Compliance and Privacy11 min read

AI Legal Intake: Unauthorized Practice of Law

Can AI do legal intake without the unauthorized practice of law? A Nevada script table of 12 allowed and prohibited lines tied to NRS 7.285 and NRPC 5.3.

Can AI do legal intake at a Nevada law firm without the unauthorized practice of law? Yes, if the agent collects facts and routes them, and never tells the caller what those facts mean. Under NRS 7.285 in Chapter 7 of the Nevada Revised Statutes, unlawful practice is a misdemeanor on a first offense and a category E felony on a third within seven years, and Rule 5.3 of the Nevada Rules of Professional Conduct makes supervising lawyers answerable for nonlawyer assistants. ABA Formal Opinion 512, issued July 29, 2024, applies that supervision duty to generative AI. This is general information, not legal advice.

Below is a script boundary any firm can adopt: twelve paired intake lines, an escalation ladder, and a monthly transcript audit a supervising lawyer can actually finish.

NRS 7.285 regulates what is said, not who or what says it

The statute prohibits the practice of law by a person who "Is not an active member of the State Bar of Nevada or otherwise authorized to practice law in this state pursuant to the rules of the Supreme Court," or who "Is suspended or has been disbarred from membership in the State Bar of Nevada." Penalties escalate on a seven year lookback: a misdemeanor first, a gross misdemeanor for a second offense within seven years, and a "category E felony and shall be punished as provided in NRS 193.130" for a third or later one.

The criminal ladder is not the only lever. The statute also says "The State Bar of Nevada may bring a civil action to secure an injunction and any other appropriate relief against a person who violates this section." An injunction aimed at an intake channel would be a business problem long before it was a criminal one.

Notice what the statute does not do. It does not name a safe job title. It regulates conduct. So the useful question for a managing partner is never whether software can be prosecuted. It is whether the firm put a nonlawyer channel in front of the public that delivered legal advice in the firm's name.

An intake agent speaks for the firm on every call. When it tells a caller the claim is strong, or that there is plenty of time to file, the caller hears the firm saying it. The fix is not a smarter agent. The fix is a narrower one: collect, confirm, route, and decline to interpret, every time, in the same words.

That narrowness is also what makes the agent useful. Callers mostly want three answers: does this firm handle my problem, will a lawyer call me, and when. An agent can give all three without touching the merits, and a custom AI assistant can be held to that scope with fixed lines instead of open ended answers.

Twelve intake lines, sorted into allowed and prohibited

Each row pairs a line an intake agent can say with the line it must never say on the same subject. The prohibited column is tagged to the rule it trips: NRS 7.285 for interpreting law or facts, NRPC 5.3 for conduct the supervising lawyer answers for, and NRPC 1.18, the Nevada rule on duties to prospective clients, for lines that blur client status or pull in more confidential detail than a conflict check needs.

SubjectAllowedProhibitedRule tripped
Practice areasWe handle injury claims, including car crashes. I can take your details for an attorney.Yes, you have a case.NRS 7.285
Accident date and factsWhat date did it happen, and where?It happened in their lot, so the store is liable.NRS 7.285
Filing deadlineDeadlines can apply to claims like this. An attorney will review timing with you.You still have plenty of time to file.NRS 7.285, NRPC 5.3
Whether to bring itAn attorney decides whether the firm can take this matter.That is definitely worth suing over.NRS 7.285
Case valueI cannot estimate value. The attorney will discuss that.Cases like yours settle for far more than that offer.NRS 7.285, NRPC 5.3
FeesAn attorney explains fees in a written agreement before any work starts.Our usual percentage applies and you will never owe costs.NRPC 5.3
Callback timeAn attorney or case manager will call you back by [the window the firm staffs].A lawyer will call you tonight, guaranteed.NRPC 5.3
Insurer releaseI cannot advise on documents. I am marking this urgent for an attorney.Go ahead and sign the release, it is standard.NRS 7.285
Insurer contactI will note that the adjuster called and flag it for an attorney.Do not talk to the adjuster, just say you have a lawyer.NRS 7.285
Police reportWas a report made? You can give me the report number.The other driver was cited, so you will win.NRS 7.285
Conflict checkWho is the other party, so we can check for conflicts?Tell me everything, including what you told the police.NRPC 1.18
Client statusYou are not a client until an attorney confirms in writing.You are all set, we are your lawyers now.NRPC 1.18, NRPC 5.3

Two patterns run through the prohibited column. The first is interpretation: every NRS 7.285 row takes a fact the caller gave and says what it means legally. The second is commitment: the fee, callback, and client status rows promise what only a lawyer can promise.

