Conflicts Screening With AI in a NJ Small Firm: 5 Things to Check Before You Trust the Output

Conflicts screening is one of those intake tasks that feels administrative right up until it isn't. Miss a conflict, and you're looking at a disciplinary complaint, a fee forfeiture, or disqualification mid-case. For NJ solo attorneys and small firms running lean, the appeal of letting an AI tool handle the initial screen is obvious. Feed in the new client's name, run it against your matter history, get a green light. Done.

Except it's not quite that simple, and a number of attorneys are learning that the hard way by building intake workflows around AI conflict-checkers that have meaningful blind spots.

Here's what you actually need to verify before an AI-assisted conflicts screen is something you'd stake your bar card on.

1. Does the Tool Search Phonetic and Name Variant Matches?

AI conflict-checkers that rely on exact or near-exact string matching will miss "Mohammed Al-Rashid" if your file is saved under "Mohammad Alrashid." They'll miss married name changes, name inversions, and DBA entities. Before you trust any tool with your conflict workflow, run a few deliberate test searches using known variants from your existing files. If the results differ when you change spelling, that's a gap you need to close with a manual second pass.

This isn't hypothetical. NJ RPC 1.7 requires an actual conflicts check, not a best-effort one. The mechanics of how you perform that check aren't prescribed, but "the AI said it was fine" is not going to satisfy a grievance committee if a real conflict slipped through because the software couldn't handle a hyphenated surname.

2. Is Your Matter History Actually Complete in the System?

AI conflict tools are only as good as the data you've given them. If you migrated to a new practice management platform in 2021 and didn't import all legacy matters, your pre-migration clients don't exist as far as the tool is concerned. Same goes for matters you handled as an associate before going solo, former clients you added manually to a spreadsheet but never entered into the system, and one-off consultations where you created a paper file but no digital record.

Before you treat AI output as authoritative, map out where your client and matter data actually lives. A conflict screen that only covers 80% of your history is worse than useless because it creates a false sense of completeness.

3. Are Entity Relationships and Principals Captured?

A new prospective client is "Greenway Holdings LLC." Your AI tool clears it. But one of Greenway's principals is a former adverse party from a 2019 commercial dispute you handled. If you haven't structured your matter records to capture principals, guarantors, related entities, and key individuals, the tool can't flag what it doesn't know.

This is where AI-assisted screening breaks down most often for small firms. Large firm software like Intapp or Closing Folders is purpose-built to map entity relationships. Many of the lighter tools marketed to solos are not. Check whether the tool you're using lets you add related parties to each matter record, and whether it actually searches those fields.

4. Is There a Human Review Step Before You Formally Clear the Conflict?

Even if you're confident in the tool, the professional obligation sits with you personally. NJ RPC 1.9 requires that you not represent a new client in a matter materially adverse to a former client in the same or substantially related matter. "Substantially related" requires legal judgment, not a keyword match.

A reasonable workflow looks something like this: run the AI screen first to catch obvious hits and reduce your manual review time, then have a lawyer review any flagged results and sign off on any cleared matter. Document both steps in the file. If you're solo with no second set of eyes, that final review is still yours to do, but you've at least narrowed the field.

The AI screen is a first pass, not a final answer.

5. What Happens When a Conflict Surfaces Mid-Matter?

Most conflict-screening conversations focus on intake, but the more complicated scenario is the one that develops after the engagement starts. A new adverse party is added to litigation. Your client tells you mid-representation that their business partner is someone you represented three years ago. Opposing counsel files a motion to disqualify.

AI tools generally don't run continuous background checks on active matters. That's not what they're built for. You need a protocol for what happens when new names enter a matter, including a defined step where someone reruns a conflicts check. For a solo, that protocol is a calendar reminder or a checklist item in your case management software. It doesn't have to be sophisticated, but it has to exist.

---

The NJ Supreme Court's Advisory Committee on Professional Ethics has not issued specific guidance on AI-assisted conflicts screening as of this writing, but the underlying RPC obligations haven't changed. What's changed is the risk of over-relying on a tool that gives you a false green light faster than any manual process ever could.

If you're going to use AI for intake conflicts screening, treat it as a filter, not a clearance. Run it, document it, and then do the legal analysis yourself.