GC AI dependent on platform vendors

This setup makes GC AI strongest where lawyers already work, but weakest where platform owners can copy the same workflow and bundle it into software the customer already pays for. Its best experience lives inside Microsoft Word, and its core intelligence depends on third party model providers. That means the interface where contracts get marked up, and the model layer that produces the answers, both sit on infrastructure GC AI does not control.

• GC AI is explicitly built around Word. Its own product materials describe an official Microsoft Word add in for drafting, redlining, comments, and playbooks inside .docx files. If Microsoft improves Copilot enough for contract compare, redlines, and clause edits in Word, it can attack the exact user moment where GC AI creates value.
  1 gc 2 gc 4 microsoft
• The model layer is also rented, not owned. GC AI routes work through major model vendors, while Anthropic is already packaging legal workflows across contract review, document search, NDA triage, and briefings. As base models get better, more of the legal magic moves down into the model vendor’s product instead of staying in GC AI’s prompt layer.
  5 sacra 6 sacra 7 anthropic
• This is a common pattern in vertical AI. Harvey faces similar pressure from OpenAI, Anthropic, and Google moving directly into legal, and even Harvey has had to shift emphasis from proprietary model claims toward workflow orchestration. The defensible part of the stack becomes customer specific playbooks, approvals, and habits, not raw reasoning alone.
  7 anthropic 8 sacra 9 sacra

The next phase is a race to own legal workflow memory before the platforms become good enough. GC AI is heading toward deeper playbooks, team standards, and in house counsel specific processes that are harder for Microsoft or model vendors to replicate with a generic assistant bundled into Word or a general purpose legal plugin.