ChatGPT for lawyers: where it ships, where it sinks

Most lawyers asking "what's the best ChatGPT for me?" are buying a chat window when the firm needs an operating system.

This is the article we'd write for a law-firm partner about to spend money on AI and wanting to spend it well. It covers the cost math, what fits which kind of firm, the ethics-rule constraints, the alternative tools, and when building your own system beats operating someone else's tool. Every claim here comes from work shipped inside live mid-market companies, including hands-on work in the legal vertical.

It's not the chat window, it's the build

ChatGPT is a tool. A law firm is a system. The reason most ChatGPT-for-lawyers articles read like a snake-oil prescription is that they're answering the wrong question. They ask "is ChatGPT good for lawyers?" when the actual question every working partner is asking is "what do I buy, build, or hire so my firm runs on AI instead of around it?"

The 1995 analogy lands here cleanly. Companies in 1995 who thought of "the internet" as a thing to buy (a website, an email server) got a small bump. Companies that built the business around the internet got everything else. The same shape is playing out with AI, just faster. ChatGPT is the 1995 website. The intelligence layer is the business built around the internet.

Here's the part most coverage misses. The advantage of AI isn't the chat window everyone can buy, it's that AI now lets a firm build custom software around its own work, fast and cheap enough to finally make sense for a single practice. Five years ago, custom software for a 12-person firm was absurd: too slow, too expensive, out of date by the time it shipped. AI collapsed that cost, and because the software is built and maintained with AI, it keeps improving instead of freezing the day it ships. That continuously improving custom software is the actual product. ChatGPT is the raw material.

Most ChatGPT advice for lawyers operates two layers below the real question. Prompt tips, sample prompts, use-case grids: tool-handling content, useful the first month and irrelevant the second. The strategy question is upstream of the tool question, and the playbook for working that out doesn't change much between a software company and a law firm.

Where ChatGPT works (and where it gets you sanctioned)

ChatGPT ships real value for three things in a law firm: first-pass drafting (correspondence, internal memos, demand-letter shells), document summarization (depositions, discovery files, long contracts), and non-billable admin (intake routing, scheduling logic, internal training material). All three are situations where the failure mode is wasted attorney time, not malpractice.

It fails, sometimes catastrophically, at three other things: citation accuracy (it invents cases), client-data confidentiality (the public version stores prompts), and audit trail (you can't reconstruct who reviewed what). To be fair, those are failures of the raw chat window, not of AI built properly. A custom system handles all three: it grounds citations against real case law so it stops inventing them, runs on isolated infrastructure the public model never touches, and logs who reviewed what. The chat window can't do those things, but a system built around it can. The bright line between "ChatGPT helped me draft this" and "ChatGPT got me sanctioned" runs along the citation-accuracy axis. In 2023, a New York lawyer in Mata v. Avianca cited fabricated cases ChatGPT had hallucinated; the sanction was $5,000, plus an order to notify his client and each real judge whose name had been attached to a fabricated opinion. Every legal-tech vendor has told that cautionary tale a hundred times. The honest version: it wasn't ChatGPT's fault, it was a process failure. The lawyer skipped the verification a paralegal would have done in 1995, and the workflow didn't catch it.

ABA Formal Opinion 512 (2024) codified the obvious: under Model Rules 1.1, 1.6, and 5.1/5.3, you have to know what your AI tool can and can't do, keep client data out of tools with unclear retention, and verify outputs the way you'd verify a paralegal's. Every legal-tech article quotes it; almost none describe what compliance looks like at the workflow level. That answer is unglamorous: a review step on every AI workflow before it ships, client data isolated from any public model, and a documented supervision chain. Same shape as supervising a junior associate, with a tool instead of a person.

