IBM's neural network explainer and the simplest analogy you have ever heard for how they work

Neural networks are the concept that stops a lot of people from understanding AI at any depth. They sound technical and the name suggests a level of biological fidelity that is not really there. IBM's explainer https://www.ibm.com/think/topics/neural-networks covers the actual mechanics in plain language: connected layers of nodes, weights that get adjusted through training, and the patterns that emerge from many examples.

The key mental model shift: a neural network does not have an explicit rule for any decision. It has a vast web of weighted connections that were calibrated through exposure to labelled examples. The decision emerges from the activation pattern rather than from programmed logic. That is alien to the way most software has historically worked.

The applications section covering image recognition, natural language processing, fraud detection, medical diagnosis, recommendation systems, and weather prediction shows the breadth of what the same underlying architecture can do when trained on different data for different tasks.

The invitation here is the one in the article angle: what is the simplest analogy you have ever heard for explaining neural networks to someone who has never thought about AI? The brain comparison is technically inaccurate but intuitively useful. What else works?