Conflicts Screening With AI in a NJ Small Firm: Can You Actually Trust It?

Conflicts screening is one of those tasks that feels administrative right up until it becomes a malpractice claim or a bar grievance. For NJ solo attorneys and small firms, the promise of AI-assisted conflicts checking is genuinely appealing: faster intake, fewer missed connections, less time combing through old matter files. Several practice management platforms now advertise AI-powered conflicts search as a core feature. Some attorneys are also running intake data through general-purpose tools to spot potential conflicts before a new matter opens.

The question worth asking honestly is: can you actually trust any of this, and what does "trusting it" even require under New Jersey's professional responsibility framework?

What AI Conflicts Tools Actually Do

Most AI-powered conflicts features inside practice management software (Clio, MyCase, Smokeball, and others) work by fuzzy-matching new client and matter data against your existing database of clients, adverse parties, and related entities. The "AI" label often describes fairly straightforward string-matching with some tolerance for typos, name variations, and related-entity inference. A few newer tools use embeddings or semantic search to catch conceptual matches, such as flagging that "ABC Holdings LLC" and "ABC Real Estate Partners" might be the same beneficial owner.

That's genuinely useful. Manual searches miss things. Attorneys forget that a client from 2019 had a spouse named the same as the adverse party in a new matter. A well-configured AI-assisted search catches some of those.

But the ceiling on what these tools can catch is set entirely by what's in your database to begin with.

The Database Problem Nobody Talks About

Here's the core issue: AI conflicts tools are only as good as your intake records. If you've been in practice for ten years and your early matters were tracked in a spreadsheet, a paper file, or a prior firm's system you no longer have access to, those matters simply don't exist for your AI conflicts tool. It will return a clean result. That result is not reliable.

The same problem appears with adverse parties. Most solo practitioners do a thorough job capturing client names but a much sloppier job capturing the full name of every adverse party, related entity, guarantor, corporate parent, or beneficiary from each matter. If those names aren't in your system, the AI won't flag them.

NJ RPC 1.7 and 1.9 require you to identify actual conflicts, not just the ones a software tool surfaces. "The AI said there was no conflict" is not a defense recognized anywhere in New Jersey ethics jurisprudence, and the ACPE hasn't issued any opinion suggesting it would be. The duty to screen belongs to the attorney.

Where Small Firms Are Most Exposed

Small firms face two specific risks that larger firms can paper over with process.

First, the two-attorney firm where each partner runs their own intake is a common setup. If conflicts data isn't centralized, attorney A's check won't surface attorney B's matters, and vice versa. AI tools accelerate the search of a shared database, but they can't substitute for the database being complete and actually shared.

Second, transactional practices carry a particular blind spot. Litigation matters tend to generate clear adverse-party records because the other side's name appears in every filing. Transactional work is messier. A real estate closing might involve a buyer, a seller, a lender, a title company, and several guarantors. If intake only captures the client and the counterparty, and the AI only searches those fields, you could miss a conflict with someone who was a non-client third party in a prior deal.

What a Defensible Conflicts Workflow Looks Like

A few things that actually reduce your exposure, regardless of which tool you're using:

Audit your database first. Before trusting AI-assisted conflicts screening for new matters, spend time cleaning up what's already in your system. Add adverse parties retroactively where you can. If you migrated from another platform, verify that the migration captured everything.

Run the search on multiple name variations manually. The AI's fuzzy match is good but not exhaustive. Search the new client's name, any known aliases, the entity's registered agent if it's a company, and any individual with a controlling interest. Do the same for the adverse party.

Document the search you ran. Not just the result, but the terms you searched, the date, and the tool you used. If a conflict is later alleged, your file should show a reasonable process, not just a clean output from a button you clicked.

Keep a running adverse-party log outside your matter management software. A simple spreadsheet updated at every new matter opening, capturing every adverse party and key related entity, gives you a fallback data source that doesn't depend on any one vendor's import quality.

The AI tools will keep improving. Some of the newer platforms are meaningfully better at entity resolution than they were two years ago. But the workflow discipline required to make those tools reliable hasn't changed. For NJ solo and small-firm attorneys, the safest posture is to treat AI-assisted conflicts screening as a first pass that makes your manual review faster, not one that replaces it.

The ACPE's existing opinions on conflicts are clear that the duty is substantive, not procedural. Clicking a button satisfies neither.