Glossary of Terms

Federal regulations providing extra privacy protections specifically for substance use disorder (SUD) treatment records — stricter than standard HIPAA. These rules restrict how SUD treatment information can be shared, even between healthcare providers. AI systems in addiction treatment must comply with both HIPAA and 42 CFR Part 2.

The percentage of callers who hang up before reaching a live person or AI agent, typically due to long hold times. Industry average in behavioral health is 20-30%. AI eliminates hold times entirely, reducing abandonment to near zero for facilities using voice AI as their first responder.

A specialized staff member who builds rapport with prospective patients and families, addresses concerns and objections, and guides them toward committing to treatment. Unlike intake coordinators who focus on data collection, admissions counselors focus on motivation and relationship-building.

The ability to handle admissions inquiries outside of normal business hours (evenings, weekends, holidays). Studies show 40-60% of behavioral health inquiries come after hours. Without AI or outsourced staff, these calls go to voicemail — and 80% of those callers never call back.

A set of protocols that allows different software systems to communicate with each other. AI admissions tools use APIs to connect with EMRs, CRMs, scheduling systems, insurance verification databases, and phone systems. The quality and breadth of a vendor's API determines how well it integrates with your existing tech stack.

The American Society of Addiction Medicine's multidimensional assessment framework for determining appropriate level of care for substance use disorders. It evaluates patients across six dimensions including withdrawal potential, medical conditions, emotional/behavioral conditions, and recovery environment. Some AI systems incorporate ASAM-aligned screening.

The process of determining which marketing channel or touchpoint deserves credit for generating an admission. Multi-touch attribution tracks the entire patient journey — from first ad click through phone call through admission. AI systems that integrate with call tracking enable more accurate attribution.

The average duration of an admissions call from pickup to completion. For human agents, AHT in behavioral health is typically 8-15 minutes. AI agents may have shorter AHT for qualification calls but should not rush sensitive conversations. AHT is less important than conversion rate as a success metric.

A legal contract between a healthcare provider and any vendor that handles PHI on their behalf. Every AI vendor processing patient data MUST have a signed BAA. This is non-negotiable — using an AI tool without a BAA is a HIPAA violation regardless of how secure the technology claims to be.

Real-time tracking of open treatment slots across a facility or network of facilities. AI systems can check bed availability during conversations and immediately confirm whether a patient can be admitted, eliminating delays caused by manual bed board checks.

The systematic evaluation of admissions phone calls against defined quality criteria. AI call scoring analyzes 100% of calls (vs. the 3-5% typically reviewed by humans) across dimensions like needs discovery, empathy, process adherence, and urgency creation. Scores identify coaching opportunities and top performers.

Technology that assigns unique phone numbers to different marketing channels (Google Ads, website, billboard) to measure which sources generate admissions calls. AI systems often integrate with call tracking platforms (CallRail, CallTrackingMetrics) to attribute conversions and optimize marketing spend.

The current number of patients occupying beds in a treatment facility at any given time. Census management is a primary business concern — empty beds represent lost revenue, while full census means turning away patients. AI admissions tools help maintain optimal census through better conversion and follow-up.

A cloud-hosted Private Branch Exchange — a virtual phone system that manages call routing, voicemail, hold queues, and extensions without physical hardware. AI voice agents typically integrate with cloud PBX systems (like RingCentral, 8x8, or Vonage) to intercept and handle calls.

AI systems designed to engage in multi-turn, context-aware dialogue with humans. Unlike simple chatbots that match keywords to canned responses, conversational AI maintains context across an entire conversation, remembers what was said earlier, and adapts its approach based on the flow of discussion.

The percentage of leads that progress from one stage to the next in the admissions funnel. Key conversion rates include: call-to-assessment scheduled (30-50% is good), assessment-to-admit (40-60% is good), and overall inquiry-to-admit (10-20% is typical). AI tools aim to improve each stage.

The total marketing and operational cost to acquire one admitted patient. This includes advertising spend, staff salaries, technology costs, and overhead divided by total admissions. Behavioral health CPA ranges from $2,000-$15,000 depending on level of care and market. AI can reduce CPA by improving conversion rates.

