Live in Claude Cowork or ChatGPT, or get left behind
Why your team should do its AI work in Claude or ChatGPT, connected to Clio and your files, instead of in a separate legal AI app.
Founder of Conduit Law (non-attorney), COO · Co-Founder, FirmOps
One workflow
Request demo
Bring one recurring bottleneck. If it is a good first firm-brain build, the next step is a live demo — not a pitch deck.
Your lawyers and paralegals should do their AI work inside Claude or ChatGPT, connected to the firm's real files. Not inside a separate legal AI app. This page explains why, in plain terms.
Last updated: October 29, 2026
TL;DR
- Claude (from Anthropic) and ChatGPT (from OpenAI) are where the best AI is built and released first. Staff who work there get every improvement the day it ships.
- A separate legal AI app asks your team to leave that workspace and work in the app's own browser tab instead. You then wait on that vendor to catch up, and you pay them on top of the AI itself.
- The reason firms get pulled into those apps is simple: the app already has the files. The fix is to connect your files to Claude or ChatGPT instead. That connection is what we call a firm brain.
- Keep Clio or Filevine. They stay your system of record. The firm brain sits on top of them.
- We run this way at Conduit Law, a working personal injury firm, every day.
Why do lawyers and paralegals have to live in Claude Cowork or ChatGPT?
Because that is where the best AI is, and it keeps getting better there first.
Anthropic and OpenAI build the AI models. They also build the apps your staff use to work with them: Claude (including Claude Cowork, its tool for longer, multi-step tasks) and ChatGPT. When a better model comes out, it shows up there first. A legal software company is always a step behind those two.
It does not matter much which of the two you pick. Choose the one your team will actually use every day. Practical safe-use notes are on Claude for lawyers and ChatGPT for lawyers.
One caveat: Claude or ChatGPT on its own does not know anything about your cases. That is the gap a firm brain fills, covered below.
What happens if the firm lives in a legal AI browser app instead?
The firm ends up working inside someone else's product, on their schedule.
Many legal AI products are built on top of the same models Anthropic and OpenAI already offer, with a legal-looking screen added. That can make for a good demo. But every time Anthropic or OpenAI releases something better, your team has to wait for the vendor to add it. The vendor also has to make money on top of the AI it is reselling, so it decides which model you get and how much of it you can use.
The result, six months later: staff are working in the vendor's tab, the AI is behind what is available in Claude or ChatGPT, and the firm is paying extra for it.
This is not an argument against Clio or Filevine. Those are where your matters, documents, and billing live. Keep them. The question is only where your people do their AI work.
What about Clio's AI?
Clio's AI features, including Clio Work, are Clio products. They are useful for questions about what Clio already holds. But they mostly see Clio. Your inbox, your document folders, your follow-ups, and your other systems are not all in Clio, so an assistant inside Clio can only answer part of the question.
Our advice: keep Clio, use its AI where it helps, and do the firm's daily AI work in Claude or ChatGPT connected to Clio and everything else. The longer comparisons are practice management AI vs a firm brain and Clio Work vs Filevine vs a purpose-built firm brain.
How does a firm brain keep you in those workspaces?
It connects Claude or ChatGPT to the firm's real systems, so staff can work on a case without leaving.
Without that connection, your team copies facts out of Clio, the case folder, or email, pastes them into Claude or ChatGPT, then copies the draft back. That copy-paste work is exactly what legal AI vendors use to sell their apps: "our app already has your files."
With a firm brain, Claude or ChatGPT can look up the matter, the documents, and the email thread directly, then prepare a draft that a person reviews before anything goes out. The technical name for the connection is MCP, an open standard both Anthropic and OpenAI support. FirmOps decides with you what the AI is allowed to see and do, builds it on the systems you already have, and keeps it running. That service is the Managed Firm Brain. It is built for your firm, not installed from a box.
If you want to see what this looks like for a staff member, read Claude Desktop plus MCP. It is one part of broader law firm operations work.
Is this proven on a live firm, or a slide deck?
It runs every day at Conduit Law, a personal injury firm.
We built it for ourselves first. At Conduit, marketing, intake, onboarding, case management, settlements, trust, and accounting are all connected, and the team does its AI work in Claude or ChatGPT with those systems behind it. Our staff are not stuck re-typing data or jumping between twenty browser tabs. Details we can share publicly are in the Conduit Law case study.
Can Claude and ChatGPT actually reach the firm's systems today?
Yes. Both companies support it and document how.
MCP is an open standard that lets AI apps like Claude or ChatGPT connect to data, tools, and workflows.1 Anthropic says these custom connections work in Claude, Claude Cowork, and Claude Desktop on every plan (free accounts are limited to one).2 OpenAI's documentation shows how to add one in ChatGPT.3
Two practical points for a law firm:
- Where the connection comes from. Claude connects from Anthropic's servers, not from the computer on your desk. A system that only works on the office network or VPN will not connect unless your IT allows Anthropic's addresses through.2 OpenAI offers a secure tunnel option for private systems.3
- Who sees what. On Claude's Team plan, only the account owners can add a connection. Each staff member then signs in on their own, so the AI sees only what that person is allowed to see, unless the connection is set up with one shared login.2 Getting that right is a setup decision, and it is part of what FirmOps does.
