Watch a matter intake run end to end
Four unstructured intake files to a reviewed matter summary, with the conflict flagged, in under five minutes.
Full transcript
Spoken by Scott Midgley, Chief Information Officer & Co-Founder, Wellforce IT. Lightly edited from the recording for readability; wording is otherwise unchanged. Timestamps link to that moment in the video.
Why matter intake takes so long 0:00
My name is Scott Midgley. I'm the co-founder of Wellforce IT. We're a managed IT and cybersecurity firm working with professional services organizations across the DC metro area and central North Carolina, and law firms make up a big part of that practice. This video is part of our series on Claude Cowork, and today we're getting into the specific work of running a law firm.
Every new matter starts with fragments. An intake email. A phone call. Voice memos dictated between meetings. Conflict notes. Fee and scope details. Before anything can move forward, someone has to turn all of that into a clean, accurate summary.
That first pass can take a lot of time, and it has to be right. Miss a deadline, overlook a conflict, or mishandle a document issue, and the firm may lose the matter or face something worse. Today I want to show you how Claude Cowork can handle the first pass, keep the key details organized, and surface the issues you cannot afford to miss.
What Cowork is, briefly 1:15
If you watched our first video, this will be a quick refresher. Claude Cowork is a mode inside Claude desktop that can read the files you give it, plan a multi-step task, and carry that task through on your machine. Regular Claude chat is a back-and-forth conversation. Cowork is built for you to hand work off and review when the output is ready.
For a law firm, that's the important distinction. You're not copying intake materials into a chat window one at a time. You're pointing Cowork at a folder of real files and asking it to return a usable work product.
We covered the full setup in our first video, so I will keep this quick. Cowork is not available on the web or the mobile app, so you need the Claude desktop app, available on Mac, Windows, or Linux, and you also need a paid Claude plan. Once you are in Claude desktop, look for the mode selector, choose Cowork, and you are ready to go.
The Chen matter scenario 2:30
Let me show you a scenario I built. This is for a fictional firm, Hartwell Bryce LLP, and a fictional prospective client, Sarah Chen, who reached out about a potential employment matter.
In the last 24 hours the firm has generated four separate artifacts: her intake email, the attorney's voice memo dictated after the call, the results of the conflicts check, and the attorney's fee and scope notes.
Under normal circumstances a paralegal is going to have to read all four of these and produce a matter intake summary before the engagement letter goes out. Instead, I am going to have Cowork do it.
Running the intake prompt 3:55
Here is the prompt: "Read through everything in the input folder and produce a clean matter intake summary for the Chen file. Include client information, factual background, potential claims, deadlines, and open items. Flag anything that blocks engagement partner review before we proceed."
Cowork lays out a plan before it does anything. It is going to read each file, then organize the output. For our purposes that transparency matters, because this is client work and I want to see what it intends to do before it does it.
While we wait, notice the progress bar in the upper right corner. It points to the actual folder in use, shown with the folder icon, and it tells us which tools and referenced files were used in the task.
It read all the intake files. It is now analyzing the claims, deadlines, and blockers. Then it will draft the matter intake summary as a Word document, and it ends by verifying the deadline math and the flags.
Something I have noticed with a lot of the customers I consult with on this is that in the beginning they like to watch how it is going, because as Cowork works it shows you what it is doing and the system it follows. That helps people get comfortable before they use it the way it is really intended, where you set it going and work on other things while it works. That is where you really see the time savings.
Reviewing the summary 5:30
Here we go. The Chen matter intake summary is opening now. It lists the firm name, marks it privileged and confidential, and lists Sarah Chen and all of that information at the top.
Look what it caught. The conflicts check flagged Meridian Ventures LLC, a former client of the firm, which is a minority investor in Meridian Stack, Sarah Chen's former employer. That is exactly the kind of thing that is easy to miss when you are skimming under deadline pressure. Cowork puts it at the top of the summary as a blocker, not buried in a list.
It also picked up the EEOC filing deadline math from a date mentioned in passing in the voice memo. It carried forward the note about not letting the client send documents through regular email. It picked up the litigation hold reminder. All from the four unstructured files, in less than five minutes.
Global instructions for firms 7:30
Now I want to show you a feature I want to call out specifically for law firms: global instructions. Click into settings, click Cowork, and there are the global instructions. You can give Claude standing rules that apply to every session.
In my case I use it for format preferences, but for a law firm this is where you enforce your practice-specific guardrails. You can put in things like: never include client-identifying information in file names or task titles; always surface potential conflicts at the top of every summary, like we did here; never make legal judgments, only organize and flag.
The good thing is those instructions apply to every session automatically. After you put them in, you are setting them one time. You do not need to go back and set them again unless you want to change them.
Confidentiality controls 8:50
Now let me address the confidentiality considerations that matter most when working with client information. When Cowork performs actions involving code, it runs those actions inside an isolated virtual machine that is separate from your operating system. Claude can only access the folders you explicitly connect to a session. It does not have unrestricted access to your computer, it does not see unrelated client folders, and it does not connect to your document management system unless you specifically provide that access.
