The problem is currently all the "competing" ideas have a ton of tradeoffs and is rarely one size fits all. Furthermore, it's not clear if the idea you choose w

The problem is currently all the "competing" ideas have a ton of tradeoffs and is rarely one size fits all. Furthermore, it's not clear if the idea you choose will be become obsolete by the underlying model architecture getting better. On top of all that, you are essentially competing with Anthropic/OpenAI/Google, where your only advantage is "privacy". Anyone who deeply cares about privacy, and is willing to pay for it, may just likely do it on their own (especially if your CTO is pouting money into "investing" in AI). Anyone who doesn't will likely not want to 6-7 months behind what you can get at OpenAI or Google.