For medical-group patient access, referral management, laboratory and imaging operations, medication-access, prior-authorization, care-coordination, revenue-cycle, and practice-operations leaders who need a reliable answer to one question before work moves downstream: is this referral or order administratively ready, and who owns what is not?
Make Every Referral and Order Ready Before Work Moves Downstream
Turn disconnected referrals, laboratory and imaging orders, medication-authorization work, records requests, and payer responses into one evidence-linked operating record. Staff can see the current version, the exact missing item, the accountable owner, and the human decision still required—without treating a scheduled appointment, a sent fax, or a portal receipt as proof that the work is complete.
The Referral Readiness Control Team is a group of six bounded administrative AI agents. It extracts explicit facts, compares approved sources, prepares focused work, and reconciles returned evidence. It does not diagnose, choose a specialist or service, create or change an order, establish medical necessity, prescribe, obtain consent, assign a code, determine coverage, make an adverse decision, or establish legal or financial liability. Qualified clinicians and authorized payer, access, referral, laboratory, imaging, pharmacy, coding, compliance, privacy, revenue-cycle, and legal staff retain those decisions.
A referral can look scheduled while the order, payer path, and receiving packet still disagree
Before: work arrives through EHR tasks, fax queues, portal messages, phone notes, document folders, and specialty inboxes. Staff retype the same facts, use sent or scheduled as a stand-in for readiness, and learn late that an order is incomplete, a plan requires a different path, the destination did not accept the packet, or a result never returned. After: every referral or order has a versioned readiness record with separate component states, source links, named owners, focused exceptions, human approvals, transmissions, responses, and an explicit administrative closure condition. The workflow coordinates the evidence; it never promotes document presence into clinical, coverage, or payment certainty.
Before: intake starts several copies of the same work
A primary care practice may create an EHR referral, send supporting records, open an authorization task, notify a central team, and ask the destination to schedule. Laboratory, imaging, and medication work may follow different queues. Each copy can carry a different payer, destination, service description, note version, or owner, and staff cannot tell which one is controlling the next action.
Operational consequence
Teams repeat data entry and outreach, two people can work the same exception, and a later response may attach to a superseded version. Leaders see queue counts but cannot distinguish real progress from duplicated or stale work.
Before: order presence is mistaken for order completeness
A record can contain an order while still lacking an authenticated source, explicit requested service, ordering-professional context, destination-required information, supporting documentation, or a resolvable conflict. Different services and payers ask for different evidence. An administrative agent cannot infer a diagnosis, protocol, modality, test, medication, frequency, site, or clinical rationale from habit or nearby text.
Operational consequence
Laboratories, imaging sites, pharmacies, and specialists send the work back for clarification; authorization staff build a packet for the wrong version; scheduling proceeds on assumptions; and clinicians receive broad inbox messages instead of one answerable question.
Before: one eligibility or authorization answer is reused too broadly
Eligibility, benefits, network, a primary-care referral requirement, prior authorization, a pharmacy benefit, a payer determination, and final claim payment are different states. The applicable path may change with the product, service, site, destination, ordering professional, date, or liability context. A successful lookup or authorization identifier does not prove that every component is covered or aligned.
Operational consequence
Staff may send a packet to an out-of-scope channel, miss a plan-specific referral step, communicate more certainty than the evidence supports, or discover a coverage or coordination question only after the next team has acted.
Before: sent, scheduled, accepted, completed, and closed mean the same thing
A transmission receipt shows that a channel accepted a message, not that the destination accepted the referral. An appointment shows scheduling, not that the planned service occurred. A completed service does not prove that a report returned, a clinician reviewed it, follow-up was assigned, or the administrative and claim records agree. Each event needs its own source and owner.
Operational consequence
Outbound referrals disappear after transmission, unreturned consult information competes with new work, duplicate outreach reaches patients and destinations, and quality or revenue-cycle teams cannot explain which closure condition was actually met.
After: readiness becomes a component decision, not a colored badge
The future-state record shows which intake facts matched, which source controls the current version, which payer and destination checks apply, which documents are present, what a qualified human approved, what was transmitted, what came back, and what remains outside the software's authority. A ready status always names its scope and time.
Operational consequence
Patient access can route one focused exception, central operations can balance work by reason and owner, and local practices can answer requests without rebuilding the entire packet. A material change reopens the affected components instead of silently preserving an obsolete ready state.
A named agent team with visible decision boundaries
Each agent handles a defined part of the referral and order readiness workflow. Confidence signals and human escalation remain visible rather than being hidden behind an automation label.
Intake and Identity Reconciler
Creates the administrative case shell from approved inbound channels, links artifacts to the intended referral or order, and separates a plausible match from a verified match. It prevents downstream agents from acting on a document whose person, request, source, or version is uncertain.
Inputs
- Approved EHR referral and order work queues with source references and timestamps
- Authorized fax, document, portal-message, interface, and structured intake metadata
- Organization-approved identity-match fields and existing case identifiers
- Current source-system status, correction, cancellation, and duplicate indicators
Checks
- Whether required match fields agree under the organization's approved identity policy
- Whether the artifact belongs to a referral, lab order, imaging order, medication request, records request, or another queue
- Whether a newer, corrected, cancelled, or duplicate version already exists
- Whether the sender, receiving entity, timestamp, and source reference are explicit rather than inferred
Outputs
- Versioned case shell with source pointers and a provisional work type
- Duplicate, conflict, unmatched-artifact, and wrong-queue exceptions
- Focused intake task naming only the fields or source decision that need review
- Confidence method
- Scores each extracted field and the overall match separately using source reliability, exact agreement, duplicate candidates, version recency, and configured matching rules. A strong document extraction cannot compensate for an uncertain identity or case link.
- Low-confidence action
- Quarantines the artifact from downstream automation, preserves every candidate without merging records, and sends the smallest explainable comparison to authorized intake staff. It never selects the nearest-looking person or request to clear a queue.
- Human escalation
- Authorized patient-access or health-information staff resolve identity, duplicate, sender, wrong-record, and legal-record questions. Clinicians resolve whether an item expresses a clinical request; privacy and compliance staff handle inappropriate or misdirected disclosures.
Referral and Order Completeness Steward
Compares the current referral or order version with an organization-approved checklist for the specific work type and destination. It identifies explicit missing or conflicting administrative elements without inventing clinical meaning or changing what an authorized professional requested.