A February 2026 ruling made the confidentiality risk concrete, and it's sharper than most lawyers assume. In United States v. Heppner, the first federal case to address it head-on, the Southern District of New York held that materials a defendant created by feeding his case details into a public consumer AI tool were not protected by attorney-client privilege or the work-product doctrine, and that 31 of those AI-generated documents had to be turned over to prosecutors. The tool there was Anthropic's Claude, but the reasoning covers any public model, ChatGPT included: there is no attorney-client relationship with an AI platform, and the platform's own terms let it use and disclose your inputs, so there is no reasonable expectation of confidentiality. The court did leave one door open. Had counsel directed the AI use, privilege might have held. That is the whole point: a public chat window is a third party, but the same work done inside a counsel-directed, controlled system stays protected.

Confidentiality is one axis a controlled system fixes. Accuracy is the other, and it's the one vendors oversell hardest. No AI system is accurate on day one, whatever a vendor promises. The realistic curve: around 60% on day one, 85% by month three with team feedback, and past 95% on the workflows the team uses daily. The 95% is the number that matters; the day-one number is sales theater.

And because a custom build is software that's actively maintained, it keeps getting better at exactly those things over time, not stuck where the public tool stops. The limitations aren't a ceiling, they're the part of the job a build is for.

The three options, and what they really cost

The headline prices are easy to find. The real cost is the work the firm has to do around each option, and which one fits the way the firm actually runs.

The three options most firms compare:

option	direct cost	hidden cost	best fit
ChatGPT Team	$30/seat/mo (~$7K/yr for a 20-person firm)	You enforce the guardrails by hand: a review step on every output, citation checks, keeping client data out of the public model	You want the raw tool for drafting, summarization, and admin, and you can run your own review step. The entry point at any size
Legal-specific SaaS (CoCounsel, Spellbook, Lexis+ AI, Paxton, MyCase IQ)	$75–$500/seat/mo (~$18K–$120K/yr for a 20-person firm)	Vendor lock-in, and you live with their default workflow even where it's two clicks from what you actually do	A mainstream workflow (legal research, contract review) that a legal-tuned tool already fits well
Custom intelligence layer (fractional CAIO retainer, builds included)	~$15K–$25K/mo, scaled to scope (a solo build sits at the low end)	Onboarding time (a few weeks to the first live build), and senior-partner attention to shape what gets built	You want an AI hire that builds systems around your own data, maintains them, and grows them over time, instead of operating a tool yourself. Fits any firm, solo or large, even one whose work an off-the-shelf tool could cover

The mistake most firms make here is treating the price tiers as a budget question. They're not. They're a fit question. Every firm has repetitive, high-value work, so that isn't the dividing line. The dividing line is what you want done with it. A firm paying $25K/mo for a custom build it won't give the partner attention to shape and adopt is wasting money. A 30-attorney firm hand-building its workflow on top of raw ChatGPT, when what it actually wants is those systems built and maintained for it, is wasting senior-partner attention. The deciding question isn't how many attorneys you have, it's whether you want an AI hire building and maintaining custom systems around your data, or a static tool you operate yourself.

When ChatGPT is enough, when a legal tool fits, when you build

Here's the call, option by option.

Reach for ChatGPT Team when you want the raw tool for faster drafting, summarization, and admin, and you're fine running your own review step. Get fluent here first, whatever your size.

Reach for legal-specific SaaS (CoCounsel for research, Spellbook for contract review, Paxton for litigation prep, Lexis+ AI for citations) when your workflow is genuinely standard and a tuned tool already fits it. The models are trained on legal corpora, which lowers the hallucination rate. But the confidentiality and audit controls these tools lead with aren't unique to them, a custom build delivers the same, and SaaS adds two costs of its own: lock-in, and a workflow shaped to the vendor instead of to you. It's the right call when your work really is standard. The moment it isn't, you're bending the firm around someone else's product.

Build a custom intelligence layer when you want an AI hire building systems around your own data, maintaining them as the models and the firm change, and growing them over time, rather than operating a fixed tool yourself. Every firm has work worth automating, so that isn't the signal, and neither is whether an off-the-shelf tool could technically cover the work. The signal is what you want done with it: systems built and maintained around your own data, not a vendor-shaped tool you run by hand. The case is sharpest when your work is specific to you (the same prompt typed by four people this week, outputs that need the same manual fix every time, a SaaS default flow two clicks from what you actually do, client-data handling blocking the team from leaning in), but it holds even where a legal tool would fit, because a build delivers the same workflow without the lock-in, keeps improving shaped to the firm, and folds that workflow into one system instead of one more subscription. What qualifies a firm for a custom build is wanting that built and maintained around its own data, not its size.

