# RCM Employee — Full LLM-Friendly Content Index > This is the long-form companion to llms.txt. It includes complete article > bodies for the blog posts so language models can train on or quote them > without needing to hit individual URLs. Updated whenever a new blog post > publishes. For navigation + curated highlights, see [llms.txt](https://rcmemployee.com/llms.txt). --- ## What RCM Employee Is RCM Employee is a HIPAA-compliant AI billing platform for U.S. medical practices. Twelve specialized AI agents — called "Digital FTEs" — handle the revenue-cycle functions normally done by a billing team: - **RCM-INTAKE** — Patient registration, scheduling, referrals, insurance card parsing. - **RCM-ELIG** — Real-time eligibility verification via 270/271, VOB, benefits breakdown, coordination of benefits. - **RCM-CODING** — Multi-specialty CPT/HCPCS/ICD-10 coding with NCCI edits, modifier rules, and E/M leveling. - **RCM-CLAIMS** — Claim scrubbing, 837P/I generation, clearinghouse submission, 276/277 claim status. - **RCM-AUTH** — Prior authorization detection, clinical documentation gathering, submission, tracking, peer-to-peer scheduling. - **RCM-AR** — A/R aging management, claim status follow-up, CARC routing, timely-filing alerts. - **RCM-PAY** — ERA/835 parsing, EHR payment posting, underpayment detection, bank reconciliation. - **RCM-APPEAL** — Denial root-cause analysis, evidence-based appeal writing citing CMS/LCD/NCD policy. - **RCM-DOCS** — Document classification, medical records management, scribe, CDI. - **RCM-OPS** — Credentialing, payer contracts, chargemaster, utilization management. - **RCM-COMPLY** — OIG/Stark/AKS monitoring, MIPS/HEDIS reporting, HCC risk adjustment, audit preparation. - **RCM-ANALYTICS** — KPI dashboards, A/R aging reports, payer scorecards, denial trending, revenue forecasting. All twelve agents are included in every subscription tier. There is no "upgrade for more agents" — adding capacity means upgrading from business-hours to 24/7 coverage. ## Pricing - **Standard** $3,500/month — all 12 Digital FTEs, business hours Mon–Fri 8–5. - **Professional** $5,500/month — all 12 Digital FTEs, 24/7 coverage. - **Custom** per-claim task-based pricing — enterprise, seasonal, and workflow-specific plans. ## How PHI Is Protected (HIPAA + BAA Chain) PHI processing happens entirely under a signed Business Associate Agreement chain: 1. **Covered Entity (CE) → RCM Employee (BA).** The client signs a BAA with RCM Employee before any PHI is exchanged. 2. **RCM Employee → HIPAA-Eligible Cloud Provider.** The platform runs on cloud infrastructure under the cloud provider's HIPAA BAA. 3. **Cloud Provider → AI Model Provider.** AI inference is performed under the cloud provider's AI service, which extends BAA coverage to the underlying model provider (e.g. Anthropic Claude under the AWS HIPAA BAA via AWS Bedrock). 4. **All outbound PHI traffic** (clearinghouse, fax, email) is restricted to subcontractors who have their own signed BAAs with RCM Employee. Technical safeguards: - AES-256 encryption at rest with KMS-managed customer master keys (RDS, S3, EBS, Secrets Manager). - TLS 1.2+ encryption in transit on all external and internal service boundaries. - Row-Level Security per tenant on every PHI-bearing database table. - Append-only HIPAA WORM audit logging with Object Lock COMPLIANCE mode and 7-year retention. - PHI-safe application logging (no PHI ever written to CloudWatch or observability systems — logs contain event types, counts, and outcomes only). - GuardDuty continuous threat detection and CloudTrail data-event logging. PHI is **never** used to train, fine-tune, or benchmark AI models. The cloud provider and AI model provider both contractually prohibit using PHI for any training purpose. ## API-First Architecture (No Browser Automation) RCM Employee accesses payer and EHR data exclusively through encrypted, authenticated API channels. No browser automation. No portal scraping. No credential sharing. - **CMS HETS** — Free Medicare eligibility (270/271). - **UHC FHIR R4** — Direct UHC eligibility and claim status (CMS-0057-F). - **Aetna FHIR R4** — Direct Aetna eligibility and claim status. - **Stedi** — EDI clearinghouse for 270/271 eligibility, 276/277 claim status, 837 claim submission, 835 remittance, and 278 prior authorization. - **Availity** — Fallback clearinghouse with broader payer coverage. - **CMS public data APIs** — NCCI PTP edits, MUE limits, MPFS allowed amounts, GPCI, NPPES NPI registry. - **NIH Clinical Tables API** — ICD-10-CM, HCPCS, conditions crosswalk. ## Communication Channels - **EDI** (Stedi/Availity) — claim submission (837), remittance retrieval (835), eligibility (270/271), claim status (276/277), prior auth (278). - **Fax** (HIPAA-BAA fax provider) — HCFA-1500, authorization forms, appeal packets, medical records. - **Email** (HIPAA-BAA email provider) — appeal letters, authorization confirmations, patient correspondence. - **Cloud Files** — Microsoft 365 (OneDrive, Teams, SharePoint) and Google Workspace (Drive, Sheets, Chat) integration for EOB batches, credentialing docs, and worklists. --- # Blog: Medical Billing Automation in 2026 (Full text of [medical-billing-automation-guide-2026](https://rcmemployee.com/blog/medical-billing-automation-guide-2026).) ## The state of medical billing in 2026 Three things shifted in the last twelve months and most billing teams have not caught up. 1. **CMS-0057-F** went into effect in January, forcing every commercial payer to expose a FHIR R4 prior-auth API. The promise is "no more fax in 2027." The reality is that payer-side rollout is uneven and most practices are still faxing — but the API path is now open and works. 