Inputs
- Current authenticated referral or order source and its version history
- Approved service-, destination-, and payer-specific administrative checklists
- Supporting note, result, history, form, and records references available in approved systems
- Clinician, laboratory, imaging, pharmacy, and specialty-office clarification responses
Checks
- Whether the source expresses an explicit request and identifies the authorized ordering or referring professional
- Whether required administrative elements are present, legible, current, and internally consistent
- Whether supporting records correspond to the same request version and relevant time context
- Whether a service, site, destination, frequency, or medication conflict requires clinician review
- Whether signature or authentication handling follows the current source-specific rule rather than a universal signature assumption
Outputs
- Component-level readiness checklist with evidence links and controlling version
- Focused clinician, practice, destination, or records request describing the explicit gap
- Hold reason that distinguishes missing evidence, conflicting evidence, and a decision outside software authority
- Confidence method
- Combines field-level extraction confidence with checklist version, exact source location, date alignment, conflict count, and whether the requested element is stated directly. Clinical implication, ambiguous abbreviations, and inferred intent always lower confidence regardless of document quality.
- Low-confidence action
- Marks only the affected component as unverified, shows the competing passages or missing source, and asks a qualified reviewer for a bounded decision. It does not complete a field from context, copy a prior request, or rewrite an order.
- Human escalation
- The ordering or referring clinician and authorized clinical staff decide clinical appropriateness, intent, requested service, rationale, and corrections. Laboratory, imaging, pharmacy, specialty, compliance, or coding experts interpret their own current requirements and terminology.
Coverage and Authorization Pathfinder
Routes each component to the current documented payer and benefit path, distinguishes medical from pharmacy authorization work, and assembles explicit requirements and response evidence for human review. It never treats eligibility or preauthorization as a promise of coverage or payment.
Inputs
- Current coverage response, payer, plan, product, benefit-channel, and effective-date references
- Versioned referral or order facts approved for administrative use
- Organization-approved payer, plan, program, portal, form, and policy sources with effective and retirement dates
- Submission acknowledgements, information requests, determinations, denials, and appeal-related correspondence
Checks
- Whether Original Medicare, Medicare Advantage, Medicaid, commercial, pharmacy-benefit, or liability logic is being applied to the correct context
- Whether network, referral, authorization, notification, and benefit questions remain separate
- Whether the source is current for the plan, product, service, site, provider, jurisdiction, and request date
- Whether a response applies to the same version and states an end date, condition, quantity, site, or other scope
- Whether a drug request has been incorrectly routed through a medical-service rule or API path
Outputs
- Payer-path record with source, scope, retrieval date, owner, and unresolved questions
- Human-review packet containing only approved facts and source-linked evidence
- Submission, additional-information, determination, expiration, mismatch, and escalation tasks
- Confidence method
- Rates payer and product identity, source scope, effective-date fit, fact-to-requirement mapping, response match, and channel acknowledgement independently. Any missing plan, ambiguous benefit channel, stale policy, or version mismatch prevents a high-confidence release.
- Low-confidence action
- Stops submission or release, displays the uncertain payer fact or competing source, and routes verification to authorized payer or authorization staff. It does not choose a plan, interpret silence as approval, or convert a portal status into final coverage.
- Human escalation
- Authorized eligibility, authorization, pharmacy-access, contracting, coding, billing, compliance, clinical, and appeal staff verify current requirements, communicate with the payer, decide what to submit, make representations, and select any reconsideration or appeal path.
Destination and Network Coordinator
Maintains the operational handoff to internal departments and outside specialists without selecting a clinically appropriate destination. It checks explicit directory and acceptance evidence, tracks referral requirements, and makes ownership visible across centralized and local teams.
Inputs
- Clinician-approved destination or destination-selection task
- Current provider-directory, roster, network, service, location, and access-channel references approved by the organization
- Internal specialty routing rules, external destination requirements, and receiving-team responses
- Patient-access communication status recorded through approved operational channels
Checks
- Whether a qualified human selected or approved the intended destination when clinical judgment is involved
- Whether the destination appears to offer the requested service at the relevant location and current operational context
- Whether network, accepting-new-referrals, referral-form, records, and scheduling facts have direct current evidence
- Whether internal ownership, financial-relationship, patient-choice, contract, or self-referral questions need qualified review
- Whether the receiving party accepted the packet rather than merely receiving a transmission
Outputs
- Destination-readiness record with source dates and unresolved verification items
- Internal or external handoff task with named sending and receiving owners
- Acceptance, redirection, unable-to-accept, patient-choice, and legal-review exceptions
- Confidence method
- Uses the recency and scope of the directory or roster source, direct receiving-party evidence, exact service and location match, patient-choice state, and presence of authorized destination approval. Search proximity, historic use, or organizational ownership never raises confidence by itself.
- Low-confidence action
- Leaves the destination unverified, preserves available options as evidence rather than recommendations, and assigns a verification task. It does not steer work to an internal or familiar entity to clear the queue.
- Human escalation
- Clinicians and authorized access staff make destination and patient-choice decisions. Contracting verifies network terms; compliance and counsel evaluate self-referral, inducement, conflict, and state-law questions; receiving organizations decide whether they accept the referral.
Evidence and Records Courier
Assembles the approved minimum packet, manages records requests and transmissions, and reconciles acknowledgements and returned artifacts. It keeps a pointer to authoritative records rather than creating a new clinical chart or treating successful delivery as acceptance.
Inputs
- Approved packet manifest from referral, order, payer, and destination checks
- Authorized document and record references from EHR, document-management, laboratory, imaging, and other approved systems
- Consent, authorization, disclosure, purpose, and recipient controls supplied by qualified staff where applicable
- Fax, direct message, interface, portal, mail, phone, and manual handoff acknowledgements
Checks
- Whether each packet item is the approved version and belongs to the same referral or order
- Whether the recipient, purpose, disclosure authority, and organization-specific minimum-necessary policy have been addressed
- Whether the channel is approved and the destination can receive the intended artifact type
- Whether a transmission acknowledgement, destination acceptance, information request, and returned report are being recorded as different events
- Whether a failed, partial, duplicate, or misdirected transmission requires containment and human response
Outputs
- Packet manifest with source pointers, version hashes or identifiers, approval, recipient, channel, and transmission state
- Focused missing-record, disclosure-review, failed-delivery, and acknowledgement-reconciliation tasks
- Non-destructive receipt and returned-artifact history for authorized operational review
- Confidence method
- Rates packet-item identity, approved-version match, recipient and channel validation, disclosure-control completion, acknowledgement type, and returned-artifact match separately. A technical success response never substitutes for recipient acceptance or legal disclosure authority.
- Low-confidence action
- Prevents automatic release, isolates a questionable or misdirected artifact, and asks authorized records or privacy staff to review the recipient, purpose, packet, and channel. It does not broaden a packet simply because more records are available.
- Human escalation
- Health-information management, privacy, compliance, legal, clinical, and referral staff authorize disclosures, decide minimum-necessary scope where applicable, resolve misdirected information, and determine the correct response to record or transmission incidents.
Closure and Exception Controller
Maintains the accountable queue after handoff, reconciles response evidence, reopens affected components after change, and applies an organization-approved administrative closure reason. It does not claim that a service occurred, a result was clinically reviewed, or payment is due without the responsible source and human confirmation.