The release row needs extra care. Both answers to "should I sign this" are advice, including "do not sign," because either one applies law to the caller's document. The only safe line is to decline and escalate fast, since the caller may be about to act.

The escalation ladder decides what happens when a caller pushes

Most slips will not come from the agent volunteering an opinion. They come from a caller asking the same question three ways until the model answers. The ladder gives the agent a fixed move at each level, so declining is a procedure, not a judgment call.

LevelTriggerAgent doesOwner
1. RoutineCaller describes what happenedCollects intake fields, confirms callback windowAgent
2. Advice requestAsks about merit, value, deadline, or feesReads the fixed decline line once, offers an attorney callbackAgent
3. Repeat or urgencyAsks again, or mentions a release, court date, or served papersStops collecting, marks urgent, transfers live in hours, pages on call after hoursStaff
4. SafetyInjury now, threat, or crisisDirects caller to emergency services, hands to a humanStaff
5. SlipAgent said a prohibited lineAttorney calls the caller to correct it, incident loggedSupervising lawyer

Level 3 is where the design work lives. A caller who mentions a release or served papers is on a clock the agent cannot evaluate, so the answer is speed to a human, not a better script. If nobody can be paged after hours, the agent should say so and give the next staffed window rather than imply a lawyer is on the way.

Level 5 is the one firms skip, and it matters most under the supervision rule. A slip nobody corrects is a different problem from a slip a lawyer caught and fixed the next morning. Our AI receptionist page shows how routing rules like these are configured on a live phone line.

NRPC 5.3 turns unreviewed transcripts into the real exposure

Rule 5.3(b) says "A lawyer having direct supervisory authority over the nonlawyer shall make reasonable efforts to ensure that the person's conduct is compatible with the professional obligations of the lawyer." Subsection (a) puts the same reasonable efforts duty on partners and lawyers with comparable managerial authority to keep firm wide measures in place.

Subsection (c) is the part to read twice. A lawyer is responsible if the lawyer "orders or, with the knowledge of the specific conduct, ratifies the conduct involved," or holds partner or supervisory authority and "knows of the conduct at a time when its consequences can be avoided or mitigated but fails to take reasonable remedial action."

An AI agent leaves a transcript of every call. That is a gift and a trap. A transcript sitting unread in a dashboard is evidence the firm could have known. A transcript a lawyer read, flagged, and acted on is evidence of reasonable efforts. The record cuts whichever way your review habits point.

ABA Opinion 512, from July 2024 and the newest national guidance in our sources, says it directly: "Managerial lawyers must establish clear policies regarding the law firm's permissible use of GAI, and supervisory lawyers must make reasonable efforts to ensure that the firm's lawyers and nonlawyers comply with their professional obligations when using GAI tools." It extends the duty to outside providers, with vendor diligence covering "understanding vendor's security policies and protocols," "using confidentiality agreements," and "understanding the vendor's conflicts check system to screen for adversity among firm clients." It also warns against "uncritical reliance on content created by a GAI tool." ABA opinions are persuasive, not binding in Nevada, but they are the yardstick a reviewer will reach for. The vendor side is covered in our ABA 512 intake post.

The skeptical owner's objection: our human receptionists already say some of this. They probably do, and the same rules apply to them. The difference is that the AI says it identically on every call and records it, so the problem scales and becomes knowable at the same time.

A monthly audit sized for a small firm and a busy one

Reasonable efforts has to be something a lawyer can finish. The call volumes, flag rates, and read times below are our own assumptions for illustration, not figures from any source.

