Technical
Zero-Shot Learning
An AI model's ability to perform a task it wasn't explicitly trained on, using only a natural language description.
Zero-shot learning refers to an AI model's ability to perform tasks it wasn't specifically trained for, based solely on natural language instructions. For example, a model trained on general text can classify customer feedback into categories it's never seen before, simply by being told what the categories are.
This capability is what makes modern LLMs so versatile — and so rapidly adopted. Employees can use general-purpose AI tools for highly specific tasks without any customization, which accelerates adoption but also means AI usage patterns are unpredictable and hard to anticipate in advance.
Why it matters
Zero-shot capability is why AI tool adoption is so fast and so hard to predict. Employees don't need specialized tools for specialized tasks — they can use ChatGPT for everything from writing code to analyzing financials. This makes comprehensive AI discovery essential.