Inputs
- All versioned component states, owners, human decisions, transmissions, and responses
- Destination acceptance, appointment, cancellation, unable-to-contact, service, report, and redirect evidence from approved sources
- Order corrections, payer changes, authorization changes, destination changes, and superseding documents
- Organization-approved closure, follow-up, escalation, retention, and reopening rules
Checks
- Whether each closure condition has its required direct evidence and authorized owner confirmation
- Whether accepted, scheduled, performed, reported, reviewed, billed, and paid states remain separate
- Whether an unresolved exception, newer version, or conflicting response invalidates administrative closure
- Whether a clinical result or consult note has an assigned qualified reviewer without interpreting its content
- Whether a case should close, remain pending, redirect, cancel, or reopen under the current approved rule
Outputs
- Reason-specific administrative closure record with source, date, scope, owner, and remaining downstream work
- Aging, no-response, returned-report, changed-request, mismatch, and unresolved-owner queues
- Override, reopen, false-release, false-hold, and rule-performance evidence for governance review
- Confidence method
- Requires direct evidence for every configured closure predicate, an accountable owner, version alignment, and no unresolved blocking conflict. It reports confidence by component and reason; elapsed time or repeated outreach never becomes proof of completion.
- Low-confidence action
- Keeps the case open with a precise reason, preserves the last valid state, and routes the disputed closure or reopening decision to the named owner. It does not auto-close work to improve queue age or completion metrics.
- Human escalation
- Referral, patient-access, clinical, laboratory, imaging, pharmacy, coding, billing, compliance, and legal leaders approve closure definitions and resolve disputed outcomes. Qualified clinicians alone confirm clinical review and next-care decisions; payers and authorized revenue-cycle staff retain coverage and claim authority.
The operating sequence, evidence by evidence
The sequence separates agent work from accountable human checkpoints so teams can see what moves forward, what stays pending, and why.
Central intake and local practice
Capture the request without starting duplicate work
The intake agent creates one provisional case from an approved source, identifies the work type, links artifacts, and checks for duplicates, corrections, cancellations, and unmatched documents. Nothing advances until the organization-defined identity and case match is strong enough or an authorized human resolves it.
Agent actions
- Index source, sender, received time, channel, and document or message version
- Classify the work as a referral, lab order, imaging order, medication request, records task, or review exception
- Compare against open and recently changed cases without merging uncertain candidates
- Create one focused intake exception for missing or conflicting match evidence
Evidence produced
- Versioned intake record with source pointers and provisional work type
- Duplicate and supersession comparison
- Human match decision or documented quarantine reason
Human checkpoint: Authorized intake or health-information staff confirm uncertain identity, request ownership, and wrong-record handling. A clinician decides whether ambiguous content constitutes an intended clinical referral or order.
Eligibility and coordination of benefits
Establish the current payer, plan, benefit channel, and liability context
Coverage staff verify the current context for the intended administrative work and keep eligibility, network, referral, medical authorization, pharmacy authorization, and payment expectations separate. The record includes the source and retrieval time rather than only a summarized yes or no.
Agent actions
- Associate the verified coverage response with the exact case version and relevant date context
- Route Original Medicare, Medicare Advantage, Medicaid, commercial, pharmacy-benefit, and liability questions to distinct source sets
- Flag ambiguous product, delegated entity, coordination-of-benefits, accident, or work-related context
- Create separate owners for network, referral, authorization, and pharmacy-benefit questions
Evidence produced
- Payer-path record with product, source, retrieval date, and scoped unknowns
- Coordination-of-benefits or liability review task when applicable
- Explicit statement that eligibility or preauthorization is not a payment guarantee
Human checkpoint: Authorized eligibility, payer, pharmacy, contracting, billing, and liability staff verify the actual product and current rules. Patients or representatives provide information only through approved operational channels, never a marketing form.
Referring practice and clinical operations
Validate the referral or order version and supporting evidence
The completeness agent evaluates explicit administrative elements against the approved checklist for the requested service and receiving context. It points to the gap and the controlling source; it never fills clinical intent, changes a service, or converts a prior request into a new order.
Agent actions
- Compare the authenticated source with checklist elements for the specific referral, laboratory, imaging, or medication lane
- Reconcile ordering professional, destination, service, site, date context, and supporting-record versions
- Separate missing evidence from contradictory evidence and from a decision only a clinician can make
- Draft one source-linked clarification request for authorized human review
Evidence produced
- Component readiness record with present, missing, conflicting, not-applicable, and human-decision states
- Source passages or locations supporting every administrative check
- Clinician-approved correction, clarification, or explicit hold
Human checkpoint: The authorized ordering or referring professional confirms clinical intent, appropriateness, requested service, rationale, and any correction. Laboratory, imaging, pharmacy, coding, and specialty experts decide technical or policy meaning in their domains.
Referral and authorization operations
Resolve destination, network, authorization, and packet scope
The pathfinder and destination coordinator assemble the current requirements for the approved destination and payer context. The records courier prepares a manifest of approved evidence, while separate queues preserve decisions about patient choice, network, authorization, disclosure, internal routing, and financial relationships.
Agent actions
- Verify source scope and effective date for payer, plan, service, site, provider, destination, and benefit channel
- Compare the current case version with referral, authorization, destination, and records requirements
- Prepare a packet manifest without adding broader records or inferred clinical claims
- Route self-referral, patient-choice, network, delegated-entity, privacy, and legal questions to qualified owners
Evidence produced
- Source-linked payer and destination requirement record
- Human-approved submission or handoff packet manifest
- Decision log for destination, network, disclosure, and authorization exceptions
Human checkpoint: Authorized clinical, patient-access, authorization, privacy, compliance, contracting, coding, and legal staff approve the destination, representation, packet, submission, and any high-risk exception before release.
Records and referral coordination
Transmit through an approved channel and reconcile what came back
The courier records the exact packet and approved channel, then distinguishes technical transmission, payer receipt, destination acceptance, request for more information, and returned evidence. Failed or partial handoffs keep a named owner and a tested manual fallback.
Agent actions
- Record packet version, recipient, channel, sender authority, and approved release
- Capture acknowledgements without interpreting silence or technical success as acceptance
- Match returned requests, determinations, acceptance messages, and reports to the correct version
- Route failed, partial, duplicate, misdirected, or unsupported channel events to manual recovery
Evidence produced
- Immutable transmission manifest and acknowledgement classification
- Destination or payer response linked to the controlling case version
- Manual recovery, incident, or additional-information task with accountable owner
Human checkpoint: Authorized staff review the actual response, decide whether further information may be released, correct misdirected information, and make any representation to a payer or receiving organization.
