In short, greater revenue potential, greater accountability and greater risk of expense overruns:
- Greater revenue potential: These AI investments will likely focus on driving product usage as well as tracking and predicting user behavior for both new and existing products. Better products mean more usage, more usage typically means more Ad revenue for advertising revenue-driven businesses.
- Greater accountability: Putting AI development in the business units (BUs), means more accountability as AI/ML/NLP investments will be focused to a greater degree on revenue-generating products and services. These efforts can be tracked vs. budget which creates accountability and a performance-driven component to the AI effort.
- Greater risk of expense overruns: Building AI/ML/NLP models that power products and services is an expensive effort. Models need to be trained, they require enormous volumes of data and man hours to be properly trained. This is a continuous process rather than discrete. Models are continuously refined. My guess is that META’s BUs will want to initially ramp-up AI/ML/NLP investments to better align them with each BU’s product roadmap. BUs will likely share the AI/ML/NLP development effort in areas where model development is not product-specific.
Here is the original META AI announcement: https://ai.facebook.com/blog/building-with-ai-across-all-of-meta/