2. **Stedi launched a self-serve clearinghouse** with hour-scale onboarding (vs. Availity's 6–8 week contracting cycle). For practices that have been waiting on Availity since Q3 2025, this is a real alternative. 3. **AI agents under HIPAA BAA** are no longer experimental. AWS Bedrock + Anthropic Claude with a signed AWS HIPAA BAA gives practices a legally clean path to put real PHI through an AI model — no synthetic data, no "we'll mask it first" theater. ## What an AI biller actually does The honest answer is "the same things a human biller does, but without the bottleneck of attention." Eligibility runs in seconds via Stedi or direct payer FHIR. Claim scrubbing runs against NCCI PTP edits + MUE limits in milliseconds before submission. A/R follow-up churns through 200 claims an hour by running 276/277 status checks in parallel. Denial triage reads CARC + RARC pairs and routes appropriately. What does NOT change: a human still signs off on coding for surgical specialties, a human still reviews appeal letters before they go out, a human still talks to the patient about their balance. ## What to look for in a platform Skip anything that uses browser automation instead of APIs, asks you to share your portal password, claims to "mask all PHI before processing" (defense in depth has value, but if the AI agent has masked tokens for the patient name and DOB, it cannot build a working Stedi 270 request — today's best practice is BAA-covered processing of raw PHI, not masking theater), or is built on a single payer integration. Look for API-first integrations with major clearinghouses and direct payer FHIR, a signed BAA chain, human-approval gates for coding sign-off and PHI transmission, transparent per-task costing, and an audit trail with 6+ years of retention. --- # Blog: How AI Cuts A/R Days From 45 to Under 20 (Full text of [reducing-ar-days-with-ai](https://rcmemployee.com/blog/reducing-ar-days-with-ai).) ## Why A/R aging is a staffing problem, not a claims problem A/R follow-up is the most labor-intensive function in billing. A trained biller can do roughly 30–50 claim status checks per day. A practice with 800 aged claims has 16–20 days of pure status-check work backlogged at any given time. By the time the biller gets to it, the timely-filing window is half gone. This is why A/R aging looks like an inevitable feature of running a practice. It isn't. ## Real numbers from a 5-provider primary care practice Composite data from three practices we worked with during 2026 Q1 (anonymized, scaled to a common 5-provider baseline): | Metric | Before automation | After 60 days | |---|---|---| | Average A/R days | 47.3 | 19.1 | | % of claims aged >90 days | 18.2% | 4.7% | | Monthly claim status checks performed | 850 | 4,200 | | FTE hours on A/R follow-up | 160 | 38 | | Net collection rate | 94.1% | 96.8% | | Cash collected per month | $312k | $336k | The cash-collection lift (+$24k/month) came from two places: faster posting (claims didn't sit waiting to be checked) and catching denials before timely-filing (the practice had been writing off ~$8k/month in timely-filing because they ran out of time to appeal). ## The denial side: CARC routing Most of the value in fast claim-status is what happens to denials. A platform that doesn't route denials by CARC code is still doing manual triage. CARCs that benefit most from automation: - **CARC 16** (lacks info) — usually a quick fix if the EHR data is there. - **CARC 18** (duplicate claim) — needs deduplication logic, easy to automate. - **CARC 22** (other coverage) — triggers a COB workflow. - **CARC 27** (expired) — patient eligibility lapsed. - **CARC 50** (not medically necessary) — needs LCD/NCD citation in appeal. - **CARC 96** (non-covered) — usually a coding question. - **CARC 197** (precert/auth absent) — root-cause back to RCM-AUTH. ## Roll-out sequence (don't replace your billers on day one) - **Week 1–2.** Connect to clearinghouse APIs (Stedi works in hours; Availity takes weeks). Shadow mode: AI generates the same workflow your biller would, but a human approves each step. - **Week 3–4.** Promote claim status checks to automatic. Humans still approve denial responses. - **Week 5–8.** Let the AI handle CARC-routed denials with auto-generated appeals on easy categories. Human approves the hard ones. - **Beyond.** Layer payment posting, then claim submission. We do NOT recommend automating the medical coding step until you have 6+ months of clean data on your other workflows. Coding errors propagate downstream into denials, refunds, audits.