Patient access and service-line operations
Release only the component that is ready, then watch for change
A scoped ready state tells downstream teams exactly what was verified, from which sources, for which version, and what remains unresolved. Scheduling or submission may proceed only under organization-approved release rules. A material change creates a new version and reopens the affected checks.
Agent actions
- Evaluate configured administrative release predicates without collapsing component states
- Display unresolved non-blocking items and all decisions that remain with humans
- Monitor for payer, plan, service, site, destination, professional, date, or source-version changes
- Reopen only dependent checks and notify the current accountable owner
Evidence produced
- Scoped ready, pending, or blocked decision with evidence and human approval
- Change-impact map showing what was reopened and why
- Override and manual-proceed record that preserves the prior agent recommendation
Human checkpoint: Authorized operations leaders approve release criteria and any manual proceed decision. Clinicians retain service and safety authority; payer and revenue-cycle staff retain coverage, authorization, coding, and claim authority.
Referral closure and revenue-cycle follow-up
Close the administrative loop without claiming a clinical outcome
Closure uses a reason-specific evidence rule: accepted, redirected, unable to accept, patient choice recorded, cancelled by an authorized owner, service evidence received, report returned, or another approved administrative endpoint. Appointment, service, report, clinical review, claim, and payment remain separate.
Agent actions
- Reconcile destination, appointment, service, report, review-assignment, and downstream claim evidence as distinct events
- Apply the current approved closure definition and prevent aging-based auto-closure
- Assign returned clinical information to a qualified reviewer without reading it as a care decision
- Keep unresolved records, coding, claim, liability, and follow-up work visible after referral closure
Evidence produced
- Reason-specific closure record with source, owner, scope, and remaining dependencies
- Returned-report and downstream-work assignment
- False-release, false-hold, reopen, override, and unresolved-owner audit evidence
Human checkpoint: Referral and clinical leaders approve closure and follow-up. A qualified clinician confirms clinical review and next steps; billing, coding, payer, and legal staff decide claim and liability matters.
Separate payer paths instead of one universal rule
Coverage, notice, authorization, and documentation requirements vary by program, plan, jurisdiction, service, and date. These paths show where teams must verify current authoritative instructions.
medicare
Original Medicare: do not turn general access into a universal no-check path
Medicare's current comparison says Original Medicare generally does not require a referral to see a specialist, but that does not remove service-specific order, documentation, coverage, enrollment, billing, or coordination checks. CMS's current laboratory guidance ties diagnostic lab testing to a treating professional's order or documented intent and supporting records. Part D drug-plan rules, physician self-referral rules, and local or national coverage context remain distinct from a specialist-referral question.
- Identify whether the work is a specialist referral, diagnostic laboratory or imaging order, outpatient drug request, DME or another service before applying a checklist.
- Keep an authenticated order, intent-to-order evidence, medical-record support, ordering/referring professional, and claim-related documentation as separate source elements; do not invent a universal signature rule.
- Treat Medicare Part D medication prior authorization and exception work as a separate drug-plan path rather than a medical-service authorization.
- Route financial-relationship and designated-health-service questions to compliance or counsel; operational software cannot decide whether an exception to the physician self-referral law applies.
Human handoff: Medicare, laboratory, imaging, pharmacy, coding, compliance, and clinical specialists verify the current service-specific source and make all clinical, legal, coverage, coding, submission, and appeal decisions. The agent may show evidence and gaps, not certify Medicare compliance or payment.
Sources for this path: Medicare.gov, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services, Medicare.gov, Centers for Medicare & Medicaid Services
medicare advantage
Medicare Advantage: verify the plan's network, referral, and determination path
Medicare Advantage is not Original Medicare with a different card. Medicare's comparison notes that a plan may use networks and may require a specialist referral. CMS defines an organization determination as a plan decision about authorization or payment for an item or service, member cost, or a quantity limit. Current plan, benefit, referral, authorization, drug, and appeal sources must be verified separately.
- Resolve the exact MA organization, plan, product, network, delegated entity, service, destination, site, and date before routing work.
- Keep primary-care referral, medical prior authorization, Part D drug rules, organization determination, claim payment, and appeal evidence as separate states.
- CMS's prior-authorization interoperability rule has a defined impacted-payer scope, operational dates, API dates, and an exclusion for drug prior authorizations; it does not prove that a particular plan API or integration is available.
- A portal receipt, authorization identifier, or favorable determination has its own scope and does not guarantee network status, patient cost, claim coding, or payment.
Human handoff: Authorized MA, authorization, pharmacy, contracting, coding, clinical, and appeal staff verify the current plan material, request the appropriate determination, interpret the response, communicate with affected parties, and choose any reconsideration or appeal path.
Sources for this path: Medicare.gov, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services, Medicare.gov
medicaid
Medicaid: keep state, delivery system, plan, program, and service together
Medicaid is administered through state programs that may use fee-for-service, managed care, primary-care case management, waivers, and program-specific arrangements. Federal managed-care authorities preserve material state flexibility. A referral, PCP assignment, network, prior-authorization, pharmacy, transport, form, or appeal path from one state or plan cannot be copied into another without current verification.
- Version the state, program, delivery system, managed-care plan, product, service, destination, assigned provider context, and effective date with every source.
- Distinguish Medicaid fee-for-service from managed care and separate medical, pharmacy, behavioral-health, dental, transport, and other carved or delegated channels when the current program requires it.
- CMS-0057-F applies defined prior-authorization provisions to specified Medicaid and CHIP payers and excludes drug prior authorizations from those provisions; it does not create one national portal, form, documentation list, or integration.
- Route enrollment, eligibility, program eligibility, clinical need, adverse decisions, fair-hearing, and state-law questions to qualified humans rather than deriving them from a checklist.
Human handoff: Authorized state-program, plan, pharmacy, clinical, referral, coding, and appeal staff verify current program and contract materials, decide submissions and representations, and handle adverse actions or fair-hearing work. The software abstains when state or plan scope is incomplete.
Sources for this path: Medicaid.gov, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services
commercial
Commercial coverage: read the actual product and contract, not a market label
Commercial and employer coverage can differ by issuer, product, funding arrangement, network, pharmacy benefit, employer plan, state law, and contract. HealthCare.gov explains that preauthorization may be required but is not a promise that the plan will cover cost. Department of Labor claims guidance applies to ERISA-covered benefit plans within its scope, not every commercial arrangement.
- Verify the exact issuer, product, network, group or individual context, funding and administrator details available to authorized staff, service, destination, site, pharmacy channel, and date.
- Keep eligibility, benefits, referral, network, prior authorization, predetermination, estimate, claim adjudication, patient responsibility, and appeal as separate evidence states.
- CMS clarifies that CMS-0057-F reaches Qualified Health Plans on the Federally-facilitated Exchanges within its defined scope, not all commercial or employer plans; voluntary adoption is not a universal requirement or integration promise.
