The DIY route teaches you a lot but getting everything to work together reliably takes longer than most people expect. Speech recognition, intent handling, API connections, wake word detection, it all adds up. Came across the
brawe ai app builder recently which builds out the product for you and also handles the marketing, sales and support side while you keep the code and revenue. That combination removes a huge chunk of the usual overhead. If the goal is a working product more than just understanding the stack, worth a look before going fully custom