A note on names: Anthropic is merging Claude's regular chat and Claude Cowork into one experience. It is rolling out on Pro and Max plans and can be turned on for Team and Enterprise plans.4 Whatever it is called, the advice is the same.
Where do you work, and who takes the margin?
Here is the choice side by side.
| Work in Claude or ChatGPT, connected by a firm brain | Work in a separate legal AI app | |
|---|---|---|
| Where staff work | Claude or ChatGPT, directly. | The vendor's browser app. |
| Who you wait on for improvements | Anthropic or OpenAI, who build the AI. | The legal AI vendor, who resells it. |
| Which AI you get | The latest model, when it is released. | Whatever model the vendor chooses to offer. |
| Who you pay | Anthropic or OpenAI for the seats, plus the firm brain built on your systems. | The vendor, who pays for the AI and adds their markup. |
| Connected to Clio, email, and files | Yes, through the firm brain. | Sometimes, often partly. Staff may still copy and paste. |
| Who controls the setup | Your firm. Built with you, run for you. | The vendor. Their roadmap, their choices. |
A focused legal tool can still be worth buying for a specific task. The mistake is making it the place your whole team works. For more on buying a tool versus building your own layer, see legal AI software vs a firm brain.
What should the firm do instead of shopping another legal AI tab?
Keep Clio or Filevine. Keep your team in Claude or ChatGPT. Connect the two.
Start with one recurring task that keeps falling between the case system, the document folder, and the inbox. Connect just that task. The AI reads first, then drafts, and a person approves anything before it is sent. Phone calls stay with people. The step-by-step approach is in read-first AI for law firms, and the approval rule is in AI approval gates for law firms.
If you are comparing specific vendors, see our MSO and legal tech alternatives overview, plus Torticity, Eve, Finch MSO, and Orion Legal MSO. When you have that first task in mind, request a demo.
Questions owners actually ask
Do we have to pick Claude or ChatGPT?
Pick the one your team will actually use. Both work. The real choice is between those two and a separate legal AI app.
Should we just use the AI built into Clio?
Use it where it helps. But it mainly sees what is in Clio. For work that also involves email, documents, and other systems, connect Clio to Claude or ChatGPT. See practice management AI vs a firm brain.
Can we just buy a legal AI app instead?
You can. Your team would then work in that app, wait on that vendor for improvements, and pay their markup. See legal AI software vs a firm brain.
Is a firm brain a product we install in Claude?
No. It is built with your firm, on your systems, and then run for you. MCP is just how it connects. See what is a firm brain.
Is Claude Cowork still a separate product?
Less and less. Anthropic is combining chat and Cowork into one Claude experience, rolling out on Pro and Max and available on Team and Enterprise where it is turned on.4 Connections work across Claude, Cowork, and Claude Desktop either way.2
Does this replace Clio or Filevine?
No. Keep them. The firm brain sits on top. See Clio Work vs Filevine vs a purpose-built firm brain.
How do we start?
Pick one task and request a demo. We will show whether Claude or ChatGPT can handle it with your systems connected.
Sources
Keep moving up the operating system
Get the Managed Firm Brain Notes
Follow the practical path: live workflow builds, CRM-aware automation when the data is ready, and supervised work after approval.
Demo Context
What to watch for in the Managed Firm Brain walkthrough
Live-Lab Proof
How Conduit informs the first external deployments
Fit Criteria
API-ready systems, workflow scope, and approval gates
About Jonathan Mahler
Jonathan Mahler is the founder of Conduit Law (non-attorney) and COO, and co-founder of FirmOps. He runs the systems of a live PI firm every day, then turns reusable patterns into practical Managed Firm Brain workflows: approved context, supervised drafts, approval gates, and a path into deeper automation when the first workflow proves value. Meet Jon and Elliot.
Related Articles
What is a firm brain?
A firm brain is a purpose-built operating layer for one law firm, implemented in partnership, then run for that firm. Not a chatbot. Not a boxed product.
Legal AI software vs a purpose-built firm brain
Buying legal AI software gives you a tool. A firm brain is a custom implementation built with the firm, then operated for them.
Law firm MSO alternatives: stay in Claude or ChatGPT
The alternative to lit-tech vendors, legal-AI browsers, and MSOs is the same move: live in Claude Cowork or ChatGPT with a firm brain.
Request demo
Bring one recurring bottleneck and see whether it is a good first firm-brain build. If the first workflow is concrete, the next step is a live build session, not a broad advisory project.
Live demo | Human approval before external action