Is AI Conflicts Screening Reliable Enough for a New Jersey Solo Practice?

Conflicts screening is one of those tasks that feels administrative until it isn't. For a solo attorney in New Jersey, a missed conflict doesn't just create an uncomfortable conversation, it can mean disqualification, a grievance, and in bad cases, malpractice exposure. So when practice management platforms started advertising AI-assisted conflicts checking, it was only a matter of time before NJ solos started asking whether this was actually worth trusting.

The short answer: it depends heavily on what your data looks like, and most attorneys haven't thought carefully about that.

What AI Conflicts Tools Actually Do

Most AI-assisted conflicts features in platforms like Clio, MyCase, or Smokeball don't run some sophisticated reasoning engine. What they typically do is fuzzy-name matching across your matter database, combined with relationship mapping, pulling in party names, opposing counsel, known affiliates, and sometimes corporate structures if you've connected external data sources.

That's genuinely useful. A human running a manual search through a spreadsheet of past clients will miss a maiden name variant or a business entity connected to a former client. A well-configured conflicts tool catches that. But "well-configured" is doing a lot of work in that sentence.

The system only knows what you've put into it. If your intake process didn't consistently capture adverse parties, related entities, or guarantors on matters from three years ago, the AI has nothing to match against. Garbage in, garbage out, and for conflicts, incomplete in is almost as bad.

Where RPC 1.7 and 1.9 Require More Than a Clean Search Result

Under New Jersey RPC 1.7, a current conflict exists when representation of a client is directly adverse to another client, or when there's a significant risk that the representation will be materially limited by the attorney's responsibilities to someone else. RPC 1.9 extends that analysis to former clients.

Neither of those rules is purely a name-matching exercise. The question of whether a "significant risk" of material limitation exists requires judgment about the nature of prior representations, the subject matter overlap, and the confidential information you actually received. An AI conflicts tool can surface a flag. It cannot tell you whether the flagged relationship rises to the level of an actual conflict under NJ's standards.

This matters because some solos are treating a clean conflicts search as a conflicts clearance. Those are not the same thing. The search is step one. The attorney still has to evaluate what the search returns, and evaluate the absence of a result with appropriate skepticism.

The Specific Scenarios AI Struggles With

A few conflict patterns that automated tools handle poorly:

Organizational clients with complex structures. If you represented an LLC and are now being asked to sue its sole member individually, a basic name-match won't flag that unless you've manually linked the entities in your system. Some platforms let you build organizational maps, but few solo attorneys take the time to do it during intake.

Positional conflicts. If you've argued one legal position in a brief for Client A, you may have a conflict taking the opposite position in an unrelated matter for Client B, not because the clients are adverse, but because your advocacy could undermine a legal theory you've already advanced. No conflicts tool screens for this, because it requires reading what you wrote and understanding the legal argument. That's human work.

Prospective client consultations that didn't become matters. Did you log that 20-minute phone call where someone described their business dispute before you declined the matter? If it's not in the system, it doesn't exist for conflicts purposes, even if you received enough information during that call to be disqualified later.

What to Actually Check Before You Trust the Workflow

If you're using or considering AI-assisted conflicts checking, run through this before treating it as reliable:

First, audit one year of closed matters and verify that adverse parties, related entities, and referral sources were captured consistently. If you find gaps, that's your data quality problem, not the tool's.

Second, check whether your platform lets you flag prospective client consultations as a distinct record type, separate from opened matters. If it doesn't, you need a manual process running alongside the automated one.

Third, run a few test searches on matters you know well: former clients with married names, business entities that share an owner, individuals who were adverse parties but not your direct clients. See what the tool surfaces and what it misses. That stress test will tell you more about the system's actual reliability than any vendor demo.

Finally, document your conflicts process. Under RPC 1.7, the adequacy of a conflicts check can become relevant in a discipline proceeding. "I ran it through Clio" is a thinner record than a written protocol describing what sources you searched, who reviewed the results, and what analysis you applied.

The tools are improving, and for a solo attorney without a dedicated conflicts clerk, AI-assisted screening is better than a manual spreadsheet search done at the end of a busy intake appointment. But "better than the alternative" isn't the same as sufficient. Keep the judgment in-house, and build a system that can prove it.