AI's ability to identify when a caller or chat visitor is in immediate danger — expressing suicidal ideation, experiencing overdose symptoms, or facing domestic violence. Proper crisis detection triggers immediate escalation to trained crisis counselors or emergency services, bypassing normal conversation flows.

The automated process of immediately connecting a person in crisis with appropriate emergency resources — trained crisis counselors, the 988 Suicide & Crisis Lifeline, or local emergency services. AI systems must have robust crisis routing protocols that activate instantly when danger is detected.

Software that tracks all interactions with prospective and current patients throughout the admissions funnel. In behavioral health, CRMs like Salesforce, HubSpot, or specialized platforms (Dazos, New Resilience) manage leads from first contact through admission. AI tools must sync data to the CRM in real-time.

A real-time query to determine whether a patient's insurance is currently active and what services are covered. Unlike full VOB which details specific benefits, an eligibility check is a quick yes/no on whether coverage exists. AI can perform this instantly during a call to qualify leads faster.

Digital systems that store patient clinical information including diagnoses, medications, treatment plans, and progress notes. In behavioral health, common EMRs include Kipu, Sunwave, Netsmart, and Credible. AI admissions tools must integrate with the facility's EMR to avoid double-entry of patient data.

The process of further training a pre-trained AI model on domain-specific data to improve its performance in a particular context. Behavioral health AI vendors fine-tune models on thousands of real admissions calls to understand industry jargon, common objections, and appropriate empathetic responses.

The percentage of inquiries resolved (scheduled, admitted, or appropriately referred) on the first contact without requiring a callback. Higher FCR means fewer leads lost to follow-up gaps. AI systems with real-time VOB and scheduling capabilities can significantly improve FCR.

Safety constraints programmed into AI systems to prevent harmful, inaccurate, or off-topic responses. In behavioral health AI, guardrails ensure the system never provides medical advice, always escalates crisis situations, stays within its knowledge base, and doesn't make promises about treatment outcomes.

When an AI model generates information that sounds plausible but is factually incorrect or fabricated. In behavioral health, hallucinations are dangerous — an AI might invent insurance coverage details, fabricate program offerings, or provide incorrect clinical information. Vendors must implement guardrails to prevent this.

The Health Insurance Portability and Accountability Act — federal law requiring healthcare organizations to protect patient health information (PHI). Any AI system handling admissions data must be HIPAA-compliant, meaning it encrypts data, limits access, maintains audit logs, and has a Business Associate Agreement (BAA) in place.

Healthcare data exchange standards. HL7 (Health Level Seven) is the legacy standard; FHIR (Fast Healthcare Interoperability Resources) is the modern replacement. These standards define how clinical data is structured and transmitted between systems. AI vendors that support FHIR can integrate more easily with clinical systems.

Whether a treatment facility has a contracted rate with a specific insurance company (in-network) or not (out-of-network). In-network typically means lower patient costs and faster authorization. AI intake systems must quickly determine network status to set accurate expectations about costs.

The legal and ethical requirement to inform patients about how their data will be used, including whether they are speaking with AI vs. a human. Regulations vary by state, but transparency about AI use during admissions conversations is increasingly expected and may become legally required.

A staff member responsible for handling initial patient inquiries, collecting demographic and insurance information, conducting preliminary clinical screening, and guiding prospective patients through the admissions process. AI intake tools aim to augment or partially automate this role.

The AI's ability to determine what a user is trying to accomplish from their words. For example, distinguishing between someone seeking information about programs, someone ready to admit immediately, someone calling about a loved one, or someone in active crisis — each requiring a different conversation path.

Traditional phone tree systems that route callers through menu options ('Press 1 for admissions, press 2 for billing'). IVR is being replaced by conversational AI that understands natural speech. However, some AI systems still use IVR as a first layer before engaging the AI agent.

