I believe they are betting on being able to monetize things around their "Large Action Model".
Could be market for the automations, where people can for example sell "buy tickets on ticketmaster" actions.
They could also monetize all the data they have and will have training that particular model, which is very lucrative if the AI bubble continues to inflate.
Voice skill discoverability was a huge issue for Amazon's alexa, and I don't see how this is any different except it's more expensive for both hardware and processing.