AI Agent Muse Takes Over Everyday Life: A Self-Experiment with Bank Access

Notice: This article was created with AI.

What’s It About?

Eli Tan of the New York Times integrated an AI agent called Muse into his life over a period of two weeks. The system was given access to sensitive data such as his bank account and email inbox in order to take on administrative tasks independently. By his own account this was the most useful AI application he had tried so far.

Background & Context

The AI agent took on various administrative activities: it made calls to the dental insurer, organised expenses and arranged appointments. The system proved particularly helpful at tracking down duplicate subscriptions, which it cancelled and for which it requested refunds. Not everything went smoothly in the process, though: on some websites the automatic sign-in failed, and when buying tickets the agent needed support.

Use was restricted to a limited group of users, which curtailed the availability of certain functions. Muse also produced a personalised podcast, which Tan rated as not very inspiring. Despite the positive overall assessment, the self-experiment also showed moments in which the wish for human interaction came to the fore.

What Does This Mean?

AI agents need extensive access to personal data in order to work effectively – a considerable leap of faith is required.
Practical benefit shows itself above all with recurring administrative tasks such as subscription management and scheduling.
Technical limits still exist: not all websites can be operated automatically, some tasks require human intervention.
The balance between automation and the need for human interaction remains an open question in integrating such systems into everyday life.