90-Second Conflict Check During Intake

Running a conflict check after the intake call is how you get a ten-minute callback telling a prospect “sorry, we can’t help you.” This workflow moves that check to the first ninety seconds of the call itself.

This is for solo attorneys and firms of two to ten lawyers who keep their client records in Clio, MyCase, or Smokeball and field new-matter intake calls personally or through a paralegal. The tooling is Claude (claude.ai or API), a fresh CSV export of your client/matter list, and a browser tab you can glance at while you talk. No third-party conflict software required. The whole loop — prospect names in, confidence rating out — runs in under ninety seconds once you have the pieces in place.

What You’ll Need

Claude Pro or API access — claude.ai ($20/month) works fine for low-volume intake; API is better if you want to script this later. GPT-4o also handles the prompt below, but Claude’s instruction-following on structured output is more consistent in my testing.
A CSV export of your client/matter list — Clio: Reports → Matters → export all fields. MyCase: Settings → Export Data → Contacts. Smokeball: Contacts → Export. Aim for columns: client name, opposing party (if captured), matter type, matter status, responsible attorney.
A shared or pinned browser tab — open Claude in a second monitor or a split-screen window you can reach while on the call. A tablet propped beside your primary screen also works.
Your intake form — paper, Clio Grow, or a basic Google Form. You need the prospect’s full name, any entity name, and the names of the opposing or adverse parties before you run the check.

Step 1: Export and Prep the CSV Before the Call Week

Refresh your CSV export every Monday morning — or after any matter opening or closing. Stale data is the single biggest failure point in this workflow. A conflict check against a list that’s two months old is not a conflict check.

Clean the CSV before you save it. Remove columns you don’t need (billing rate, trust balance, invoice history). Keep: client name, opposing party name, matter status (open/closed/pending), responsible attorney, and matter number. Shorter file means faster paste into the prompt window and fewer tokens consumed if you’re on API.

Save the cleaned file somewhere you can open and copy from in three seconds — desktop shortcut, pinned in Finder/File Explorer, or a TextExpander snippet if your list is small enough to fit. The goal is zero fumbling when a call comes in.

Step 2: Collect the Names in the First Sixty Seconds of the Call

Before you can run anything, you need names. Train yourself to ask for them early — before the prospect tells you their whole story. A simple script: “Before we go further, I need to grab a few names for our records — your full legal name, any business entities involved, and the name of the other party or parties.”

Type the names directly into your intake form as the prospect speaks. Don’t wait until after — the typing is the capture step. If the entity name is ambiguous (“my LLC” or “the company”), ask for the full registered name. LLC name variations are where fuzzy matching earns its keep, but “the company” is not matchable against anything.

Note every party name you hear: the prospect, their spouse or business partner if mentioned, the opposing individual, any named entity on the other side. More names fed into the prompt means fewer missed matches.

Step 3: Build and Run the Prompt

Once you have the names — usually ninety seconds to two minutes into the call — tab to Claude. Paste your CSV data into the prompt first, then append the names block and the instruction below. The full prompt structure looks like this:

You are a conflict-check assistant for a law firm. Below is the firm's current client and matter list in CSV format. After the CSV, I will give you a list of names from a prospective new client intake call. Your job is to compare the intake names against every name in the CSV — client names AND opposing party names — and return a confidence-rated match report.

MATCHING RULES:
- Flag exact matches as: DEFINITE CONFLICT
- Flag near-matches (name variations, abbreviations, LLC vs Inc vs Ltd, first-name-only matches on uncommon names, maiden vs married name patterns) as: POSSIBLE CONFLICT — explain why
- Flag no match found as: CLEAR
- Check individuals against entity names and vice versa (e.g., "James Whitfield" should flag if "Whitfield Holdings LLC" appears in the list)
- Ignore punctuation differences, capitalization, and common abbreviations (Co., Corp., & vs and)
- If a matter status is "closed," still flag it but note it as CLOSED MATTER — attorney review required

OUTPUT FORMAT:
For each intake name, return one line:
[NAME] → [DEFINITE CONFLICT / POSSIBLE CONFLICT / CLEAR] — [one-sentence reason or "no match found"]

Then after all names are processed, return a single OVERALL ASSESSMENT line:
OVERALL: [DEFINITE CONFLICT / POSSIBLE CONFLICT / CLEAR] — [one sentence]

Do not give legal advice. Do not tell me whether I can take the matter. Only report matches.

---CSV DATA START---
[PASTE YOUR CSV HERE]
---CSV DATA END---

INTAKE NAMES TO CHECK:
Prospect (client): [full name]
Prospect entity: [entity name or "none"]
Opposing party individual: [full name or "none"]
Opposing party entity: [entity name or "none"]
Additional parties mentioned: [names or "none"]

Fill in the bracketed fields from your intake form. The paste-and-run takes about fifteen seconds. Claude’s response comes back in another fifteen to thirty seconds depending on CSV length. You’re looking at a forty-five to sixty second turnaround from the moment you tab to the browser.

