The short answer is that there are a small number of common use cases, but not for lack of ability. One of the most common use cases is with developers using an agent to write code and execute tasks for a program. Many developers using agents are buying web-based credits for things such as API calls, AWS credits, or Azure compute credits. If you're a developer and will need to pay for these things anyway, it makes sense to have an agent that's already making the API call or using the credits to have the ability to replenish these things as needed.
These all make sense from the perspective of a developer who constantly uses and needs to repurchase things like these. Now what about for the rest of the general population?
What seems to be a relatively simple question can only be bounded by the user's imagination. People have used AI agents to buy pizzas, cleaning supplies, back to school gear for their kids, necessities for an upcoming vacation, the list goes on and on. Though they are new, agents are being explored for tons of non-technical applications and will certainly see usage in the future both for companies and for regular people.
Imagine you're a marketing professional. Wouldn't it be effortless for an agent to compile a list of promotional materials, receive your approval and then purchase them by itself? Or put yourself in the shoes of a grocer managing the store's inventory. As long as your agent has data, it can save plenty of hours of surveying and taking stock of how much milk, produce, and other foodstuffs are in stock. It can take over that burden of queueing and executing orders after planning for the demand of certain seasons and leave you, the manager, with more time to tackle the other tasks of the business.
Even on the individual level, someone with a busy life may want an agent to restock things like their paper towels and trash bags. Sure, they can have recurring purchases setup, but why don't they just avoid the possibility of overstocking altogether? Having an agent ask you periodically if you are low and need more will eventually train it enough where it doesn't have to ask anymore, it just knows when and when not to buy.
Use cases are fast approaching, and it's important that the security on this new frontier is emphasized. We at Authoryze took that into account with our platform, giving agents selective access to funds to protect the user they are acting on behalf of. At the end of the day, an agent doesn't bear the repercussions of a bad decision, a human does. It is imperative for the human to have the proper controls on their agent if they want it buying things on their behalf, and Authoryze provides those controls.