- Use the governing plan document, contract, payer instruction, state requirement, and authorized advice for the real work rather than a generic commercial-payer checklist.
Human handoff: Authorized payer, contracting, benefits, pharmacy, coding, billing, clinical, compliance, and legal staff verify current terms, make representations, communicate estimates carefully, and select any review or appeal route. The software cannot promise coverage, cost, or payment.
Sources for this path: HealthCare.gov, U.S. Department of Labor Employee Benefits Security Administration, Centers for Medicare & Medicaid Services
workers comp auto liability
Workers' compensation, auto, and liability: preserve the event and payer order
An injury-related referral or order may require an employer, carrier, adjuster, claim, jurisdiction, authorization, causal, legal, or coordination path that is separate from ordinary health coverage. CMS explains that workers' compensation, no-fault, or liability coverage may pay before Medicare and that Medicare conditional-payment and recovery questions can arise. That federal context does not replace state rules or carrier instructions.
- Create a distinct liability review from the minimum approved event, claim, carrier, jurisdiction, represented-party, service, and date context; do not copy unverified allegations into unrelated clinical or marketing systems.
- Separate health-plan eligibility, liability acceptance, workers' compensation authorization, no-fault benefits, Medicare Secondary Payer status, conditional payment, settlement, and final responsibility.
- Preserve the exact source and authorized owner for every carrier communication, deadline, request, denial, and payment-order decision.
- Do not infer causation, compensability, legal representation, coverage, settlement effect, or which party must pay from diagnoses, notes, or a generic accident flag.
Human handoff: Qualified workers' compensation, auto, liability, Medicare Secondary Payer, billing, compliance, clinical, and legal professionals verify jurisdiction and case facts, communicate with authorized parties, and decide causation, compensability, recovery, settlement, authorization, and billing strategy.
Sources for this path: Centers for Medicare & Medicaid Services
Govern readiness as evidence, not as an AI opinion
A useful control model makes source, version, scope, confidence, ownership, human authority, and failure behavior visible. It also keeps operational PHI inside approved environments, leaves marketing analytics free of case data, and gives the group a safe manual path when an interface, source, or model is unavailable.
Versioned source registry
Every payer, plan, program, destination, laboratory, imaging, pharmacy, disclosure, closure, and internal checklist source carries scope, owner, effective date, review date, retirement date, and supersession history. Expired, conflicting, or unscoped sources create a hold rather than a guessed rule.
Separated states and decision rights
Identity, order, clinical, coverage, network, referral, authorization, destination, scheduling, service, report, coding, claim, payment, and closure states remain distinct. The control matrix names which agent may prepare evidence and which qualified human may approve, override, submit, sign, or decide.
Field-level confidence and mandatory abstention
Confidence is reported for the exact extraction, match, source fit, and response link—not as one opaque case score. Uncertain identity, clinical meaning, destination, payer scope, disclosure authority, adverse action, or legal question triggers abstention and an evidence-specific human task.
Privacy, access, retention, and disclosure controls
Role, purpose, environment, minimum-necessary policy where applicable, authentication, authorization, encryption, retention, deletion, incident response, and vendor duties are configured by the organization. HHS notes exceptions to the HIPAA minimum-necessary standard, including treatment disclosures; privacy professionals decide applicability rather than a universal software rule.
Independent validation by lane and failure type
Before release, teams test routine, missing, conflicting, stale, changed, duplicate, wrong-person, wrong-payer, wrong-destination, unsupported-channel, downtime, and malicious-input scenarios against independently recorded human decisions. False releases, false holds, bad matches, missed changes, and wrong-owner routes are reviewed separately.
Manual operation, rollback, and non-destructive override
Every automated action has a documented manual path, safe stop, queue owner, rollback condition, and recovery test. Human overrides require a reason and evidence, preserve the earlier recommendation, never erase history, and feed controlled review rather than silently retraining production behavior.
Referral choice and financial-relationship review
Internal and external destinations are never ranked from organizational preference alone. Patient choice, clinician judgment, network facts, access, contracts, financial relationships, federal and state rules, and applicable exceptions receive their own evidence and qualified review before a high-risk routing decision.
- Human authority
- Authorized humans retain identity correction, clinical assessment, order creation or change, referral selection, patient choice and consent, destination approval, disclosure, medical necessity, payer representation, authorization, prescription, coding, claim, adverse action, appeal, self-referral, liability, and closure decisions. An agent may prepare, compare, route, and abstain; it cannot sign, attest, direct care, conceal uncertainty, or override a qualified decision maker.
- Audit trail
- For every state change, retain the source reference, source version, retrieval time, extracted field, confidence, rule or checklist version, agent action, human reviewer, approval or override, transmission manifest, acknowledgement, response link, reopen reason, closure reason, and downstream owner. Logs should support access review, incident investigation, reproducible validation, data correction, source retirement, and rollback without duplicating unnecessary clinical content.
- Data boundary
- Operational PHI, referral and order content, payer-member values, clinical notes, authorization data, destination communications, portal credentials, reports, claims, and free text stay only in approved operational systems under the group's privacy and security controls. Credentials and secrets stay in managed credential stores. After publication, GA4 may receive only approved route, page-family, specialty slug, workflow slug, content-cluster, CTA-label, and CTA-location values. Search Console review stays aggregate and page-scoped to query, clicks, impressions, CTR, and average position; neither system may receive patient, referral, order, payer, authorization, document, appointment, portal, service, or claim values.
Connect the handoffs without pretending every system is integrated
The workflow can sit across approved systems while each remains authoritative for its own record. Integration scope is established lane by lane: source, field, direction, timing, acknowledgement, write-back, permissions, downtime, error recovery, vendor dependency, and cost. No EHR, payer, laboratory, imaging, pharmacy, referral network, portal, or clearinghouse connection is promised by this page.
EHR and practice-management systems
Information in scope
Approved referral and order references, clinician and practice context, source-document pointers, task status, human approvals, appointment context, and returned-report assignments may enter or receive scoped status updates.
Boundary
The EHR remains the clinical record and the authorized professional remains responsible for orders, diagnoses, medications, consent, results, and care decisions. Every read and write path, supported field, latency, acknowledgement, and vendor permission requires local validation.
Fax, direct messaging, referral platforms, HIEs, and document intake
Information in scope
Source, sender, recipient, received time, document type, transmission state, acknowledgement, and artifact reference can support intake, packet delivery, response matching, and records retrieval.
Boundary
Optical extraction and message delivery can be incomplete or misdirected. A technical receipt is not destination acceptance, identity proof, disclosure authority, or clinical review. Unsupported channels stay manual and every lane needs a tested recovery procedure.
Laboratory, imaging, specialty, and result-delivery systems
Information in scope
Destination requirements, approved order references, appointment or access state, request-for-information events, service-status evidence, report-return metadata, and review assignments can be reconciled where authorized interfaces exist.
