Key takeaways
- Qualification answers three questions: is this our kind of matter, is it in our jurisdiction, and how soon does an attorney need it.
- Turn every advice question into a fact question, then book the attorney; that is the core screening skill.
- Florida's guidance requires intake chatbots to say they are AI and not lawyers[1]; the same transparency suits human screeners.
- Stanford found general chatbots wrong on 58% to 82% of legal queries[2], so no tool should improvise answers at screening.
What qualifying a legal lead means
Sales teams in other industries qualify leads by budget, authority, need and timing. Law firms need something narrower and safer. A legal lead is qualified when staff can answer three questions from facts alone:
- Fit: is this a type of matter the firm handles, under its written acceptance criteria?
- Jurisdiction: is the matter in a state, county or court where the firm's attorneys practice?
- Urgency: is there a date, custody situation or other time pressure that means an attorney needs to see it today?
Notice what is not on the list: whether the person has a good case, what it is worth or what they should do. Those are legal judgments. A screener who answers them is giving legal advice, however casually it is phrased, and the person may rely on it.
The three tests, with questions that stay factual
| Test | Factual questions to ask | The advice trap to avoid |
|---|---|---|
| Fit | What kind of matter is it? Is it about you or someone else? What outcome are you hoping for? | "That sounds like a strong case." |
| Jurisdiction | Which state and county? Where do you and the other side live? Is anything filed in a court yet? Which court? | "You'll need to file in federal court." |
| Urgency | Is there a court date, deadline or letter asking for a response? What date is on it? Is anyone in custody? | "You have plenty of time." |
| Conflicts | Who else is involved? Full names of people and businesses, please. | Hearing what each person did before running the search |
| Representation | Do you already have a lawyer on this? | Discussing the matter with someone who is represented |
Every question in the middle column has an answer the caller knows. None requires the screener to interpret the law.
Convert advice questions into fact questions
Callers ask advice questions constantly, often in the first minute. The skill that separates good screeners from risky ones is turning each advice question into a fact question, then pointing to the consultation.
- Caller: "Can my landlord evict me for this?" Screener: "Have you received any written notice from your landlord? What date is on it?"
- Caller: "Do I need a lawyer for a DUI?" Screener: "Has a court date been set? Which court is it in?"
- Caller: "Is my ex allowed to take the kids out of state?" Screener: "Is there a custody order in place now? Which county issued it?"
- Caller: "How much is my injury case worth?" Screener: "When did the accident happen, and in which county?"
- Caller: "Should I talk to the insurance adjuster?" Screener: "Has the adjuster asked you for a statement, and by when? I'll make sure an attorney sees that today."
Then close the loop with one sentence: "Those are exactly the questions the attorney will answer with you. Let me book a time."
This approach follows the line drawn in Florida's Ethics Opinion 88-6, which the Florida Bar cites for intake chatbots: nonlawyers doing intake identify their status, limit questions to facts, give no legal advice and refer legal questions to the lawyer[1].
Jurisdiction questions that trip screeners up
Jurisdiction is the test that looks easiest and goes wrong most often, because callers describe where they live, not where the matter belongs. Give screeners a short set of follow-ups for the common traps, and an escalate rule for anything outside them.
- The caller lives in one state and the incident happened in another. Record both, plus where the other party lives. An attorney decides which matters.
- The caller has moved since the matter started. Record the county where any existing order or case was issued, not just the new address.
- A business matter involves a company registered in one state and operating in another. Record the state of registration and the states where it does business.
- A family matter crosses state lines, such as a parent who has moved with a child. Record where the child has lived and for how long, and flag it for an attorney.
- Anything already filed. Ask which court, which county and, if the caller has it, the case number from the paperwork.
The screener's job is to capture these facts accurately, not to decide which court has jurisdiction. Where the answer is not obvious from the firm's written list of states and counties, the outcome is escalate. A five-minute attorney review is far cheaper than a consultation booked for a matter the firm cannot handle, or a matter declined that the firm could have taken.
Decision rules after screening
Screening ends with one of four outcomes, chosen by rule rather than mood.
Escalate now
Any date within the firm's urgent window, any arrest or detention, any protective order or child safety issue. The on-call attorney is contacted before the caller hangs up, if possible.
