Should we pursue law firms as clients?
I directed research into the legal obligations, documented failures, available products, and operating realities behind that question.
How I worked with AI
I set the research question and evidence standard, directed where the research went, challenged assumptions, and made the commercial recommendation. Claude handled much of the collection and synthesis, then helped pressure-test the market hypothesis.
What is making AI adoption difficult for law firms?
Before judging the market, I needed to know whether the operational problem was real. I looked at where legal AI use was failing and what professional guidance actually required, then compared that with the controls vendors already provide.
What the evidence showed
A review requirement is weak if the workflow can’t show it happened.
Across the AI-sanctions cases I sampled, a recurring failure was inadequate human review, including one case where the supervising attorney signed off without reading the filing. ABA guidance makes clear that using AI does not remove a lawyer’s existing ethical responsibilities. Making review explicit and, where useful, recording that it happened is my governance response, not a rule-mandated workflow.
A small, non-exhaustive sample, not a census of published cases; case details are as reported by secondary legal-AI trackers, not independently re-verified against primary court records.
The visible failures cluster in high-consequence work.
Litigation was one of the least common AI use cases in the adoption data I reviewed, yet every case in this sanctions sample involved a court filing. That’s not proof litigation is uniquely risky. It’s more likely a visibility problem: litigation produces public court orders, while a bad AI-drafted contract clause or internal memo usually doesn’t leave a public trail. The real lesson is that governance can’t be prioritized by how often something is used. Low-frequency, high-consequence work still needs the stronger controls.
Confidentiality earns its own section.
Every firm AI policy in my sample gave data confidentiality a dedicated section, which lines up with attorneys’ existing confidentiality duties, including ABA Model Rule 1.6. It’s one of the most solidly evidenced pieces of this research.
A lawyer must verify AI-generated citations.
Citation-dependent work cannot be finalized until the authorities are checked.
Record who verified the citations and when.
The lawyer still decides whether the authority supports the argument.
Once the problem was clearer, I shifted from “is there a real gap?” to “does that gap create a client market worth pursuing?”
Given those problems, are law firms worth pursuing?
I looked at adoption evidence and firm structure to understand who would own and buy this work. That made me question the initial market idea.
Most solo and small firms don’t have dedicated IT support.
ABA technology survey data shows that solo attorneys usually make technology decisions themselves. Internal tech support is rare at small firms and becomes much more common only at larger ones. My initial thinking had quietly assumed a firm with a named policy owner and a second reviewer to catch mistakes. That assumption does not hold for most of the addressable market. It also raises a harder question: are firms that size worth pursuing as clients, given the cost of selling to and supporting them one at a time?
A dedicated legal-AI governance offering could serve law firms broadly as clients.
Solo and small practices appear most structurally underserved, but they are also the hardest segment to justify pursuing one relationship at a time.
Real gap. Do not propose an offering yet.
The operational problem is real: review is hard to prove, and a generic governance layer could miss the difference between low-risk and high-risk work. The commercial case isn’t strong enough yet, though. My recommendation is not to propose an offering until direct customer discovery shows that a reachable segment has both the need and the budget for consulting help.
What I ruled out: treating "law firms" as one market or treating a real problem as proof of a viable client base. I also stopped assuming every firm has dedicated governance and IT owners.