Legal AI Workflow Split
GC AI
This split is really a fight over default workflow, not just model quality. Law-firm platforms are strongest when the job starts with legal research and polished memo style output, while in-house-native tools win when the job is a fast business decision inside Word, Slack, or an intake queue. GC AI is built around that second loop, where a legal team needs clause redlines, fallback positions, and stakeholder ready answers without turning every request into outside counsel style work.
-
Harvey and Legora come from the law-firm side, with much larger scale, capital, and agent libraries. That gives them enterprise sales muscle and broad legal workflows, but their center of gravity is still the firm workflow, where research depth and long form work product matter more than quick internal routing and business translation.
-
The in-house-native set is more fragmented. Spellbook and LegalOn are closest on contract review inside Word, with LegalOn especially strong for buyers who want dozens of prebuilt playbooks on day one. Wordsmith is attacking the front door instead, catching requests from Slack, email, and Jira before a lawyer even opens the matter.
-
Incumbents like Ironclad and platform vendors like Anthropic compete from workflow ownership and distribution. Ironclad already sits inside many legal departments as the contract system of record, and Anthropic is shipping legal specific plugins into a general enterprise AI surface that many companies have already approved.
The next phase is likely to reward vendors that own an entire in-house operating loop, from intake, to review, to policy memory, to system of record. As legal AI buying shifts from experimentation to consolidation, the strongest products will be the ones that save a lawyer a real handoff, not just generate a better paragraph.