Commoditization of Patent Drafting

The real threat is not that general AI can do patents perfectly, it is that it can make first pass drafting feel cheap and interchangeable. Solve Intelligence wins today by turning invention notes, call transcripts, office actions, and technical documents into patent specific work product inside one editor, but horizontal platforms are improving fast on the raw reasoning and citation layer that powers those first drafts, which squeezes the premium on basic generation.

• Harvey is the clearest proof point. It grew to $195M ARR in 2025, moved beyond a custom legal model, and now leans on the best available frontier models plus citation backed LexisNexis content. That means specialized vendors are no longer just competing with niche tools, they are competing with large distribution and research stacks.
  2 sacra 3 sacra 5 harvey
• In practice, basic drafting gets commoditized when the user workflow starts with a general prompt, paste in source materials, get a coherent draft with citations, then edit. Solve Intelligence still has an edge in patent specific controls, like jurisdiction styles, office action parsing, chemical sequences, drawings, and claim chart generation, which are harder for horizontal products to copy quickly.
  1 sacra 2 sacra
• This is why the market is fragmenting around workflow ownership rather than pure model quality. Legal AI leaders are shifting toward document systems, search, integrations, and repeatable task flows, because frontier models have already erased much of the advantage once claimed by fine tuned vertical models.
  3 sacra 4 lexisnexis 5 harvey

The next layer of competition will center on who owns the full patent workflow after the first draft. General AI platforms will keep absorbing generic drafting, while the durable winners in IP will be the products that capture attorney edits, enforce patent specific structure, connect to prosecution and litigation tasks, and become the place teams actually finish the work.