Zero-shot agents can come very handy when we want to adapt our models for something specific by defining different tools.

By using chain of thoughts approach, model will know what information it wants and can automatially call for it using different tools which are defined.

This approach can even be better especially with time series llm's