There are a couple of use cases I have come across.
One is supply chain, using what is now called an NFT (the application predates that term). The case mentioned in the article is based on assuring that a physical item (a bottle of wine) is not a fake. This doesn't work, for the reason given. However a different supply chain application is avoiding real items of forbidden provenance being introduced into the supply chain - specifically conflict diamonds. Large diamonds are marked with a microscopic serial number, which addresses the problem of how to associate the physical item with a blockchain token. The company which first sells the diamond adds a record of a transfer to a new owner, associating this serial number. [Point of trust: you have to assume that De Beers et al. will not introduce a conflict diamond]. Every time the diamond is sold, the transaction is recorded. If someone tries to sell a diamond with a new serial number, it is clear that the origin was never recorded. If someone tries to duplicate a serial number, it is evident that they are not the owner. This was implemented for diamonds, and I have seen it in at least prototype form for other materials such as tropical hardwood.
I'm not sure if the second case was ever implemented, but it was receiving serious commercial attention. There is a type of disaster insurance which does not insure against damage, but against circumstances which can cause damage. So you might want to insure against a flood if you have a house next to a river. Rather than a loss adjuster checking damage and months later you get a cheque, an alternative is that the contract says if the water level is above level x, you get $y. Because this does not require any judgement, just an oracle giving the water level, it can disburse funds quickly. A smart contract is one way to handle this. It's one of the few cases where suitable machine-readable oracles often exist already. The advantage to the policy holder is quick access to fund when needed for emergency response. The insurer can in theory reduce operational cost, though most of the saving comes from the type of policy (no loss adjuster needed) rather than the mechanism.
One is supply chain, using what is now called an NFT (the application predates that term). The case mentioned in the article is based on assuring that a physical item (a bottle of wine) is not a fake. This doesn't work, for the reason given. However a different supply chain application is avoiding real items of forbidden provenance being introduced into the supply chain - specifically conflict diamonds. Large diamonds are marked with a microscopic serial number, which addresses the problem of how to associate the physical item with a blockchain token. The company which first sells the diamond adds a record of a transfer to a new owner, associating this serial number. [Point of trust: you have to assume that De Beers et al. will not introduce a conflict diamond]. Every time the diamond is sold, the transaction is recorded. If someone tries to sell a diamond with a new serial number, it is clear that the origin was never recorded. If someone tries to duplicate a serial number, it is evident that they are not the owner. This was implemented for diamonds, and I have seen it in at least prototype form for other materials such as tropical hardwood.
I'm not sure if the second case was ever implemented, but it was receiving serious commercial attention. There is a type of disaster insurance which does not insure against damage, but against circumstances which can cause damage. So you might want to insure against a flood if you have a house next to a river. Rather than a loss adjuster checking damage and months later you get a cheque, an alternative is that the contract says if the water level is above level x, you get $y. Because this does not require any judgement, just an oracle giving the water level, it can disburse funds quickly. A smart contract is one way to handle this. It's one of the few cases where suitable machine-readable oracles often exist already. The advantage to the policy holder is quick access to fund when needed for emergency response. The insurer can in theory reduce operational cost, though most of the saving comes from the type of policy (no loss adjuster needed) rather than the mechanism.