Collectly

Billie is a conversational agent in the full sense, handling billing questions and taking payment across chat, text, email, and voice. A substantial part of the platform, though, is payments infrastructure, statement delivery, and plan management that functions without AI. The company's own framing is accurate: the EHR is the system of record, and this is the system of action layered on top.

The division of work is described, and the rule that sends a patient to a person is not. Billie answers billing questions, explains balances, resolves disputes, takes payments and enrolls patients in payment plans around the clock by chat, text, email and voice, 'and escalates to staff only when needed'; payment plans are limited to options the provider preapproves. The eligibility agent works autonomously 'with human review only when needed' and flags edge cases.

When payer data is incomplete or ambiguous it 'applies learned reasoning patterns to infer missing details', so some copay and benefit figures shown to staff or patients may be inferred rather than returned by the payer. Every interaction is logged, time-stamped and reportable, and agent activity can be monitored like a live team. Nothing published says what counts as a case for a person, whether inferred benefit details are marked as inferred, or what a patient disputing a balance hears before reaching staff.

The approach and its training data are described; no model, version or accuracy figure is. Collectly says its eligibility agent applies large language model reasoning to 271 eligibility responses, combining them with CPT, ICD and taxonomy codes, historical claim trends and appointment context, and infers missing details when payer data is incomplete.

It says Billie was trained on hundreds of thousands of real-world patient billing questions and reads live data from the record system, the insurance system and Collectly's billing platform. No model or provider is named, no version or update practice is published, and the only accuracy figure is a 30 percent improvement in first-pass accuracy with no baseline.

The host is named, and the model layer and other processors are not. Collectly names Amazon Web Services for hosting and a PCI DSS Level 1 certified payment processor, unnamed, for card data. It describes large language model reasoning in the eligibility agent and a voice channel for Billie, but names no model provider, speech or telephony service, and publishes no subprocessor list or the retention and training terms that govern data those parties receive. Billing conversations and eligibility data reach each of them.

A named deployment with figures exists, and the headline numbers are unnamed. A 2021 case study published through the athenahealth marketplace credits Collectly at CleanSlate Centers, an addiction treatment provider with 76 centers in 10 states, with patient collections up by over 250 percent, paper statements down by over 80 percent and a return on investment above 650 percent.

Collectly's home page reports $96.42 million billed and $88.97 million collected (a 92.3 percent collection rate) and 12.6 average days to collect without saying across which customers or period, plus two to three times collection increases, 66 percent lower cost to collect and 95 percent patient satisfaction.

The claims for Billie are more than 475,000 resolutions and up to 85 percent fewer staff handled billing inquiries; for the eligibility agent, a 90 percent plus automation rate and 30 percent better first pass accuracy; and for an unnamed orthopedic practice, collections up 180 percent in six months. None of the AI figures names a customer, baseline or method.

General protections are published, and the questions a billing agent raises are not answered. Collectly describes encryption, access control, audit logs and card tokenization through its processor. It also says Billie 'has been trained on hundreds of thousands of real-world patient billing questions', without saying whose conversations those were, whether they were de-identified, or whether customers' patient interactions continue to train the agent.

No retention period is published for billing records, call recordings or transcripts, nor what happens to data when a provider leaves. The demonstration call is simulated rather than a real patient. Safety engineering, meaning guardrails, hallucination mitigation and how a safety event is handled, is not described, although Billie explains balances and resolves disputes directly with patients.

Business associate status is stated and supported by a published description of how protected information is handled; the agreement itself is not published. Collectly's security page says it supports customers' HIPAA compliance as a business associate, signs business associate agreements, and protects protected health information with encryption, role-based access, audit logs, trained staff, secure handling and a documented incident response procedure, with AWS hosting and key management named.

No agreement text, subcontractor terms, breach notification period or tier information is published. The platform holds billing information that reveals treatment, card payments through a processor, and patient conversations with Billie.