Small firm: assume a two attorney injury practice whose agent handles 300 intake calls a month. Assume a keyword filter flags 15 percent for words like release, deadline, worth, settle, or sign: 300 times 0.15 is 45 calls. Read all 45, plus one in ten of the remaining 255, which rounds to 26. That is 71 transcripts. At an assumed 3 minutes each, 71 times 3 is 213 minutes, about 3.5 hours a month.

Busier firm: assume 1,200 calls a month at the same flag rate, which is 180 flagged calls. Sample one in twenty of the remaining 1,020, which is 51. That is 231 transcripts and 693 minutes, about 11.5 hours. At that size a paralegal does the first pass, and the supervising lawyer reads only the calls marked as possible slips.

The conclusion flips in one place. If more than a small share of audited calls contain a prohibited line, you have a script problem, not an audit problem, and reading more transcripts will not fix it. Tighten the decline lines and retest before next month. The checklist:

  1. Pull every flagged transcript plus the random sample.
  2. Mark any line that interprets facts, quotes timing, values a claim, states fee terms, or implies representation.
  3. For each slip, confirm a lawyer contacted the caller, and log the date.
  4. Check that intake captured conflict check fields and nothing the caller did not need to share.
  5. Confirm urgent calls reached a human inside the promised window.
  6. Record the slip count and any script change, and have the supervising lawyer sign the log.

Where this analysis does not hold

This framework is built for Nevada. A firm taking calls on matters in another state needs that state's statute and rules, which may draw the line differently.

It also assumes the agent holds a conversation. A system that only takes a message and transcribes it carries far less risk, because it never answers anything. At the other end, an agent giving lawyer approved status updates to existing clients is different territory: those callers are clients, the content is lawyer authored, and the questions turn more on confidentiality and verification than on unlawful practice.

Two limits on our sources. The facts here do not settle who the "person" is when software speaks, and we do not guess; the safe working assumption is that the firm owns the channel. And ABA Opinion 512 is persuasive only. This is general information, not legal advice, and your ethics counsel should review your actual script.

Questions owners ask

Can an AI receptionist tell a caller the statute of limitations?

No. Quoting a filing deadline applies law to the caller's facts, which is the conduct NRS 7.285 reaches when it comes through a nonlawyer channel. The agent can say deadlines may apply and an attorney will review timing, then escalate quickly if the caller sounds close to one. Under NRPC 5.3, a deadline quote the firm knew about and left uncorrected lands on the supervising lawyer.

In practice, the firm and its lawyers. NRS 7.285 regulates conduct, and NRPC 5.3 makes partners and supervising lawyers responsible for nonlawyer conduct they ratify, or know about in time to fix and leave alone. The State Bar can also seek an injunction. ABA Opinion 512 applies the same supervision duty to generative AI tools and the vendors behind them.

Does AI intake make a caller a prospective client?

It can. NRPC 1.18 governs duties to people who consult a firm about possibly hiring it, and an intake conversation is that kind of consultation even when software runs it. That is why the agent should collect only what a conflict check and a callback need, and never tell a caller they are already represented. Have counsel confirm how the rule applies to your flow.

What to do this week

  1. Export twenty recent intake transcripts and mark every line that matches a prohibited row in the table above.
  2. Write fixed decline lines for merit, value, deadlines, fees, and documents, and load them into the agent's instructions word for word.
  3. Name the supervising lawyer for the intake channel, put the monthly audit on their calendar, and decide who gets paged for level 3 calls after hours.

If you want a second set of eyes on your intake script, the free 15-minute audit on our law firm page is a quiet place to start.

Drafted with AI assistance, researched, edited, and fact-checked by Elias Musleh on November 8, 2026.

Free 15-min automation audit

Want this handled for you?

We build the automation, run it, and hand you the numbers. Book a fifteen minute call and we will tell you straight whether it is worth doing for your business.

Or reach us directly: vendors@vegasbusinessai.com 702.826.9055