A structured repository of information that an AI system references when answering questions. For behavioral health facilities, this includes program details, insurance accepted, staff credentials, visiting hours, location info, and FAQs. The quality and completeness of the knowledge base directly impacts AI accuracy.

A deep learning model trained on massive text datasets that can generate human-like text, answer questions, and conduct conversations. Models like GPT-4, Claude, and Gemini power many behavioral health AI chatbots and voice agents. LLMs require careful fine-tuning and guardrails to handle sensitive clinical conversations appropriately.

The time delay between when a user finishes speaking and when the AI begins its response. In voice AI, latency below 500ms feels natural; above 1 second feels awkward and robotic. Leading behavioral health voice AI vendors achieve 300-800ms response latency. This is a critical evaluation criterion.

The process of re-engaging prospective patients who expressed interest but didn't complete the admissions process. This includes people who called but didn't schedule, started an online form but abandoned it, or were previously assessed but didn't admit. AI enables persistent, personalized follow-up at scale.

The percentage of initial inquiries (phone calls, web chats, form submissions) that ultimately result in a patient being admitted to treatment. Industry average for behavioral health is 10-20%. AI tools aim to improve this by ensuring faster response, better qualification, and persistent follow-up.

The intensity of treatment a patient requires, typically determined by clinical assessment. Common levels include: Detox, Residential/Inpatient (RTC), Partial Hospitalization (PHP), Intensive Outpatient (IOP), and Outpatient (OP). AI triage systems help determine appropriate LOC during initial contact.

The process of following up with callers who didn't connect with a live person or AI agent. This includes callbacks, automated texts, and outbound AI calls. Given that each missed behavioral health call represents $8,000-$15,000 in potential revenue, recovery systems provide significant ROI.

A clinical counseling approach that helps people resolve ambivalence about behavior change. Some AI systems incorporate MI principles — using open-ended questions, affirmations, reflective listening, and summaries to help callers move toward treatment readiness rather than using high-pressure sales tactics.

A system design where multiple specialized AI agents work together on a single task, each handling a different aspect. For example, one agent might handle speech recognition, another manages conversation flow, a third performs insurance lookup, and a fourth scores call quality — all coordinating in real-time.

A branch of AI that enables computers to understand, interpret, and generate human language. In behavioral health admissions, NLP powers chatbots and voice agents that can comprehend caller intent, detect emotional cues, and respond naturally to questions about treatment options.

A subset of NLP focused specifically on comprehension — determining what a user means rather than just what they said. NLU enables AI agents to understand that 'my son needs help with drinking' is a request for substance abuse treatment for a family member, not a literal statement about beverages.

A metric measuring patient satisfaction and likelihood to recommend the facility. While primarily used post-treatment, some AI systems measure caller satisfaction during the admissions process itself — tracking whether the AI interaction felt helpful, empathetic, and efficient.

AI systems that proactively make phone calls or send messages to leads who didn't convert on initial contact. Unlike inbound systems that wait for calls, outbound AI initiates contact — following up with people who called but didn't admit, re-engaging alumni, or reaching out to web form submissions.

The insurance company or entity responsible for paying for a patient's treatment. Common behavioral health payers include Aetna, Blue Cross Blue Shield, Cigna, UnitedHealthcare, Medicaid, Medicare, and Tricare. AI systems must interface with multiple payer systems for real-time verification.

Any individually identifiable health information including names, dates, phone numbers, diagnoses, treatment records, and insurance details. AI systems processing admissions calls handle PHI extensively — every caller's name, substance use history, and insurance information is PHI that must be protected.

Approval required from an insurance company before certain treatments can begin. Many behavioral health admissions require prior authorization, which can delay treatment by days. Some AI systems initiate the pre-auth process automatically during intake, reducing delays between decision and admission.

The systematic monitoring and evaluation of admissions team performance to ensure consistent quality. Traditional QA involves supervisors listening to recorded calls; AI QA automates this by scoring every interaction across multiple dimensions and flagging calls that need human review.

A technique where an AI model retrieves relevant information from a knowledge base before generating a response, rather than relying solely on its training data. This reduces hallucinations and ensures responses are grounded in accurate, facility-specific information like current bed availability or accepted insurance plans.

