Agentic identity management is happening at the frontier model a lot sooner than I expected. Smart! Anthropic is productizing their stack up the stack to guarantee vendor lock-in. But more important, Claude is being embedded inside the organization to learn the policies & workflows. I expect they’ll offer an internal Claude model for organizations that will be fine tuned inside the enterprise & serve as the learning loop with vendor lock-in in. Only that way will they be able to optimize their own service in a bundle that’s less token based.