Collectly holds one current attestation of its own and describes the rest as alignment or as its processor's. Its security page says it maintains HITRUST i1 validated status with an annual third-party assessment. It says it maintains 'controls aligned to' SOC 2's security, availability and confidentiality criteria, which is not a statement that a SOC 2 report has been issued.

Card data is tokenized through its PCI DSS Level 1 certified payment processor, so that certification is the processor's. Third-party penetration tests are described, as are AWS hosting, AWS KMS key management and GitLab static analysis; the Billie page separately claims SOC 2 compliance. No report, certificate date or trust portal is published.

No device claim is made, and the product is financial and administrative. What governs Billie and the outreach engine is consumer law: the Telephone Consumer Protection Act, which covers AI-generated voices on outbound calls following the FCC's February 2024 ruling; the Fair Debt Collection Practices Act and state collection rules where an account has gone to collection; card network rules on payments; and the No Surprises Act's good faith estimate requirements, which the cost estimation product produces. Inbound calls from patients raise less of this than outbound voice outreach about balances. Nothing published states a position on consent, calling hours or who carries these obligations.

Multilingual support is sold as a strength, and nothing is published on how the agent performs across languages or patient groups. The Billie page says it communicates in more than 250 languages across chat, text, email and voice. No per-language accuracy, governance framework or evaluation of the agent or the eligibility model is published.

Segmentation is a second exposure: Collectly's Epic announcement says it segments patient populations and builds a tailored financial journey for each, deciding who to reach, on which channel, with which message and when. Nothing says what drives those decisions or whether outreach intensity or payment plan offers vary by predicted ability to pay; the same prediction can offer a plan earlier or pursue a balance harder. Collectly does not claim equal performance across languages.

Mechanisms let staff review and correct the agent, and nothing stands behind a wrong answer. Every Billie interaction is logged, time-stamped and reportable; agent activity can be monitored like a live team; identity is verified before billing details are shared; payment plans are limited to options the provider pre-approves; and eligibility edge cases are flagged for review.

No warranty, remedy, service credit or measured error rate is published, and the 475,000-plus resolutions and up to 85 percent fewer staff-handled inquiries measure volume, not correctness. A wrong balance explanation can lead a patient to pay what they should have disputed, and an inferred copay can produce a surprise bill.

Integration breadth is the strongest element: more than 20 leading EHRs via API or FHIR, spanning ambulatory and enterprise systems, plus a bidirectional Epic integration listed in Connection Hub on the Epic Showroom as of July 2026, meaning live Epic customers and pre built integration rather than a claimed capability. EHR agnostic deployment matters here because patient billing sits across every system a group runs.

A single hosted platform on a named cloud, with the location implied rather than committed. Collectly's security page says it hosts on Amazon Web Services, deploys on Kubernetes and manages keys with AWS KMS. No region, residency commitment, tenancy model or backup and recovery posture is published. Billie's voice channel adds telephony and speech services, and card payments pass through a separate processor; neither is named.

Cost is not disclosed before a sales conversation. Collectly sells through demo requests to medical groups, health systems, billing and revenue cycle companies (which white-label it) and record system vendors (which embed it); the demo form asks for provider count.

The company publishes no price, unit of charge, tier or implementation fee, and nothing from it says whether fees are a subscription or a share of collections; the subscription model repeated in software directories is not the company's statement. For a product that works patient balances, that leaves open whether the vendor earns more when patients pay more.

Coverage is named across provider types, with validation behind part of it. Collectly sells to solo and group practices, health systems, urgent care, billing and revenue cycle companies and record system vendors, and reports more than 3,000 healthcare facilities. Named validation is thin: CleanSlate Centers (addiction treatment) through a 2021 case study, Action Urgent Care in a testimonial, and an unnamed federally qualified health center on Epic; other stories are unnamed.

The product spans pre-visit, point-of-service and post-visit patient financial work, which behaves similarly across specialties. Its AI eligibility and benefits agent verifies coverage, retrieves benefits from payer portals and identifies prior authorization requirements, and its previsit intake automation handles digital check in, insurance card capture, demographic validation and consents.