If there's one thing we learned from this morning's presentation from Zillow's Olly Downs, it's that artificial intelligence is like a rooster. Yes, a cock-a-doodle-do rooster. Downs, the senior director of consumer analytics, grew up on a farm in southeast England, so he knows from what he calls cockerels. And one thing he knows is that when a rooster wants to attack you, it is "relentless." As is AI.

"AI will relentlessly pursue whatever metric you've set up for it," he said, so "be careful what you wish for." The most important thing to get right about AI is defining the success metric. AI approaches work, but the impact can be unexpected. The impact is directly related to the success metrics you choose to optimize. Think about that success metric. Then, align customer experience and business metrics.

Zillow is in the middle of a three-year journey to change its business model, from simply listing available properties to one in which it takes part in the process of buying, selling and renting, and has developed a home loans department. It has four foundations to that end: 

1. Rich user understanding informed by user identity (sense/think)

2, Massively scalable measurement, experimentation and optimization (think)

3. The Right Metric (AI), expected value (feedback)

4. Cross-channel integration (act)

The foundations are innate in machine learning, which starts off being a static process. Sensing, Downs said, is the training of the ML model. Action is how you curate action on the signal generated. When it becomes AI, he said, it is the like the analogy between search and results.

Results are what you get using machine learning. AI is when you have a conversation, and you have to sense, think and act. Based on the effect your interaction has, you go around in a feedback loop. It is part of an ongoing success, with continual improvement. It is, at bottom, an adaptive process.