To start branching out into additional software. Rather than shoehorning a feature into a product that doesn't need it, make that feature be a standalone product.
This was the premise behind various component platforms, e.g. OLE, CORBA, OpenDoc which were, in many ways, the GUI equivalent of command line shells.
We could see a renaissance in that area if technologies similar to Electron and ubiquitous app stores make these more feasible.
(To toot my own horn, I think https://tenet-lang.org/ would be helpful in building well versioned APIs to allow intercommunication with shared semantics.)
It's a bit late for this thread, but Android Intents and the related WebIntents were kind of similar to OLE, but it seems neither captured the attention of the market.
