Building a Legal Practice Intake Agent: Client Screening and Appointment Scheduling

The Legal Intake Bottleneck

Law firms lose potential clients at the intake stage more than anywhere else. A prospective client calls, gets voicemail, and moves on to the next firm. Studies show that the first firm to respond wins the client over 70 percent of the time. An AI intake agent answers every call immediately, asks the right qualifying questions, checks for conflicts of interest, and books a consultation — turning a missed call into a scheduled meeting.

This tutorial builds a legal intake agent that handles the complete client screening workflow while respecting the unique ethical requirements of legal practice.

Legal Intake Data Model

Legal intake involves collecting structured information about the potential client, their legal matter, and any parties involved. The data model must capture enough detail for an attorney to decide whether to take the case.

flowchart LR
    CALLER(["Prospective Client"])
    subgraph TEL["Telephony"]
        SIP["Twilio SIP and PSTN"]
    end
    subgraph BRAIN["Legal Intake AI Agent"]
        STT["Streaming STT<br/>Deepgram or Whisper"]
        NLU{"Intent and<br/>Entity Extraction"}
        TOOLS["Tool Calls"]
        TTS["Streaming TTS<br/>ElevenLabs or Rime"]
    end
    subgraph DATA["Live Data Plane"]
        CRM[("CRM and Notes")]
        CAL[("Calendar and<br/>Schedule")]
        KB[("Knowledge Base<br/>and Policies")]
    end
    subgraph OUT["Outcomes"]
        O1(["Consultation booked"])
        O2(["Conflict check passed"])
        O3(["Attorney callback queued"])
    end
    CALLER --> SIP --> STT --> NLU
    NLU -->|Lookup| TOOLS
    TOOLS <--> CRM
    TOOLS <--> CAL
    TOOLS <--> KB
    NLU --> TTS --> SIP --> CALLER
    NLU -->|Resolved| O1
    NLU -->|Schedule| O2
    NLU -->|Escalate| O3
    style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
    style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
    style O1 fill:#059669,stroke:#047857,color:#fff
    style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
    style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937

from dataclasses import dataclass, field
from datetime import datetime, date
from enum import Enum
from typing import Optional

class PracticeArea(Enum):
    FAMILY_LAW = "family_law"
    PERSONAL_INJURY = "personal_injury"
    CRIMINAL_DEFENSE = "criminal_defense"
    BUSINESS_LAW = "business_law"
    ESTATE_PLANNING = "estate_planning"
    EMPLOYMENT_LAW = "employment_law"
    REAL_ESTATE = "real_estate"

class IntakeStatus(Enum):
    NEW = "new"
    SCREENING = "screening"
    CONFLICT_CHECK = "conflict_check"
    QUALIFIED = "qualified"
    DISQUALIFIED = "disqualified"
    CONSULTATION_SCHEDULED = "consultation_scheduled"

@dataclass
class PotentialClient:
    name: str
    phone: str
    email: str = ""
    practice_area: Optional[PracticeArea] = None
    matter_description: str = ""
    opposing_parties: list[str] = field(default_factory=list)
    urgency: str = "normal"
    statute_of_limitations: Optional[date] = None
    referral_source: str = ""
    status: IntakeStatus = IntakeStatus.NEW
    intake_timestamp: datetime = field(default_factory=datetime.now)
    notes: str = ""

@dataclass
class ConflictRecord:
    client_name: str
    opposing_party: str
    case_number: str
    practice_area: str
    is_active: bool

Intake Questionnaire Engine

Each practice area requires different screening questions. We define question flows that the agent follows to gather the information attorneys need.

INTAKE_QUESTIONS = {
    PracticeArea.PERSONAL_INJURY: {
        "required": [
            "When did the incident occur?",
            "Where did the incident take place?",
            "What type of incident was it (car accident, slip and fall, medical, other)?",
            "Have you sought medical treatment?",
            "Have you spoken with the other party's insurance company?",
            "Are there any witnesses?",
        ],
        "statute_years": 2,
        "qualifier": "incident_date",
    },
    PracticeArea.FAMILY_LAW: {
        "required": [
            "What type of family matter (divorce, custody, adoption, support)?",
            "Are there minor children involved?",
            "Is there a current court order in place?",
            "Has the other party retained an attorney?",
            "Are there concerns about safety or domestic violence?",
        ],
        "statute_years": None,
        "qualifier": "matter_type",
    },
    PracticeArea.CRIMINAL_DEFENSE: {
        "required": [
            "Have you been charged, arrested, or are you under investigation?",
            "What is the charge or suspected offense?",
            "When is your next court date, if any?",
            "Are you currently in custody or on bail?",
            "Have you spoken with law enforcement about this matter?",
        ],
        "statute_years": None,
        "qualifier": "court_date",
    },
}

Conflict Checking

Conflict of interest checking is an ethical obligation for law firms. The agent must verify that the firm does not already represent the opposing party in any matter.

