Should firms/builders focus on AI-native tools or use AI as a means to build traditional software, given the risk that AI/token costs could rise sharply?

One builder described deliberately building their first product version with no AI in it at all, out of concern that token costs could rise dramatically and make AI-dependent tools too expensive to rely on. Their philosophy: use AI-assisted development (e.g., Claude Code for frontend work) to build durable software, and treat AI subscriptions like Claude Max as a resource to extract maximum value from now, while asking 'if AI went away tomorrow, what tools would still work?' Another participant raised the concern that if firms train staff heavily on AI-driven accounting workflows and then AI availability or affordability changes, that would create a rough transition, especially if the firm hasn't built up staff who understand the fundamentals independent of AI.