First-draft support
Status letters, client updates, engagement letter sections, internal memos, and plain-English explanations of a process. ChatGPT gets you from a blank page to a reviewable draft in seconds. A person still edits and signs.
ChatGPT for lawyers
ChatGPT for lawyers: use it safely, then connect it to the case file.
ChatGPT drafts and summarizes well, but it does not know your matters. FirmOps, a Vertana brand, builds and runs an MCP server that connects your case system (Clio, Filevine, SmartAdvocate, others), Gmail or Microsoft 365, documents, billing, and phones to ChatGPT. The brain drafts and stages. A human always releases.
By Jonathan Mahler · Updated November 1, 2026
Practical uses
Where ChatGPT earns its seat in a law firm.
Used inside a business workspace, ChatGPT is a fast first-draft and summary tool. Everything below assumes a person reads the output before it leaves the firm.
Summaries of material you give it
Deposition excerpts, a carrier email thread, a long policy, a medical bill ledger. Ask for a summary with page or paragraph references so the reviewer can check each point against the source.
Checklists and SOPs
Turn a messy process into a step list, then have the people who do the work correct it. Good for intake scripts, file-opening steps, and new-hire training.
Research starting points
Issue spotting, questions to ask, and a map of what to look up. Never cite a case, statute, or rule from ChatGPT output without pulling and reading the source yourself.
Running PI demands? Demand from the File is a free ChatGPT skill that drafts a personal-injury demand from uploaded records and lists the gaps a paralegal still has to fill.
Plans and confidentiality
Pick the plan before you pick the prompts.
The account matters more than the model. OpenAI says its business products, such as ChatGPT Business, ChatGPT Enterprise, and the API, are not used to train its models by default, and give admins control over the workspace. Personal accounts are different: they have their own data and training settings, and the firm cannot see or control them. Client facts never go into a personal account.
Read the current business terms yourself and keep a copy in the file. Ask three questions: is our data used for training, how long is it kept, and who can see it inside our workspace.
The ethics side is settled enough to act on. ABA Formal Opinion 512 (July 2024) applies duties lawyers already have: competence, confidentiality, communication with clients, supervision of lawyers and nonlawyers, and reasonable fees. The key lines are summarized with sources on our law firm AI statistics page. Turn them into rules your staff can follow with a law firm AI policy. This page is operator guidance, not legal advice.
What it cannot do alone
A chat box does not know your file.
Out of the box, ChatGPT knows the public world and whatever you paste. That gap is why staff start copying matter notes into chat windows, which is the shadow AI pattern: no log, no permissions, no review.
- It does not know your matters. Ask about the Garcia file and it will guess or ask you to paste it.
- It cannot see your inbox, your document folders, your billing system, or your phone logs unless something connects them.
- It does not know your firm’s fields, stages, deadlines, or who owns which task.
- It has no idea which staff member is allowed to see which matter.
- It cannot hold a draft for approval inside your case system. It hands you text and you carry it from there.
Connecting the case file
What ChatGPT can connect to natively, and where MCP comes in.
MCP, the Model Context Protocol, is an open standard for connecting AI apps to other systems. Think of it as one controlled door the AI walks through to read a system and, with permission, prepare work in it. The plain-English version for firms is in what is MCP for a law firm.
Projects and custom GPTs
Group instructions and reference files (your SOPs, templates, style rules) so staff get consistent output. Good for firm knowledge. Not a live view of any matter.
Connectors
On business plans, ChatGPT can connect to common sources such as cloud drives and email through connectors (OpenAI also calls these apps). What is available depends on your plan and admin settings, and it changes often, so check your workspace before you plan around one.
Custom connectors over MCP
ChatGPT can also talk to custom connectors built on MCP, the Model Context Protocol. That is the door a case system, a billing system, or a phone platform comes through when no off-the-shelf connector exists.
What the native options leave open
- Your case system with your fields: matter stage, injury type, treating providers, liens, statute dates, the custom fields your firm actually runs on.
- Scoped permissions per role, so the intake team does not see what only the attorney should.
- Approval gates before anything sends, files, or writes back to the record.
- A log of what the AI read and what it prepared, so a supervising lawyer can check the work.
- One view across systems: the matter, the carrier email, and the bill in a single answer.
How a firm brain fills the gap
FirmOps builds and runs the MCP server for your firm. It connects your case system, Gmail or Microsoft 365, document storage, billing, and phone platform into one layer ChatGPT reads through. Each role sees what it is allowed to see. Every read and draft is logged.
The brain drafts and stages: email drafts in the sender’s drafts folder, proposed task updates held behind approve and reject controls. A human always releases. Your team keeps working in ChatGPT, and the same layer works in Claude.
We build it and run it. The brain is yours. See the Managed Firm Brain and the integrations we connect.
Example prompts
What a connected ChatGPT can answer, and what a person still releases.
