How should firms protect themselves against the risk of being over-reliant on a single AI provider like Claude/Anthropic changing its pricing or product?

A member raised concern about putting 'all your eggs in one basket' with Claude, since the provider could change pricing or pull features at any time, potentially making time invested in building skills and workflows feel pointless. The response was that the effort is 'still worth it' regardless. One participant agreed on the need to be model agnostic: token pricing is likely still subsidized at a loss currently, so firms should extract as much value as possible now while building things that don't depend permanently on Claude. Skills were pointed out as essentially just markdown/text files, meaning they can be moved to any other LLM in the future. That firm is also starting to focus more on running LLMs locally rather than relying solely on cloud-based LLMs, specifically to stay model agnostic and avoid being exposed to token price increases or a single vendor's decisions.