While Claude processes, keep the prospect talking — ask about their timeline or how they found the firm. You need maybe thirty seconds of filler. Most callers fill it themselves.

A close-up detail shot of two hands resting on a laptop keyboard, a legal notepad with handwritten names visible only as

Step 4: Read the Output and Decide in Real Time

The output gives you one of three lanes. CLEAR means proceed — finish intake, schedule the consult, or open the matter per your normal process. POSSIBLE CONFLICT means pause the call briefly: “Let me put you on hold for just a moment while I pull up a file.” Then check the flagged matter manually before continuing. DEFINITE CONFLICT means stop intake, take contact info only, and follow your firm’s conflict-closing procedure before any further conversation.

Do not read the AI’s output verbatim to the prospect. The output is an internal triage signal. You make the call on whether a conflict actually exists — Claude is pattern-matching names against a list, not applying conflict-of-interest law to your jurisdiction or your professional responsibility rules.

Log the result. A one-line note in the matter record or intake form — “conflict check run [date], result: CLEAR, CSV dated [date]” — gives you a paper trail. If you’re ever questioned on your conflict-check process, that log is your evidence it happened.

Step 5: Handle the POSSIBLE CONFLICT Cases

POSSIBLE CONFLICT outputs need a thirty-second manual look, not a full-stop conflict closure. Claude will tell you why it flagged the name — “Whitfield LLC vs. Whitfield Holdings LLC, partial entity name match” — and that reason tells you where to look. Pull up the flagged matter in Clio or MyCase right then, check the actual parties, and make the call yourself.

If the matter is closed and the flag is on a closed matter, your jurisdiction’s rules on former-client conflicts apply. That’s your analysis to do, not the AI’s. The prompt surfaces the record; you determine whether representation is permissible.

Common false-positive patterns I’ve seen: very common surnames (Johnson, Patel, Williams) triggering on unrelated clients; national chain businesses (State Farm, Enterprise) appearing on both sides across unrelated matters; and hyphenated names where the CSV only captured one half. All of these are manageable with a quick look. They are not reasons to abandon the workflow — they are reasons to train your eye on what a POSSIBLE flag actually means.

Where This Breaks

Complex corporate affiliations. If Opposing Party A is a wholly-owned subsidiary of Company B, and you represented Company B two years ago, the prompt will only catch this if both names appear somewhere in your CSV. Claude cannot infer corporate ownership chains from a name alone. If your practice touches M&A, PE-backed entities, or any matter where parent/subsidiary relationships matter, this workflow is a first-pass screen only — not a complete check.

Recent acquisitions and name changes. If a client’s business was acquired last month and is now operating under a new name, your CSV has the old name. The prompt cannot match what isn’t there. Keep a short separate log of known entity name changes if you practice in areas where this comes up.

Dropped or purged matter records. If your firm deletes closed matters from the practice management system (some firms archive off-platform), those names aren’t in the export. The CSV is only as complete as your data hygiene. If you’ve had records in multiple systems — a legacy system before switching to Clio, for example — your exported list may have gaps.

Real-time data lag. A matter opened this morning by another attorney at your firm may not be in Monday’s CSV export. For firms with active concurrent intake across multiple attorneys, this workflow needs a same-day export protocol or a shared Google Sheet updated on matter open — not a weekly refresh.

Very large CSV files. If your firm has thousands of matters, pasting the full CSV into Claude’s context window every call is slow and will eventually hit token limits. The fix is to use the API with a pre-loaded system prompt containing the CSV, or to run a local script that pre-filters the list to names starting with the same first letter before pasting. This is a solvable problem, but it requires a bit more setup than the basic workflow above.

What This Saves You

The obvious win is time: the post-call conflict check that used to take five to fifteen minutes — pull the list, scan it, run it again with a name variant — now happens during the call’s natural dead air. You’re not adding a step; you’re moving an existing step earlier and compressing it.

The less obvious win is prospect experience. Finding a conflict before you’ve had a thirty-minute consult — or worse, before you’ve taken a retainer — saves everyone a difficult conversation. The prospect gets declined in two minutes instead of two days. That’s a better outcome for them and a cleaner process for you.

For a solo running fifteen to twenty intake calls per month, moving to in-call conflict checking eliminates roughly one to two hours of post-call administrative work per month and cuts the average time-to-conflict-discovery from hours to under two minutes. For a firm of five attorneys with higher intake volume, the time savings per week become meaningful fast.

The workflow takes about thirty minutes to set up the first time — export, clean the CSV, save the prompt template, test it against two or three known matters to verify the matching behavior. After that, the marginal cost per intake call is ninety seconds and a browser tab. That’s a trade worth making.

One firm discipline this requires: actually refreshing the CSV. The workflow is only as current as the file you feed it. Build the Monday export into a calendar reminder or a paralegal’s standing checklist, and treat a stale export the same way you’d treat a broken process — because that’s what it is.

The 90-Second Conflict Check Workflow for Live Intake Calls