The AI gap at the boutique IP bar
A quiet split is opening in legal AI. At one end of the market, Big Law is procuring agentic platforms. At the other, the boutique bar is handed a chatbot subscription and wished good luck. The gap is not about ambition. It is about what each product asks of the firm.
What Big Law is getting
The platforms of the Harvey and Legora class are not chat windows. They are wired into the firm’s documents and matters, they carry playbooks tuned to the firm’s standards, and they arrive with something just as important as the software: innovation teams and dedicated staff whose job is to make adoption happen. The platform meets the firm inside its own work, and someone is paid to keep it there.
What boutiques are offered
A boutique IP firm is offered none of that. It is offered a general-purpose chat window: a blank box that knows nothing about the firm’s matters, forgets each session, and leaves the entire adoption burden on the attorney. And a boutique has no innovation department to absorb that burden — the partner is the innovation department, fitting the experiment in between consults. The predictable result: the tool gets tried twice, produces something generic, and quietly becomes another unused subscription — while the firm concludes, reasonably, that AI was overhyped.
Why a chat window can’t run a practice
A practice runs on four things a chatbot does not have. Context: it does not know your matters, your deadlines, or your clients — you have to paste your practice into it, one prompt at a time. Memory: whatever it learns about how your firm writes is gone by tomorrow; nothing compounds. Execution: it cannot update your CRM, stage a send, or sweep a docket — its output is text you still have to go act on. Audit: it keeps no record of what was reviewed, approved, or sent, which is untenable in professional work.
An operating system, not a copilot
The distinction that matters is between a system of record and a system of action. The CRM records what happened. An agentic operating system for IP practice acts: it triages the pipeline, drafts in the firm’s voice from the firm’s own notes, keeps the CRM current, and stages every client-facing artifact for review. A copilot helps you do the work you were already doing. An operating layer does the routine work so it arrives done. For a trademark, copyright, or patent boutique, that is the difference between a tool you remember to use and a practice that runs.
The adoption unlock: never write a prompt
The reason Big Law’s model does not translate down-market is that it assumes someone will operate the AI. The boutique answer has to be dashboard-first: the work arrives by email and text, the day’s decisions live in an approval queue, and the firm needs no AI subscription and no AI skills. If the first touch feels like homework, a boutique will never come back. It has to feel like relief — the machine work already handled, waiting for a yes.
The trust architecture: the approval gate
Autonomy is not what earns a law firm’s trust — a gate does. Nothing client-facing sends until an attorney approves it; approval is what authorizes the send; every resolution carries an audit line. That single constraint is what makes agentic software compatible with professional responsibility: the attorney remains the decision-maker, and the software remains software.
The boutique IP bar does not need a better chatbot. It needs the same class of operating layer Big Law is getting — built for firms with no one to deploy it, priced as flat monthly SaaS, and governed by the one rule that never bends. AI drafts. You approve.