These are typed into ChatGPT the way you would ask a paralegal. The difference is the answer comes from your systems, with sources, instead of from a guess.
| You ask | The brain does | A person releases |
|---|---|---|
| “What’s the status of the Garcia matter?” | Reads the matter, the latest notes and tasks, open records requests, and the last client contact, then answers with sources. | Nothing to release. It is a read. |
| “Draft an update for every matter quiet over 30 days.” | Finds matters with no client contact in 30 days and drafts a plain-language update for each from the file. | Staff review each draft in email drafts and send the ones that are right. |
| “Which medical records requests are still outstanding?” | Lists open requests by provider and age, and drafts the follow-up for any past your firm’s threshold. | A paralegal approves the follow-ups. |
| “Summarize the carrier’s last three emails on this claim.” | Pulls the thread from Gmail or Microsoft 365, summarizes positions and dates, and flags anything that needs a response. | The handling lawyer decides the response. |
| “List new leads from this week that have not been called back.” | Checks intake records against call logs and lists who is waiting and for how long. | People make the calls. The brain does not phone clients. |
| “Draft the provider letter from the file.” | Fills the letter from the matter: client, date of loss, provider, authorization on file. | Staff check it and send it. |
| “What is unbilled on matters over 90 days?” | Reads the billing system and lists unbilled time and costs by matter and owner. | The owner or billing lead decides what goes out. |
| “Prep the file-opening checklist for today’s signed client.” | Compares the signed intake with the opening checklist and lists what is missing. | Staff open the matter and confirm the checklist. |
Most of these rows are paralegal and staff work, which is where a connected model pays off first. More on that in AI for paralegals.
Guardrails
The tool is easy. The review path is the product.
Six rules that keep ChatGPT inside the lines a supervising lawyer can defend. People keep live client calls. AI works behind the phone.
- Use a business workspace the firm controls. No client facts in personal accounts.
- Do not ask ChatGPT to make final legal decisions, set case value, or give a client advice. Those stay with a lawyer.
- Verify every case, statute, and quote against the source before it goes anywhere.
- Keep approval gates on every client message, filing, and record change. A person releases the work.
- Write the rules down in a firm AI policy and train staff on it, including who supervises whom.
- Tell clients how you use AI when it matters to them, and do not bill AI time as if a person did the work.
FAQ
ChatGPT in a law firm: straight answers.
Is ChatGPT safe for lawyers to use?
It can be, on a business plan the firm controls, with written rules and a person reviewing the output. OpenAI says business products such as ChatGPT Business and Enterprise are not used to train its models by default. Personal accounts have their own data settings, so client information should never go into one.
Can ChatGPT read my Clio, Filevine, or SmartAdvocate matters?
Not on its own. It needs a connector. Off-the-shelf connectors cover common tools like cloud drives and email. A case system with your firm’s fields usually needs a custom connector built on MCP, with permissions scoped to what each person is allowed to see.
What is MCP, in plain words?
The Model Context Protocol is an open standard for connecting AI apps to other systems. ChatGPT and Claude can both use MCP connectors. For a law firm it is the plumbing that lets the AI read the case file, the inbox, and the billing system through one controlled door.
What does ABA Formal Opinion 512 require?
The July 2024 opinion applies existing duties to generative AI: competence (understand the tool and check its output), confidentiality (know where client information goes and get informed consent when needed), communication with clients, supervision of lawyers and nonlawyers, and reasonable fees. It is guidance, not legal advice for your firm, and your state bar may have its own.
Should staff paste client documents into ChatGPT?
Not into a personal account, and not as a habit. Copying and pasting is how shadow AI starts: no log, no permissions, no review. Connect the systems once, with scope and logging, so nobody has to paste.
Does FirmOps replace ChatGPT?
No. Your team keeps using ChatGPT (or Claude). FirmOps builds and runs the MCP server behind it so it can read your case system, email, documents, billing, and phones, and stage drafts for a person to release.
Is ChatGPT or Claude better for a law firm?
Both are frontier models and both can use MCP connectors. Pick by which one your team already works in and by the review burden on your real tasks. The connection layer should work with either, so you can switch when the next model lands.
Deciding between the two models? Read Claude vs ChatGPT for lawyers and why we tell firms to live in Claude or ChatGPT instead of another legal AI tab.
Keep reading
Related guides.
Claude for lawyers
The same question for Claude: projects, connectors, and the desktop app.
Claude vs ChatGPT for lawyers
Choose by review burden, not model hype.
AI for law firms
The 2026 guide to what AI does in a firm, by function.
Legal AI tools compared
Research, drafting, intake, and demand tools, side by side.
Clio and AI
Clio Work, Clio Manage AI, and a layer across the rest of the stack.
Law firm AI integrations
Connect Clio, mail, and files to ChatGPT or Claude.
Law firm AI statistics
Primary-sourced adoption, ethics, and time-saved numbers.
Legal AI adoption
A rollout path that starts safe and expands after proof.
Next step
See ChatGPT answer from a real case file.
The 30-minute live demo shows a firm brain answering matter questions and staging drafts for approval. We map your stack and confirm your systems have the APIs to connect.
Put a number on it
ChatGPT Team is a line that stays. The legal AI seats around it often do not. Check with the software cost calculator. It starts from a real firm's stack, tags each line Stays or Consolidates, and nets the $500 a month FirmOps maintenance; what that covers is on the pricing page.