Boundary
The agent does not choose or change tests, modalities, protocols, contrast, medications, specialists, or care. It does not interpret a result or confirm service from an ambiguous status. The performing entity and qualified clinicians retain technical, safety, and clinical authority.
Eligibility, payer, clearinghouse, authorization, and portal channels
Information in scope
Coverage-response references, payer and product context, scoped requirements, submission manifests, acknowledgements, requests for information, determinations, dates, conditions, and response sources may be linked to the case.
Boundary
Access depends on payer support, contracts, delegation, credentials, transaction type, and current technical capability. Screen automation and shared credentials are not assumed. A response is evidence for qualified review, not a universal coverage or payment guarantee.
Electronic prescribing, pharmacy, and medication-access workflows
Information in scope
Medication-request references, prescriber-owned clarification tasks, medical-versus-pharmacy benefit routing, prior-authorization status, exception correspondence, and human-approved follow-up may be coordinated.
Boundary
The software does not prescribe, select therapy, substitute a medication, infer an indication, create clinical rationale, sign, or make a formulary or exception decision. Drug authorization rules and technical paths are not assumed to match medical-service authorization.
Scheduling, contact-center, and patient-communication tools
Information in scope
Readiness scope, authorized outreach task, destination acceptance, appointment state, cancellation, redirection, and unable-to-contact evidence can inform the accountable queue without placing message content in public analytics.
Boundary
Scheduling does not prove order completeness, authorization, coverage, service, report return, or closure. Communication requires approved identity, consent, accessibility, language, channel, and privacy practices, with human help available for exceptions.
Revenue-cycle, coding, claim, and work-queue systems
Information in scope
Administrative readiness, authorization scope, order and report references, open exceptions, downstream coding assignment, claim status, and denial correspondence may be connected to prevent the referral record from disappearing at handoff.
Boundary
The workflow does not assign codes, calculate a final balance, create claim facts, determine coverage, post payment, or decide an appeal. Revenue-cycle systems and qualified staff remain authoritative; interface errors and delayed responses must be reconciled visibly.
Model released administrative capacity with your own observed inputs
Use unique referral and order cases that actually pass through the selected lane, observe the administrative minutes genuinely removed rather than shifted, and apply a finance-approved loaded labor rate. The model excludes clinical work, necessary patient communication, required payer work, new quality-control time, vendor charges, and work transferred to another team. It estimates labor-capacity value, not cash savings, revenue, approvals, denial reduction, faster care, head-count reduction, or a customer result.
Monthly unique referral and order cases
2,400 cases
Illustrative planning input only. Replace it with a deduplicated count from the one pilot lane; exclude reopened tasks, documents, status messages, and duplicate referrals so volume is not inflated.
Administrative minutes actually removed per case
7 minutes
Illustrative planning input only. Establish it with before-and-after observation and subtract review, exception, monitoring, and downstream time added elsewhere. Do not include clinical time or necessary communication.
Finance-approved loaded labor rate
34 dollars per hour
Illustrative planning input only. Finance should supply the relevant wage, benefits, and overhead convention and decide whether released capacity has economic use; this is not a quoted price or payroll assumption.
Formula
2,400 unique referral and order cases × 7 administrative minutes saved ÷ 60 × $34 loaded labor rate = $9,520 of illustrative monthly labor-capacity value.
Illustrative result
$9,520 in illustrative monthly labor-capacity value before software subscription, usage, interface, third-party, validation, governance, and change-management costs. This is not booked savings or revenue; the group must show where the released capacity goes and compare total cost with its own baseline.
Illustrative planning model—not a customer result, guarantee, or quote.
Illustrative workflow examples
These examples explain process behavior. They are not customer stories, measured outcomes, clinical advice, or promises of coverage.
Illustrative example
An outbound specialty referral arrives with two different packet versions
This hypothetical process example contains no patient or customer data. A primary care office creates an outbound specialty referral, while a later clinician-approved update changes one supporting document. The central referral queue has the original packet, the destination has only a transmission receipt, and the current network and acceptance states still require verification.
- The Intake and Identity Reconciler links both packets to the same provisional case, preserves their source times, and marks the later document as a possible superseding version without deleting the first.
- The Referral and Order Completeness Steward identifies the exact component that differs and asks authorized clinical staff which version reflects the intended referral; it does not interpret the change clinically.
- The Coverage and Authorization Pathfinder keeps network, primary-care referral, medical authorization, and benefit states separate and routes any uncertain plan fact to payer staff.
- The Destination and Network Coordinator records the human-approved destination and checks direct current acceptance evidence rather than assuming a transmission receipt is acceptance.
- After qualified staff approve the controlling version and packet, the Evidence and Records Courier sends through an approved channel and reconciles the receiving response.
- The Closure and Exception Controller leaves the referral open until an approved administrative closure condition has direct evidence; it does not claim that an appointment or clinical service occurred.
Illustrative outcome: The illustrative outcome is an explainable handoff with one controlling packet, separate payer and destination states, and named human decisions. It is not a customer result, approval, appointment, completed service, denial avoided, or promise of clinical or financial benefit.
Illustrative example
A diagnostic order is present, but the receiving checklist exposes a conflict
This hypothetical laboratory or imaging lane shows process behavior only and contains no patient, order, or customer values. The EHR has an authenticated order reference and supporting record, but the requested service wording and the receiving destination's administrative intake record do not align clearly enough for release.
- The Completeness Steward points to the two explicit source locations and labels the issue as conflicting evidence rather than missing documentation.
- The agent holds only the affected order-readiness component and sends a focused question to the authorized ordering team; it does not choose a test, modality, protocol, diagnosis, frequency, or site.
- A qualified clinician confirms or corrects the intended request in the authoritative system, creating a new version rather than editing the agent's extracted copy.
- The Pathfinder reevaluates only payer requirements affected by the new version and keeps eligibility, authorization, network, and payment expectations separate.
- The Courier prepares the minimum approved packet and records transmission and acknowledgement as distinct events.
- The closure agent assigns any returned report to a qualified clinical reviewer but does not interpret the result or decide follow-up.
Illustrative outcome: The illustrative outcome is a corrected, source-linked administrative order path with a clinician-owned decision and a visible downstream review assignment. It does not demonstrate medical necessity, safety, coverage, payment, faster service, or a customer outcome.
Illustrative example
A medication request moves from medical to pharmacy-benefit review
This hypothetical medication-access example contains no prescription, member, diagnosis, or customer data. Intake initially routes a request with general authorization work, but current approved evidence indicates that a separate pharmacy-benefit path and prescriber-owned supporting statement may apply.
- The Pathfinder identifies the benefit-channel uncertainty and prevents medical-service authorization rules from being reused as a drug rule.
- Authorized pharmacy-access staff verify the current plan and process, while the agent assembles only explicit existing evidence and source links.
