Short answer: use AI to gather the conflicts data, not to decide it. Let it search your own matter list, contacts, and closed files for every name that touches a new matter. Then a lawyer reads what it found and makes the call. Nothing gets "auto-cleared."
This sits right behind intake. A good intake call captures the story. The conflicts check decides whether you can take it at all. And like client confidentiality, it depends on what the tool can see and where that data lives.
Why the check is harder than it looks
Most small firms run conflicts with a name search in the practice management system and somebody's memory. That works until it does not:
- The new client is "Bob Miller." The old file says "Robert J. Miller, Jr."
- The adverse party is a company. Your firm represented its owner personally five years ago.
- A spouse, a business partner, or an insurer shows up in an old email but never made it into the contacts table.
- The lawyer who handled the matter retired, and the memory went with them.
The rule behind all of this is plain. Under ABA Model Rule 1.7, a lawyer may not take a representation that is directly adverse to another client, or that is materially limited by duties to someone else, unless the rule's conditions and informed consent are met. Former clients and imputed conflicts have their own rules. Your state's version is the one that binds you. Read it.
What AI may do
- Pull every name from the intake. The prospective client, the other side, spouses, companies, owners, insurers, witnesses who matter.
- Search your own records for each one. Open matters, closed matters, contacts, billing records, and old correspondence.
- Catch near-matches. Nicknames, maiden names, middle initials, a company and its former name.
- Show its work. For every hit, the record it came from and why it looks related.
- Draft the conflicts memo for the responsible lawyer: what was searched, what was found, what was not found.
That last point matters. "No hits" is only useful if you know what was searched.
What it must never do
- Clear a conflict. Deciding whether a hit is a real conflict, a waivable one, or nothing is legal judgment. It stays with a lawyer.
- Tell the prospective client anything. No "you're all set," no "we can't help you because we represent your landlord." What you say, and when, is the lawyer's decision.
- Draft a waiver on its own initiative. If a waiver is on the table, the lawyer decides that first.
- Search outside your systems. The check runs against the firm's own records. Do not send names of prospective clients to a public chatbot to "see what it knows."
How the review should work
- Intake finishes. The system builds the name list and runs the search.
- The memo goes to the responsible lawyer, with every hit and its source.
- The lawyer marks each hit: not related, related but no conflict, or conflict to resolve.
- The lawyer signs off before any engagement letter goes out.
- The memo and the sign-off stay in the file.
If your firm has a conflicts partner or a set rule for who signs, keep it. The tool changes how fast the list arrives, not who is responsible for it.
Questions to ask before you turn it on
- Which records does it actually search? Contacts only, or also closed files and email?
- Does it show the source for every hit? A hit you cannot trace is a hit you cannot judge.
- Can it mark anything "clear" by itself? You want no.
- Where does the data go during the search? Conflicts data is confidential. Ask whether it can run on hardware in your office.
A realistic first step
Take five recent intakes where you know the history. Run them through the tool. Have the lawyer who knows those clients read the memos. Did it find the related party you remember? Did it flag things that do not matter? One afternoon of that tells you more than a demo.
If you want a conflicts setup that gathers from your own records and leaves the decision with your lawyers, talk with us. If a cleaner contacts table is enough, we will say so.
Related: AI intake for law firms · our approach for law firms.