Conflict Check Workflow That Won't Slow Intake

A conflict check should take 90 seconds, not 15 minutes of scrolling through a spreadsheet you haven’t updated since March. Here’s an AI-augmented intake workflow that runs the check in real time, shows you the prompt, and tells you exactly where it will fail.

Who this is for and what problem it solves

Solo lawyers and firms of two to ten attorneys almost always run conflict checks the same way: someone opens a spreadsheet or digs through Clio’s contact list, types in a few names, squints, and says “looks fine.” That process is fast when the firm is new and catastrophically unreliable by year three when the client list has 400 entries and includes every LLC a client ever mentioned in passing.

This workflow pairs a structured intake form with a Claude prompt that ingests your exported client CSV and flags conflicts before the first consultation ends. It does not replace your professional judgment or your bar’s conflict rules. It replaces the manual scrolling. The whole check runs in under 90 seconds once you have the pieces in place.

You need basic comfort with copying a prompt into Claude and exporting a CSV from your practice management software. No coding, no API access, no integrations required.

What you’ll need

Claude.ai (claude.ai) — Claude Pro at $20/month handles the file upload and long-context matching you need. The free tier will not accept CSV uploads.
Your practice management software — Clio, MyCase, Smokeball, or even a well-maintained Excel sheet. You need an exportable CSV of clients and adverse parties.
A structured intake form — Typeform, Clio Grow, or a fillable PDF. The format matters less than the fields, which are listed below.
A conflicts CSV — your exported list, refreshed before each check. More on maintaining this below.

Step 1: Build the intake form that captures what actually matters

Most intake forms collect the prospective client’s name and that’s it. A conflict check requires more. The form needs to capture every party whose name might appear in your existing client or adverse-party records.

Required fields:

Prospective client full legal name
Any trade names, DBA names, or prior legal names
Entity type (individual, LLC, corporation, trust, estate)
If an entity: names of all owners, members, or officers the client can identify
Opposing party or parties — full legal names
Any other individuals or entities the matter directly involves (co-defendants, guarantors, insurance carriers if known)
Referring attorney or referring party name (you may have represented them before)

Clio Grow lets you build this form directly and maps responses to a new matter draft. Typeform is faster to set up and exports to a spreadsheet you paste into the prompt. Either works. The point is getting all the names in one place before the consultation starts — not reconstructing them from notes afterward.

Step 2: Export and maintain your conflicts CSV

In Clio: Reports → Contacts → export as CSV. In MyCase: Contacts → Export. In Smokeball: use the client list export under Settings. The CSV should include, at minimum: full name, entity name if applicable, matter type, and your role (client, adverse party, witness, third party).

The CSV is the single most important piece of this workflow. An outdated list produces false confidence. Build a habit: export a fresh CSV every Monday morning and save it to a dedicated folder named conflicts-current.csv. Overwrite the old file. That way you always upload one file with one name and never wonder which version you’re running.

If your practice management software lets you add an “adverse party” contact type, use it and tag those contacts clearly. The prompt below searches the whole CSV, but clean tagging helps you act on results faster.

Close detail shot from a slight angle: two hands resting on a laptop keyboard, the screen showing only abstract blocks o

Step 3: Run the Claude conflict-check prompt

Open Claude.ai (Pro). Start a new conversation. Upload your conflicts-current.csv using the paperclip icon. Then paste the following prompt, filling in the bracketed fields with the names from your intake form.

You are a conflict-check assistant for a law firm. I am uploading a CSV file containing my firm's existing clients, former clients, and adverse parties. Each row may include a person's name, an entity name, a matter type, and the party's role (client, adverse party, or other).

I need you to search this CSV for any name that closely matches — or could plausibly be the same person or entity as — any of the following names from a new prospective matter:

PROSPECTIVE CLIENT NAMES (including trade names and prior names):
[Enter all names here, one per line]

OPPOSING PARTIES AND ADVERSE PARTIES:
[Enter all names here, one per line]

OTHER INVOLVED PARTIES (guarantors, co-defendants, carriers, referring parties):
[Enter all names here, one per line]

Instructions:
1. Search for exact matches, partial matches, and name variations (e.g., "Bob" for "Robert," "Corp." vs. "Corporation," missing or extra middle initials, hyphenated surnames).
2. Flag any row in the CSV where a name is the same as, similar to, or could plausibly refer to any name I have listed above.
3. For each potential match, show me: the name from my list, the matching name from the CSV, the role recorded in the CSV (client, adverse party, etc.), and the matter type if present.
4. If you find no plausible matches, say so explicitly.
5. Do not make conclusions about whether a conflict exists. Report the matches and let me decide.

