Why Do AI Wrappers Work in the Enterprise?

Aaron Levie says a model wrapper is working out in the enterprise because organizations need a bridge from a model's capability to the workflow the enterprise already runs.

Two years ago, treating application companies as the hottest opportunity would not have meant much. The pattern is now playing out in the market, mostly because of open source. People underappreciated the gap. Probably a trillion dollars has been bet on two outcomes: the bridge stays limited, or the bridge is vast and has to take on a lot of depth within organizations. The bet is whether you are only long the model itself and super intelligence, or long the application tier, which previously would have been a pure Neo Lab. A large gap remains between the model and the workflow. Bridging the gap leads, over time, to also doing the model. Enough data and domain expertise then become their own flywheel.

The case for the application tier is one he is very bullish on. A startup can be the applied company itself, which is the Neo Lab, or the infrastructure provider to the Neo Lab. Multiple layers of opportunity are open for attacking the space, and the shift is a huge update for the market.