- The software does not infer an indication, draft clinical rationale as fact, substitute medication, select step therapy, or request an exception on its own.
- The prescriber or other authorized professional reviews any supporting statement, and payer-facing staff approve the representation and channel.
- A response is linked to the same request version and separated into receipt, information request, determination, and any human-selected follow-up.
Illustrative outcome: The illustrative outcome is correct administrative routing and preserved authority, not medication approval, clinical appropriateness, payer compliance, a shortened turnaround time, or a promise that the drug will be covered or dispensed.
Adopt one bounded lane, prove the controls, then expand deliberately
Begin with one referral or order lane whose sources, owners, exceptions, and closure rules can be observed end to end. Implementation should expose integration limits, manual work, false releases, false holds, source maintenance, and total cost before additional practices, specialties, payers, services, destinations, or automated actions are added.
Define the lane and establish the before-state
- Select one work type, practice cohort, destination pattern, payer scope, and accountable executive without using patient values in the project brief
- Map current intake, clinical, payer, destination, records, scheduling, closure, coding, and billing handoffs with system and human owners
- Define readiness components, blocking versus non-blocking exceptions, closure reasons, manual fallbacks, and prohibited agent decisions
- Measure unique case volume, administrative touches, observed minutes, queue age, duplicate work, reopen reasons, and exception distribution using approved operational methods
- Document current source maintenance, privacy, security, legal, contracting, vendor, and total-cost assumptions
Exit criteria: The group approves a non-PHI process map, source inventory, decision-rights matrix, baseline method, illustrative value formula inputs to replace, success and stop criteria, risk register, manual operating procedure, and named owners for clinical, payer, privacy, compliance, legal, security, and operational decisions.
Configure sources, checklists, integrations, and boundaries
- Create versioned referral, order, payer, destination, records, transmission, closure, and exception schemas for the selected lane
- Load only approved sources with scope, effective and retirement dates, owners, and review cadence
- Configure role-based access, data minimization, retention, audit, credential, incident, downtime, and deletion controls
- Validate every read, write, acknowledgement, error, retry, duplicate, and rollback path against the real vendor and environment
- Use synthetic test data first and keep marketing analytics, SEO evidence, logs, screenshots, and support artifacts free of PHI and credentials
Exit criteria: Every field and action has an authoritative source, owner, permission, failure path, and test result; every high-risk decision requires an authorized human; unsupported channels remain documented manual work; and security, privacy, compliance, clinical, operational, and technical owners approve entry into shadow validation.
Run shadow validation and challenge the failure modes
- Compare agent outputs with independently recorded human decisions without allowing the agents to release, submit, close, or overwrite production work
- Test routine, missing, conflicting, stale, changed, duplicate, wrong-person, wrong-payer, wrong-destination, pharmacy-versus-medical, downtime, and unsupported-channel scenarios
- Review false releases, false holds, incorrect matches, missed changes, wrong-owner routes, disclosure risks, and closure errors separately rather than relying on one average score
- Confirm evidence-specific confidence, mandatory abstention, override history, rollback, manual recovery, and incident response
- Recalculate the transparent cases × minutes saved ÷ 60 × loaded labor rate model with observed time and all added review or governance effort
Exit criteria: Independent reviewers approve performance by action and risk class, no unresolved critical failure remains, mandatory abstention works, manual operation and rollback have been exercised, source owners accept maintenance duties, and the executive owner accepts the documented residual risk and full cost of a bounded pilot.
Release gradually, monitor, and decide whether to expand
- Start with the lowest-risk approved actions and small authorized user groups while preserving human review for every defined high-risk decision
- Monitor source freshness, queue health, false releases, false holds, overrides, reopens, interface failures, privacy events, support demand, and manual fallback use
- Hold regular clinical, operational, payer, privacy, compliance, legal, security, and revenue-cycle review with authority to pause or roll back
- Compare observed released capacity and total cost with the approved baseline without converting capacity into guaranteed savings or staff reduction
- Require a new scope, source, interface, risk, validation, support, and cost review before adding another specialty, service, payer, destination, or automated action
Exit criteria: The group can operate safely during downtime, reproduce every decision, maintain current sources, resolve incidents, support users, explain total cost, and show useful administrative capacity with acceptable error and override patterns. Expansion proceeds only through a separately approved evidence package.
Authoritative sources used for coverage context
Sources support the cited coverage and administrative context. Teams must confirm the current policy, contract, jurisdiction, and effective date for each real case.
Lab Test Order Requirements
Centers for Medicare & Medicaid Services · government · reviewed
Current CMS provider guidance describing who must order covered diagnostic laboratory tests, documentation of medical necessity and intent to order, acceptable order evidence, communication methods, record responsibilities, and the need to consult current manuals and regulations rather than apply a universal signature shortcut.
Complying With Medical Record Documentation Requirements
Centers for Medicare & Medicaid Services · government · reviewed
Official CMS Medicare Learning Network material on sufficient, authenticated, accurate, and supportive medical-record documentation, including order or intent-to-order evidence when required. It supports evidence-linked checks without allowing software to certify coverage or documentation adequacy universally.
Compare Original Medicare and Medicare Advantage
Medicare.gov · government · reviewed
Official comparison of provider-choice, network, primary-care-doctor, and specialist-referral context in Original Medicare and Medicare Advantage. It supports the distinction that Original Medicare generally does not require a specialist referral while an MA plan may, without implying that all services, plans, orders, or coverage questions follow one rule.
Medicare Advantage Organization Determinations
Centers for Medicare & Medicaid Services · government · reviewed
Official CMS explanation of organization determinations involving authorization or payment for an item or service, member cost, or quantity limits, plus who may request standard or expedited determinations. It supports keeping plan decisions, request authority, and appeal work with qualified humans.
Medicare Part D Drug Plan Rules
Medicare.gov · government · reviewed
Official Medicare explanation of drug-plan prior authorization, formulary and coverage rules, exceptions, prescriber supporting statements, and step-therapy context. It supports separating medication-access work from medical-service prior authorization and preserving prescriber and plan authority.
Filing a Claim for Your Health Benefits
U.S. Department of Labor Employee Benefits Security Administration · government · reviewed
Official guidance for claims and reviews under ERISA-covered health benefit plans, including consulting the governing plan and claims procedure, retaining evidence, reviewing adverse-benefit information, and using the applicable appeal process. Its scope does not include every commercial arrangement.
Medicare Secondary Payer
Centers for Medicare & Medicaid Services · government · reviewed
Current official CMS overview of circumstances in which another entity may pay before Medicare, including certain group-health, workers' compensation, no-fault, and liability situations, plus conditional-payment and recovery context requiring qualified case-specific review.
