One underexplored reason: companies can't give AI agents real authority. The moment an agent needs to do anything beyond summarizing text — update a CRM, transfer funds, modify infrastructure — the security question kills it. No one wants an agent that can take irreversible actions with no approval chain. Until the trust architecture problem is solved, AI stays in read-only mode for most enterprises.
This is the biggest bottleneck. To realize the “replacement of white collar workers” fever dream, (which is, I still believe, technically feasible), you need the agent that replaces them to have all of the context they had. All of the emails, all of the Slacks, all of the meeting minutes, access to private corporate systems and files, etc. I can’t think of a single company that would want to turn all of that over to OpenAI.