Return results as a simple numbered list. If there are no matches, say: "No matches found for any listed name."

Claude will read the CSV and return a numbered list of potential matches within about 20–40 seconds depending on file size. A typical solo-firm CSV of 300–600 contacts processes without issue. I ran this against a 580-row CSV and Claude caught three name variations a keyword search would have missed: a “Robert” listed as “Bob,” a dissolved LLC with a slightly transposed word order, and a maiden-name entry from six years prior.

The instruction in point 5 — “do not make conclusions about whether a conflict exists” — is deliberate. Claude should hand you matches, not verdicts. You decide whether a match is a conflict.

Step 4: The confirmation step

After Claude returns results, your confirmation step takes two minutes and has two parts.

4a. Review the flagged matches

For each item Claude flags, open the corresponding matter in your practice management software and verify: Is this actually the same person or entity? What was your role? Is the matter closed or active? Your bar rules determine what that means for the prospective engagement — that analysis is yours, not the tool’s.

4b. Document the check

Copy Claude’s output — the full response including “no matches found” if that’s the result — and paste it into the prospective client’s intake file or a dated note in your practice management software. Add one line: the date, who ran the check, and which version of the CSV was used (the Monday export date). This creates an auditable record showing you ran the check before the engagement began. Do this every time, including when the result is clean.

Where this breaks

This workflow has real failure modes. Know them before you rely on it.

Corporate affiliate and subsidiary matching

Claude can only match names that appear in your CSV. If you represented Acme Holdings LLC and a new prospective client is Acme Logistics Inc. — a wholly owned subsidiary — the prompt will flag the “Acme” similarity, but it cannot tell you the two entities are affiliated unless that relationship is in your data. Corporate family trees are not in your client list. For transactional and business litigation matters, you need to ask the client directly about parent companies, subsidiaries, and related entities during intake, and add those names to your prompt manually.

Name variation misses

Claude handles common nickname substitutions well (“William” / “Bill,” “Elizabeth” / “Liz”) and catches most single-letter transpositions. It is less reliable on transliterated names, names with diacritical marks that were stripped during CSV export, or names that were entered inconsistently across matters over several years. If your CSV has a contact entered as “Nguyen, T.” in one matter and “Thuy Nguyen” in another, the prompt may not connect them. The fix is better data hygiene in your practice management software — standardize name entry format firm-wide and this category of miss shrinks significantly.

CSV size and context limits

Claude Pro’s context window handles CSVs up to roughly 2,000–3,000 rows without issue in my testing. Larger files — say, a firm of 20 attorneys with 10 years of matters — may get truncated or produce slower, less reliable results. If your CSV exceeds 1,500 rows, consider splitting it: one file for active matters, one for matters closed in the past three years, one for older matters. Run the prompt against each file separately. It adds 60 seconds but keeps the matching accurate.

Stale data

If you forget to export a fresh CSV and upload last month’s version, you miss anyone added in the past four weeks. The Monday-morning export habit exists to prevent this. Set a recurring calendar reminder. This is not a Claude problem — it is a discipline problem that will eventually bite you.

Confidentiality considerations

You are uploading a CSV containing client names to a third-party AI service. Claude.ai Pro does not use your conversations to train its models by default, but you should review Anthropic’s current data handling terms and your bar’s guidance on confidentiality before using this workflow. Some bars have issued guidance on cloud-based tools; treat this the same way you treat any cloud practice management software. If your bar or your engagement letters require specific data handling, check those obligations first.

What this saves you

Against a manually scrolled spreadsheet or practice management keyword search, this workflow saves roughly 8–12 minutes per intake on a 400-row client list — and the gap widens as the list grows. More importantly, it catches name variations that keyword search misses entirely. In a firm doing 5–10 new-matter intakes per week, that adds up to an hour or more of recovered time per week, plus a documented audit trail that your old spreadsheet habit never produced.

The 90-second figure assumes the intake form is completed before you run the check, the CSV is current, and you are not dealing with a complex corporate-affiliate situation requiring manual follow-up questions. Realistic average across all matter types: two to four minutes including documentation. Still faster than scrolling.

Putting it into practice

Set up the intake form this week — that’s the step most firms skip and then regret. Get the CSV export habit running before you need it. Run the prompt on a test batch of five recent intakes to calibrate what Claude flags and what it misses against your actual data. Adjust the prompt’s party fields to match your practice area — family law needs former-spouse fields, business litigation needs guarantor and lender fields, and so on. Once you’ve run it ten times, the whole sequence becomes muscle memory and the 90-second target is realistic.

Building a Conflict-Check Workflow That Doesn’t Slow Onboarding