Minimum Necessary Requirement
U.S. Department of Health and Human Services · government · reviewed
Official HHS guidance on the HIPAA Privacy Rule minimum-necessary standard, organization-specific policies, role-based access context, reasonable reliance, and stated exceptions including provider treatment disclosures. Privacy professionals must determine applicability to each use, request, and disclosure.
Physician Self-Referral
Centers for Medicare & Medicaid Services · government · reviewed
Current official CMS overview of the federal physician self-referral law, covered financial relationships, designated health services, billing prohibition, and applicable exceptions. It supports routing financial-relationship questions to compliance and counsel rather than allowing software to prefer or approve a destination.
Primary care and multispecialty groups workflow FAQs
Open a question to review the answer. The disclosure controls use native browser behavior and remain available by keyboard without page JavaScript.
What does healthcare referral management software mean on this page?
It means a governed administrative workflow that links intake, identity and case matching, the current referral or order, payer and benefit path, destination, network, authorization, records, transmission, acceptance, change, follow-up, and reason-specific closure. It is not recruiting referral software, a generic scheduling tool, an EHR replacement, a clinical decision system, a payer determination engine, or proof that care occurred.
Will this replace our EHR, practice-management system, referral platform, or scheduling tool?
No replacement is assumed. Those systems can remain authoritative while the workflow coordinates approved facts, evidence pointers, dependencies, and accountable tasks across them. Every proposed interface, field, direction, write-back, acknowledgement, vendor permission, error path, downtime process, and cost must be validated locally. Unsupported systems or channels remain explicit manual work rather than an invented integration.
Can the AI decide which specialist, laboratory, imaging service, or medication is appropriate?
No. It can show what an authenticated source states, compare explicit administrative elements, identify a conflict, and route a focused question. It cannot diagnose, recommend a destination, choose or modify a test or modality, set a protocol, select therapy, infer an indication, prescribe, establish medical necessity, obtain consent, or replace patient choice and clinician judgment. Qualified professionals keep those decisions.
Does a ready status guarantee authorization, coverage, network status, cost, or payment?
No. Ready means only that the configured administrative evidence and required human approvals support a named handoff for a defined version and time. Eligibility, benefits, network, a PCP referral, medical authorization, pharmacy authorization, destination acceptance, appointment, service, claim adjudication, patient responsibility, and payment are separate events. New evidence can reopen an earlier readiness decision.
How are laboratory and imaging order-completeness checks kept within administrative boundaries?
The checklist can verify explicit source, ordering-professional, requested-service, destination, authentication, supporting-record, and version elements defined by approved current sources. It can point to missing or conflicting evidence. It cannot choose a test, modality, contrast, protocol, diagnosis, site, frequency, or clinical rationale; interpret a result; or create a corrected order. Authorized clinicians and performing organizations decide those matters.
How does medication prior authorization differ from medical-service authorization?
The benefit channel, plan rules, formulary, prior authorization, step therapy, exception request, prescriber statement, response, and appeal path may differ from a medical-service request. The workflow keeps those sources and states separate. It does not reuse CMS-0057-F medical-service API claims for drug work, infer clinical rationale, draft a prescriber statement as fact, substitute a drug, or request an exception without authorized review.
What counts as closing the referral loop?
The medical group defines reason-specific administrative closure and the evidence each reason requires. Transmission, acceptance, scheduled, cancelled, redirected, unable to accept, service evidence, returned report, assigned clinical review, claim, and payment are not interchangeable. The agent cannot close work because it is old or because outreach was repeated; a named human owner approves disputed closure and any clinical follow-up.
Can the workflow route referrals to our employed or affiliated specialists automatically?
Not merely because they are internal, familiar, available in a directory, or financially connected. A qualified process must preserve clinician judgment, patient choice, current network and access evidence, contracts, destination acceptance, financial relationships, federal and state requirements, and applicable exceptions. Compliance or counsel decides self-referral and other legal questions; software cannot clear them or steer work to improve leakage metrics.
How are Original Medicare, Medicare Advantage, Medicaid, and commercial payer paths kept separate?
Each source is tagged with payer or program, plan or product, state and jurisdiction, service, destination, owner, effective date, review date, and retirement date. Missing scope creates a hold. Original Medicare's general specialist-referral context is not copied into Medicare Advantage; one state's Medicaid rule is not copied to another; and CMS rules for FFE Qualified Health Plans are not presented as applying to all commercial or employer coverage.
How does the workflow handle fax, portal, phone, and unsupported destinations?
Approved fax and message artifacts can be indexed with source and confidence, phone work can be recorded through an authorized structured process, and portal steps can be supervised where contracts and credentials permit. A sent fax, placed call, portal success screen, or interface acknowledgement is not acceptance. Failed or partial handoffs keep an owner, and each production lane has a tested manual fallback.
How do we evaluate accuracy, abstention, and human override before production?
Run shadow validation against independently recorded human decisions using routine, incomplete, conflicting, stale, changed, duplicate, wrong-person, wrong-payer, wrong-destination, medical-versus-pharmacy, downtime, and unsupported-channel scenarios. Review false releases, false holds, wrong matches, missed changes, and wrong-owner routing separately. Require field-level evidence, mandatory abstention, reasoned override, non-destructive history, manual operation, and tested rollback.
How are PHI, credentials, GA4 measurement, and Search Console data separated?
Operational PHI stays in approved systems under access, purpose, retention, deletion, audit, and incident controls; secrets stay in approved credential stores. Marketing GA4 may receive only approved route, page-family, specialty, workflow, content-cluster, CTA-label, and CTA-location context. Search Console review remains aggregate and page-scoped. Neither receives patient, referral, order, payer-member, authorization, document, appointment, portal, service, report, or claim values.
What does implementation cost, and is the referral software free?
No, the software is not free. $0 implementation fee. $0 customization charges. Those terms apply to the agreed implementation and customization scope. Software subscription and usage charges are separate and still apply. Third-party data, payer, clearinghouse, pharmacy, messaging, interface-vendor, license, storage, migration, validation, travel, internal change, and out-of-scope work may also have separate written costs.
How should a primary care or multispecialty group calculate potential value?
Use the group's own deduplicated monthly case count, observed administrative minutes actually removed per case, and finance-approved loaded labor rate: cases × minutes saved ÷ 60 × loaded labor rate. Subtract new review, monitoring, exception, and downstream time plus all software and third-party costs. Treat the output as released-capacity planning, not promised cash savings, revenue, approvals, denial reduction, faster care, or staff reduction.
Bring one referral or order lane to a working session
Choose one practice cohort, work type, payer scope, destination pattern, and non-PHI process map. We will identify the before-and-after handoffs, readiness components, authoritative sources, human decisions, integration boundaries, manual fallback, validation plan, total-cost questions, and the transparent cases × minutes saved ÷ 60 × loaded-rate model. Do not submit patient, referral, order, payer, authorization, document, appointment, portal, medication, result, claim, or credential values through the marketing form.