But as you said writing your own is also error prone.
I'm pretty sure that developing your own has much, much higher risk than using some pre-existing solution.
If I understand correctly, these are a way to automate the execution of the financial part of normal contracts without the need for undue trust between parties. Some simple examples that might help (and I invite correction):
• I and a few investors agree to buy into an enterprise in several installments. A contract can be written to the blockchain that automates these payments. Should a majority of the investors decide to withdraw support, the payments will cease, but while more than 50% wish to continue support, everyone will continue.
• I have an enterprise that I am highly confident will generate profit. I can establish a smart contact that guarantees my investors a minimum ROI at specific intervals; should the balance fail to grow by the agreed margin, all investors will instead be refunded (in full or part) based on a scheme agreed a priori to be fair — say, a function of shares owned for how long.
• I and several other Ethereum miners agree to develop a scheme built on top of Ethereum for, say, a specific method of interoperability between the Ethereum blockchain and traditional futures markets. We have a specific idea for this, but agree we need to ensure that there is momentum in the project, and want those participating to have 'skin in the game'. We build a tontine such that all of us invest a non-trivial amount, and any investor becomes ineligible for payout if their hash power abandons the protocol. If this falls below a certain threshold, remaining investors can vote to continue or divide the pot and abandon the project.
It is possible to create a smart contract, like the DAO, which implements the rules of a joint stock company. But stockholders are not concerned about the risk that these rules will be broken. The risk is that the company fails and there is nothing to distribute according to the rules. Replacing the general meeting of stockholders with a smart contract doesn't change the economics of collective investment – it just adds the risk of losing everything to a bug in the contract.
without the need for undue trust between parties
Because the program author has no admin privileges, users don't have to worry about the service provider abusing their admin privilages to secretly modify the database (which may otherwise be a concern for high-value applications such as, e.g. vote-counting, e.g. a database that stores bank account balances, etc).
The disadvantages are (1) it is much more expensive compared to other cloud computing services, (2) since the uploader doesn't have admin access, they can't fix bugs, and since everything is transparent, it's hard to store secrets.