The end-to-end financial process from patient registration through final payment collection. In behavioral health, RCM includes VOB, pre-authorization, claims submission, denial management, and patient collections. AI tools that capture accurate data during intake reduce downstream RCM errors.

The financial return generated relative to the cost of an AI investment. Calculated as: (Additional Revenue from AI - Cost of AI) / Cost of AI × 100%. For behavioral health AI, ROI is typically driven by recovered missed calls, improved conversion rates, and reduced staffing costs. Most vendors claim 3-10x ROI.

AI's ability to detect the emotional tone of a conversation — whether a caller is anxious, angry, hopeful, or in crisis. Behavioral health AI uses sentiment analysis to adjust its tone, escalate to human staff when distress is detected, and flag calls that may require immediate clinical intervention.

The percentage of scheduled assessments or admissions where the patient actually arrives. Behavioral health show rates are notoriously low (50-70%) due to ambivalence, continued substance use, or logistical barriers. AI follow-up systems that send reminders and maintain engagement can improve show rates by 15-30%.

A one-time contract negotiated between an out-of-network provider and an insurance company to cover a specific patient's treatment at an agreed rate. Some AI systems flag when an SCA might be appropriate and can initiate the request process automatically.

Session Initiation Protocol trunking — the technology that connects phone systems to the internet, enabling Voice AI to receive and make calls. SIP trunks replace traditional phone lines and allow AI systems to handle calls programmatically, including recording, transferring, and conferencing.

A security framework developed by the American Institute of CPAs that evaluates how well a company protects customer data across five trust principles: security, availability, processing integrity, confidentiality, and privacy. SOC 2 Type II certification means the vendor has been audited over time, not just at a single point.

Technology that converts spoken audio into written text in real-time. Also called Automatic Speech Recognition (ASR). Critical for voice AI systems that need to understand what callers are saying before generating a response. Modern STT achieves 95%+ accuracy even with diverse accents and emotional speech.

The time between when a prospective patient first reaches out and when they receive a meaningful response. Research shows that responding within 5 minutes increases conversion by 400% compared to responding after 30 minutes. AI enables instant response 24/7, effectively reducing speed-to-lead to zero.

Technology that converts written text into natural-sounding spoken audio. Modern TTS systems produce voices nearly indistinguishable from humans, with appropriate prosody, emotion, and pacing. Quality TTS is essential for voice AI agents to sound warm and empathetic rather than robotic.

The elapsed time from a patient's first contact to their physical admission to the facility. Shorter time-to-admit correlates with higher show rates and lower dropout. AI tools that automate VOB, scheduling, and pre-admission paperwork can reduce time-to-admit from days to hours.

The process by which insurance companies evaluate whether continued treatment is medically necessary. UR typically occurs at set intervals during a patient's stay. While primarily a clinical function, AI systems that capture thorough intake data provide better documentation for UR justification.

The process of confirming a patient's insurance coverage, benefits, deductibles, copays, and authorization requirements before admission. Traditional VOB takes 30-60 minutes per patient via phone calls to insurers. AI-powered VOB can complete verification in seconds by querying payer databases electronically.

When an AI agent connects a caller to a human staff member while providing full context about the conversation so far — the caller's name, insurance info, clinical needs, and emotional state. This eliminates the frustrating experience of repeating information. Contrast with 'cold transfer' where no context is passed.

An automated message sent from one system to another when a specific event occurs. For example, when an AI agent completes a call, a webhook can instantly notify the CRM to create a new lead record, trigger a task for staff follow-up, or update bed availability in the scheduling system.

The brief period during which a person with a substance use or mental health disorder is ready and willing to accept help. This window can close within hours due to ambivalence, external pressures, or continued substance use. The urgency of this window drives the need for instant AI response.

No-code automation platforms that connect different software tools through pre-built connectors. Many behavioral health AI vendors offer Zapier or Make integrations as a quick way to connect with systems they don't natively support — useful for smaller facilities without dedicated IT teams.