"Software Is Eating the World, but AI Is Going to Eat Software," as NVIDIA's CEO described in 2017. Yet not every product problem is a good fit for AI. As a Product Manager, focus on the user problem and business outcome. AI/ML is a powerful tool at your disposal.
Many jump on the bandwagon without assessing if their problem is the right one. Use this checklist:
1. The problem requires making a lot of decisions
It takes a couple of editors to sort articles for a printed paper. But it's infeasible for Twitter to employ editors to sort its feed per person. Examples: recommendations (people who bought this also bought...), pattern recognition (spam detection, face detection).
2. Each decision is low consequence
If we get it wrong once in a while, no one will die. ML isn't perfect. It can't drive a car on a snowy day. High stakes usually still involve humans.
3. Humans can agree on the "right" decision
Either there's ground truth (this email is spam) or we agree on the outcome (watching this video makes the user more likely to return).
4. Your organization has access to a lot of relevant data
Options: collect data in the product (video completion on YouTube), human raters (Google Search, TikTok Integrity), buy data from other companies. Without data, your models are worthless.
5. Your organization is committed
At Google, we developed a model to determine landmark importance for tourists. World-class team, advanced AI, troves of data. Still took a year to get a result we'd put in front of users, and another year to be proud of it. Plus you need to retrain models, keep data fresh, and manage technical debt.