Book
Fit, jurisdiction and conflicts all pass. The consultation is booked on the call and confirmed in writing.
Refer
The matter is real and urgent enough to need a lawyer, but not this firm's kind of work or not in its area. Give referral details and record the referral.
Decline
The matter fails the firm's criteria and there is no useful referral. Decline politely, suggest the state bar referral service, and mention that legal matters can have time limits.
If the screener cannot tell which outcome applies, the rule is to escalate to an attorney, not to guess. Borderline calls are where the firm's judgment belongs.
Screening with tools and AI
Many firms now use answering services, web forms, chat tools or AI to handle part of their screening. The same boundary applies, and the risks are sharper because tools can sound confident when they are wrong.
Stanford researchers found that general-purpose chatbots hallucinated between 58% and 82% of the time on legal queries, and that even purpose-built legal research tools produced incorrect information more than 17% of the time in one case and more than 34% in another[2].
Those figures come from legal research tasks, not intake, but the lesson carries over: a tool that improvises answers to legal questions will sometimes be wrong. For screening, tools should be configured to ask factual questions and hand off, never to answer.
Regulators have said so directly. The Florida Bar's 2024 opinion says generative AI chatbots that communicate with clients or third parties must disclose that they are an AI program and not a lawyer or employee of the firm[1]. The FTC's 2025 order against DoNotPay, which marketed "the world's first robot lawyer", bars the company from claiming its service performs like a real lawyer without sufficient evidence[3]. California's State Bar says supervision of an AI intake system must match its level of access and autonomy[4].
An SI Employee such as Ava is built for the factual side of this work: it follows the firm's approved intake questions and approved boundary wording, with no legal advice and no promise of representation, and passes the qualified matter to the firm's people.
Example: a screening call that stays on the right side
Example (fictional): a caller says, "My business partner is cutting me out of the company and I want to know if I can stop him." The screener notes the matter type (business dispute), asks which state the company is registered in and where the partners live (both in Ohio, the firm's state), and asks for the partner's full name and the company's legal name for the conflict check.
She then asks whether anything has been filed or whether there is a meeting or vote scheduled. There is: a members' meeting next Thursday to vote on removing him. That is a date, so she flags it for an attorney today.
The caller asks again whether he can stop the vote. She says: "That's exactly what the attorney will look at with you, and I've flagged next Thursday so they see it today. Can I book you for tomorrow at 11 or 3?" He takes 11. At no point did she say whether he could stop the vote.
Training and checking screeners
- Give every screener the advice-to-fact conversion list and add to it as new questions come up.
- Role-play the five most common advice questions in your practice areas until the conversion is automatic.
- Review a sample of intake notes each month for phrases like "should be fine", "you have a case" or "they can't do that".
- Track how often screeners choose "escalate" because they were unsure. A rising number means the criteria need to be clearer.
Frequently asked questions
How do you qualify a legal lead?
Test fit (is it a matter the firm handles), jurisdiction (is it where the firm practices) and urgency (is there a date or custody issue), using factual questions only, then escalate, book, refer or decline by rule.
What is the difference between legal information and legal advice at intake?
At intake, staff should avoid both interpretation and prediction. Asking what a letter says is factual; saying what it means or what the caller should do is advice and belongs to the attorney[1].
Can an AI chatbot screen legal leads?
It can collect facts if it is supervised, discloses that it is AI and not a lawyer, and is configured to hand off rather than answer legal questions[1][4].
What if a caller insists on an answer before booking?
Acknowledge the question, explain that only an attorney can answer it, record any date involved, and offer two specific consultation times.
References
The Florida Bar (2024). Florida Bar Ethics Opinion 24-1 (opens in a new tab) โ
Stanford Institute for Human-Centered Artificial Intelligence (HAI) (2024). AI on Trial: Legal Models Hallucinate in 1 out of 6 (or More) Benchmarking Queries (opens in a new tab) โ
Federal Trade Commission (2025). FTC Finalizes Order with DoNotPay That Prohibits Deceptive 'AI Lawyer' Claims (opens in a new tab) โ
The State Bar of California, Standing Committee on Professional Responsibility and Conduct (2026). Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law (opens in a new tab) โ