Maintaining Solve's workflow edge

This risk is really about whether Solve owns a workflow edge or just rents model intelligence. Patent drafting is not one prompt, it is a chain of tasks, turning inventor notes into claims, matching a firm's wording habits, parsing office actions, inserting citations, and handling drawings and sequences. Solve keeps its pricing power only if that full edited output is consistently faster, more accurate, and easier to approve than what a strong general model can produce on its own.

• The clearest warning sign comes from adjacent legal AI. Harvey abandoned its own fine-tuned legal model as frontier reasoning models improved, then shifted toward workflow orchestration, external content integrations, and custom firm workflows. That shows domain tuning alone can lose value quickly as base models get better.
  5 sacra 6 sacra
• Solve has more room to defend itself than a generic legal copilot because patent work is unusually structured. Users pick jurisdiction and style profiles, upload USPTO PDFs, inspect citations, and work with technical drawings, chemical structures, and biological sequences. That kind of post-processing is closer to product logic than raw model output.
  1 sacra 2 sacra
• Incumbents are moving the same direction. Clarivate bought Rowan Patents in July 2024 to fold drafting into a broader IP stack, and LexisNexis partnered with Harvey in June 2025 to combine authoritative legal content with AI workflows. When data owners add drafting, the standalone vendor must win on measurable output quality, not access alone.
  2 sacra 3 clarivate 7 lexisnexis 8 harvey

The next phase favors companies that turn every user edit, approval, and citation check into better workflow rules. If Solve keeps compounding firm specific styles, prosecution patterns, and multimodal patent handling into a tighter review loop, it can stay ahead even as raw models improve. The durable moat shifts from model fine-tuning by itself to workflow performance embedded in daily patent practice.