Cowork also includes a permission mode selector. For confidential matters I usually suggest asking before each action. In that mode Cowork pauses and shows you what it intends to do before proceeding. It is a slower workflow, but it is an appropriate trade-off when working with privileged client materials. Regardless of the selected mode, Cowork will always ask for confirmation before it permanently deletes any files.
These controls don't replace the firm's obligations under Rule 1.6, but they do provide a meaningful technical foundation for meeting those obligations. That distinction is important, because many AI tools do not offer the same level of operational control over what the system can access and what it can do.
And that's the walkthrough. If you are going to try Cowork with anything client-related, start with a low-stakes use case: an internal memo, a research folder, or something that is not privileged. Get comfortable with how it behaves before you point it at real client work.
The risk is the rules, not the tool.
Most firms already have people using AI quietly, which is the worst version of this: no policy, no controls, no audit trail. Claude Cowork deployed on Claude Enterprise inverts that: single sign-on, role-based access, audit logs, matter-scoped file access, and no model training on your data by default.
The work it compresses is the work that eats non-billable hours: intake summaries, conflict-check triage, document review, records prep, and routine first drafts, all with a lawyer reviewing and deciding. Our training covers where that line sits, in line with the duties framework of ABA Formal Opinion 512 (we run a plain-English guide to Opinion 512 as well).
Where firms actually put it to work
The starting menu from our legal deployments. Every output goes through a lawyer.
Client intake
Turn intake calls and forms into clean matter summaries and follow-up questions, ready for a lawyer to review.
Conflict checks
Claude scans your matter and party lists to flag possible conflicts for a human to confirm.
Document and contract review
Summarize long documents, pull key clauses and dates, and flag risks across a stack of files.
Research summaries
Summarize case law and long records into plain memos, with your team checking the cites.
Deliverable drafting
First drafts of letters, memos, and routine filings in your firm's voice and format.
Records and discovery prep
Organize and summarize medical records, discovery, and exhibits so review starts ahead.
What makes it defensible
Three things a managing partner should demand from any AI deployment, all standard in ours.
Enterprise controls on
Single sign-on, role-based access, audit logs, and retention controls, configured before the first matter file is touched. A BAA where the work calls for one.
A policy that fits your duties
Confidentiality, supervision, and disclosure written into a plain-language AI use policy. Claude drafts and reviews; your lawyers decide and remain responsible.
An audit trail partners can check
Who used it, on what, when. We verify the audit trail at day 30 and review access monthly under optional Managed Claude.
The full deployment package, timeline, and pricing (from $3,500 flat): Claude for Legal & Professional Services. Firm-wide IT and security is covered by our law firm practice.
Legal questions, answered
How do law firms use Claude Cowork?
Is Claude Cowork confidential enough for client data?
Does using Claude Cowork fit ABA guidance on AI?
Will it give legal advice or replace associates?
What does a legal deployment cost and how long does it take?
We install Claude Cowork for teams in DC & Raleigh
Design, setup, connected tools, hands-on training, and a one-page AI policy. Live in two weeks, from $2,500 flat. Nonprofit pricing available.
Or just tell us what you're trying to do — we'll point you the right way, free.
More Claude Cowork guides
What is Claude Cowork? The plain-English guide
What it is, what it actually does, who it is for, and how teams use it.
Read the guide →Claude Cowork pricing
What Claude Cowork costs, plan by plan, plus what a real team rollout runs.
Read the guide →How to use Claude Cowork
Step-by-step setup and a first-week playbook for business teams.
Read the guide →Claude Cowork vs Claude Code
Two Anthropic agent products, two different jobs. Which one fits yours.
Read the guide →Claude Cowork vs OpenClaw
Managed knowledge-work agent or self-hosted personal assistant: the practical choice.
Read the guide →Is Claude Cowork free?
The direct answer, what each paid plan includes, and what can cost extra.
Read the guide →Claude Cowork use cases
Eighteen real workflows, by team type: nonprofits, legal, and growing businesses.
Read the guide →Claude Cowork vs Microsoft Copilot
An honest comparison for SMB stacks, from a Microsoft shop that deploys both.
Read the guide →Claude Cowork security & data privacy
What it can touch, what it cannot, and the controls that make it safe for real data.
Read the guide →Claude Cowork for nonprofits
Grants, donor emails, and board packs: the nonprofit playbook.
Read the guide →Legal intake automation, step by step
Four intake files to a reviewed matter summary, with the conflict surfaced, in under five minutes.
Read the guide →Claude Cowork for small business
Official workflows, Team plan costs, and a rollout that prevents AI shelfware.
Read the guide →Claude Cowork Dispatch
Assigning desktop tasks from your phone: what it needs, which plans have it, and the trade-off.
Read the guide →Claude Cowork change log
A dated record of Cowork releases, plan changes, platform support, and fact checks.
Read the guide →Claude Cowork plugins & Skills
Skills, connectors, and plugins: what each one is, when to build, and how to govern them.
Read the guide →Claude consulting for SMBs
Implementation, training, and managed Claude operations in DC & Raleigh. Live in two weeks, from $2,500 flat.
See the service →