2nd generation chains have moved away from PoW.
Of course, it is not necessarily obvious that blockchain-powered smart contracts are the only way — or the best way — to achieve that automation and efficiency!