class ConflictChecker:
    def __init__(self):
        self.records: list[ConflictRecord] = []

    def add_record(self, record: ConflictRecord):
        self.records.append(record)

    def check_conflicts(
        self, client_name: str, opposing_parties: list[str]
    ) -> dict:
        conflicts_found = []
        all_names = [client_name] + opposing_parties

        for name in all_names:
            name_lower = name.lower()
            for record in self.records:
                if (name_lower in record.client_name.lower()
                    or name_lower in record.opposing_party.lower()):
                    conflicts_found.append({
                        "matched_name": name,
                        "existing_client": record.client_name,
                        "case": record.case_number,
                        "active": record.is_active,
                    })

        if conflicts_found:
            return {
                "has_conflict": True,
                "conflicts": conflicts_found,
                "recommendation": (
                    "Potential conflict of interest detected. "
                    "This matter must be reviewed by a senior attorney "
                    "before proceeding."
                ),
            }
        return {
            "has_conflict": False,
            "conflicts": [],
            "recommendation": "No conflicts found. Safe to proceed with intake.",
        }

conflict_checker = ConflictChecker()
conflict_checker.add_record(ConflictRecord(
    "Smith Industries", "Johnson Corp", "2025-CV-1234", "business_law", True
))

Agent Tools

from agents import Agent, Runner, function_tool

@function_tool
def get_intake_questions(practice_area: str) -> str:
    """Get the screening questions for a specific practice area."""
    try:
        area = PracticeArea(practice_area)
    except ValueError:
        areas = [a.value for a in PracticeArea]
        return f"Unknown practice area. Available areas: {', '.join(areas)}"
    questions = INTAKE_QUESTIONS.get(area, {}).get("required", [])
    numbered = "\n".join(f"{i+1}. {q}" for i, q in enumerate(questions))
    return f"Intake questions for {area.value}:\n{numbered}"

@function_tool
def run_conflict_check(
    client_name: str, opposing_parties: str
) -> str:
    """Run a conflict of interest check against the firm's records."""
    parties = [p.strip() for p in opposing_parties.split(",") if p.strip()]
    result = conflict_checker.check_conflicts(client_name, parties)
    if result["has_conflict"]:
        details = "\n".join(
            f"  - {c['matched_name']} matches case {c['case']}"
            for c in result["conflicts"]
        )
        return f"CONFLICT DETECTED:\n{details}\n{result['recommendation']}"
    return "No conflicts found. Safe to proceed."

@function_tool
def evaluate_case(
    practice_area: str, incident_date: str = "",
    description: str = ""
) -> str:
    """Evaluate basic case viability based on statute of limitations and details."""
    try:
        area = PracticeArea(practice_area)
    except ValueError:
        return "Invalid practice area."
    statute_years = INTAKE_QUESTIONS.get(area, {}).get("statute_years")
    if statute_years and incident_date:
        incident = date.fromisoformat(incident_date)
        deadline = incident.replace(year=incident.year + statute_years)
        days_remaining = (deadline - date.today()).days
        if days_remaining < 0:
            return f"WARNING: Statute of limitations likely expired ({deadline.isoformat()}). Attorney review required."
        if days_remaining < 90:
            return f"URGENT: Only {days_remaining} days until statute deadline ({deadline.isoformat()}). Expedite consultation."
        return f"Statute of limitations: {days_remaining} days remaining (deadline: {deadline.isoformat()})."
    return "Statute check not applicable for this practice area. Proceeding with intake."

@function_tool
def schedule_consultation(
    client_name: str, phone: str, practice_area: str,
    preferred_date: str, notes: str = ""
) -> str:
    """Schedule a consultation with an attorney."""
    return (
        f"Consultation scheduled for {client_name}\n"
        f"Practice area: {practice_area}\n"
        f"Date: {preferred_date}\n"
        f"Phone: {phone}\n"
        f"Pre-consultation notes: {notes}\n\n"
        f"Please bring any relevant documents, correspondence, "
        f"and a photo ID to your consultation."
    )

legal_intake_agent = Agent(
    name="Legal Intake Specialist",
    instructions="""You are a professional legal intake specialist.

1. Begin by asking about the nature of the caller's legal matter.
   Determine the practice area, then use get_intake_questions.
2. Work through the screening questions conversationally, not
   as a rigid checklist.
3. Collect the names of all opposing parties, then run_conflict_check.
4. If relevant, use evaluate_case to check statute of limitations.
5. If no conflicts and the case appears viable, offer to
   schedule_consultation.

CRITICAL ETHICAL RULES:
- Never provide legal advice. Say: "I can help you schedule a
  consultation with one of our attorneys who can advise you."
- If a conflict is detected, do NOT proceed. Explain that the firm
  may have a conflict and recommend the caller contact another firm.
- All information shared is confidential.
- Do not promise outcomes or make statements about case strength.""",
    tools=[get_intake_questions, run_conflict_check, evaluate_case, schedule_consultation],
)

FAQ

How does the agent handle callers who want immediate legal advice?

The agent instructions include a hard rule against providing legal advice. When callers press for advice, the agent responds with variations of "I understand the urgency. Our attorneys can address that specific question during your consultation. Let me get you scheduled as soon as possible." This redirects the conversation toward the actionable step — booking the meeting.

What happens when a conflict of interest is detected?

The agent stops the intake process immediately and informs the caller that the firm may have a conflict that prevents representation. It does not disclose details about the existing client or case. It recommends the caller contact another firm and can optionally provide referral numbers. The intake record is flagged for attorney review.

Can the agent handle multiple practice areas for the same caller?

Yes. Some callers have overlapping legal needs — for example, a car accident victim may need both personal injury and insurance coverage advice. The agent collects information for all relevant practice areas, runs conflict checks against all parties mentioned, and schedules the consultation with a note covering all areas so the attorney can prepare accordingly.

---

#LegalTech #ClientIntake #ConflictChecking #AIScheduling #Python #AgenticAI #LearnAI #AIEngineering