The custom layer isn't a different ChatGPT. It's the connective tissue around it (or Claude, or whichever foundation model is ahead this quarter): firm-specific intake routing, drafting agents trained on your templates, research agents that respect your jurisdiction, document QC, conflict-check automation, all behind one UI with role-based permissions. The model is a replaceable component. The intelligence layer is the durable system around it.

What a custom legal intelligence layer actually looks like

Concretely, it's usually four or five firm-specific agents wired into one system. A representative shape for a litigation-heavy boutique:

• An intake agent that classifies inbound leads by case type, runs an initial conflict check, generates the first-draft engagement letter, and routes the file to the right paralegal.
• A drafting agent trained on the firm's templates and prior winning briefs, that produces first-pass drafts a partner can edit in 20 minutes instead of starting from scratch.
• A research agent that searches Westlaw or Lexis (depending on the firm's existing subscription) and produces citation-verified memos with the case quotes inline.
• A discovery summarization agent for long document sets (depositions, contract collections, regulatory filings) that produces a paragraph summary per document with timestamps and key-passage links.
• A secure document-handling agent for any workflow that touches privileged client data, running on a hardened instance with no public-model exposure (NVIDIA's NemoClaw or similar).

Five components, one authentication layer, one data layer, one UI. A paralegal texts the intake agent like a colleague; a partner asks the research agent for a memo. Swap the underlying model when a better one ships; the system around it stays.

This is what an AI consultant who actually ships builds delivers: the system, not the tool recommendation. The custom-layer framing is the operator-shaped answer to "should we use ChatGPT?" and it's the answer nobody selling a SaaS subscription is incentivized to give.

How to test this before you commit

Pick one workflow and run a 90-day pilot. It's a paid quarter, not a free trial, but it's a low-risk way in: a quarter instead of a year, a working build at the end, and an easy exit if the value isn't clear.

The workflow should be specific. "AI for our firm" is too vague to act on. Examples that work: "draft demand letters from intake forms in our personal injury practice," "summarize incoming discovery PDFs into one-page case-fact briefs," "auto-generate engagement letters from the intake-conflict-check output," "produce client-update emails from the case-management activity log." Each is one workflow, one output, one team that uses it. The first working version is quick to build: not polished, but enough that a partner can see the shape and judge whether the value is real.

The test is the trajectory, not the demo. Full team adoption takes one to two quarters, so the pilot won't show everyone living on the tool by day 90, and it isn't supposed to. What it should show is the early signal: trustworthy output, a workflow that fits how the firm actually works, and the people who touch it starting to reach for it instead of the old way. Don't sign a monthly retainer on the strength of one working demo, sign it once that signal is real. When you do, the first month of the retainer goes straight into the next build, because the pilot already did the discovery. If the signal never shows, you walk, out one paid quarter instead of a year. A fast-moving firm on a focused workflow can see payoff inside that first month; a larger or slower one should plan for up to two quarters, because the limit isn't the build, it's how fast the team changes the way it works.

Fire me anytime.

The contract shape no SaaS vendor offers, and the one most working partners actually need, is monthly, no lock-in, where the person has to keep delivering enough value that you don't want to fire them. That's the fractional CAIO retainer. And the value isn't the first build. It's that someone keeps the workflows you already have current as the models and tools underneath them change, and ships new ones as the firm changes: a new practice area, a new bottleneck, a tool worth wiring in. AI systems drift if nobody tends them, so the CAIO's standing job is to keep the existing builds sharp and find the next workflow worth automating, month after month. The price doesn't move when the scope does. Either side can end it anytime.

If we're a fit, the first step is the 90-day pilot. Pick the workflow, build it, see if the value is real. If the answer is yes, the retainer is the obvious next step, not a sales conversation. If the answer is no, we shake hands and you've spent a quarter instead of a year.

For any firm ready to put its work into maintained, custom-built systems, the math is straightforward. A year of fractional CAIO retainer costs less than a single underperforming associate. One ChatGPT-hallucinated citation that reaches a brief costs more than two months of it. And the cost of not having an AI system in 2027, when half your peers do, is harder to quantify but isn't zero.

Start with one workflow. Run a 90-day pilot. Fire me anytime. Book a 30-minute call and we'll walk through one of the live builds.

Frequently asked questions

Is there a ChatGPT for lawyers? There isn't a single "ChatGPT for lawyers" product. There's ChatGPT itself (general-purpose, $30/seat/mo for the Team plan), there's a tier of legal-specific AI tools that wrap foundation models with legal-tuned training and confidentiality controls (CoCounsel, Spellbook, Lexis+ AI, Paxton, MyCase IQ, $75–$500/seat/mo), and there's the option of building a custom intelligence layer for the firm that uses whichever foundation model is best for each workflow. The right answer depends on whether you want the raw tool, a vendor's pre-built tool for a standard workflow, or an AI hire building and maintaining custom systems around your own data, not on which product has the best marketing.

How much does ChatGPT for lawyers cost? ChatGPT Team is $30/seat/mo. ChatGPT Enterprise (with stricter data controls, longer context, admin controls) is roughly $60/seat/mo at typical firm sizes. The cost that's easy to miss isn't the subscription, it's the review step every output needs before it leaves the building. Build that into the workflow and the tool pays for itself. Legal-specific SaaS runs $75–$500/seat/mo, and a custom intelligence layer (a fractional CAIO retainer with builds included) runs roughly $15K–$25K/mo, scaled to scope.

What is the 80/20 rule for lawyers? The 80/20 rule for AI in a law firm: 80% of the value comes from automating the right 20% of repetitive workflows. Drafting, summarization, and intake almost always sit in the 20%. Strategy, judgment, and client relationships sit in the 80% that should not be automated. The mistake most firms make is trying to automate the wrong 20%: generic prompts for legal research instead of firm-specific automation of a specific repeated task.

What happens when your lawyer uses ChatGPT? If they use it correctly (with a review step, with client data isolated from the public model, and with citations verified against an authoritative source), they ship more work in less time. If they use it incorrectly (pasting client data into the public version, citing case law without verification, treating the output as final), they risk an ethics violation under ABA Model Rules 1.1, 1.6, and 5.3, and in the worst case sanctions like the Mata v. Avianca outcome. A 2026 ruling, United States v. Heppner, added another risk: feeding case details into a public AI can waive attorney-client privilege and make those exchanges discoverable by the other side. The bright line is the review step, plus keeping client data out of any public model. ChatGPT isn't the failure mode; skipping the verification is.

Does our firm need a custom AI system or is a SaaS tool enough? It's a fit question, not a headcount one. Every firm has repetitive, high-value work, so that isn't the dividing line. If you just want the raw tool or a vendor's pre-built tool for a standard workflow, ChatGPT Team or a legal-specific SaaS tool (CoCounsel, Spellbook, Paxton, Lexis+ AI) is enough. If you want an AI hire building and maintaining custom systems around your own data, a custom intelligence layer pays back, whether you're a solo practice or a 50-attorney firm, and even where an off-the-shelf tool could cover the individual workflow, because the build delivers it without the lock-in and keeps improving around your firm. The signals to watch for: the same prompt typed by four people this week, outputs that need the same manual fix every time, a tool whose default workflow doesn't match yours, or client-data handling blocking adoption.

How long until an AI system delivers value? Plan for a one-month-to-two-quarter window. A single focused workflow at a fast-moving firm can deliver value within the first month. A larger or slower-moving firm should expect up to two quarters (about six months), because the bottleneck isn't building the system, it's how quickly the team adopts it into daily work.