Illustrative scenario · Healthcare, specialty outpatient services · Oracle Fusion Service and Amazon Bedrock
Getting a patient ready before the appointment
A specialty outpatient network connects appointment requests, preparation instructions and administrative support, while clinical decisions stay with care teams.
Illustrative scenario
- The promise
- The referral is acknowledged, missing information is requested once, and a clinical question reaches a clinician before the patient arrives.
- Where it breaks in this scenario
- Referrals arrive through several channels and missing information is tracked in spreadsheets and shared inboxes, so staff ask for the same things twice.
- What changes
- Agents summarize, check referrals against an approved checklist and route questions. Clinical decisions stay with care teams; identity and scheduling decisions stay with authorized staff.
- The business case
- Modeled below, with every assumption beside its figure and the downside if recovered capacity adds nothing. See the financial case.
A few days before her MRI, a patient receives preparation instructions. She has an implanted device and is unsure whether the instructions apply to her. Her question needs to reach the right clinician while there is still time to complete the examination.
This scenario follows that question, and the referral behind it.
Her primary care physician has referred her to an outpatient imaging center. She needs an MRI before a specialist appointment in two weeks.
The referral arrives, but the imaging center needs additional information before it can complete scheduling. The patient calls to ask whether the order was received. A scheduler searches the referral queue, checks another system for authorization status, and promises to call back.
Business days from referral
Before
4 business days to administratively ready. Information asked for more than once.
The implanted-device question waits in the same queue as scheduling questions.
After
2.5 business days to administratively ready. Missing information requested once.
The device question goes to the clinical team with her wording and the appointment context, and reaches a clinician before she arrives.
Before: the same questions, asked twice
The specialty network uses an EHR and scheduling system for clinical and appointment records. Referral documents arrive through several channels. Administrative teams track missing information in spreadsheets and shared inboxes.
A marketing platform supports general community education and provider outreach. Once someone requests an appointment, staff transfer the request into the healthcare workflow.
Patient access representatives repeatedly check whether referrals are complete, authorizations are pending, or a patient has already been contacted. Different staff members sometimes ask for the same information.
Preparation instructions are selected manually from a document library. Scheduling questions arrive in the same queue as questions requiring clinical review. When an appointment is cancelled late because information is missing, the network loses capacity and the patient loses time.
Scheduling questions arrive in the same queue as questions that need clinical review.
A coordinated path to the appointment
One request, one record of what is outstanding
QxOrbit’s proposed design centers on Oracle Fusion Service, Fusion Digital Customer Service, Fusion Data Quality, and the network’s EHR, scheduling, and referral systems.
General outreach continues through the marketing platform. Patient-specific administrative and clinical information moves through approved intake and healthcare systems.
A service request connects the appointment inquiry to referral status, outstanding tasks, and contact history. Fusion Data Quality can flag possible duplicate administrative records. Identity reconciliation follows the network’s established patient-identification process before records are linked.
The referral, read against the checklist
An agent built in Oracle AI Agent Studio for Fusion Applications prepares a work summary for the patient access representative. It identifies completed items and retrieves outstanding-task status from connected systems.
A custom Amazon Bedrock agent reviews referral documents against an approved administrative checklist. The design assumes the network already operates document ingestion and approved content repositories in AWS. Findings include source references, allowing staff to request the specific information needed without rereading the entire packet.
The device question, routed to a clinician
Through Digital Customer Service, the patient sees administrative next steps, receives approved preparation material, and submits questions. An implanted-device question is routed to the clinical team with the patient’s wording and appointment context.
The EHR and scheduling system remain authoritative for clinical documentation, orders, and appointment status.
- General outreach, then the patient enters an approved intake channelThe provider submits an order, recorded in the healthcare systemsFusion Service tracks the administrative request
- Person decides: Staff resolve ambiguous identity matches
- Agent: A Bedrock agent reviews the referral for completeness
Decision: Required information complete?
- NoPerson decides: Staff request the missing information↺ back to the Bedrock referral review
- YesPerson decides: The team checks administrative readiness
- Person decides: Scheduling and clinical teams confirm the appointment
- System or team: The portal delivers approved preparation material
Decision: Clinical question?
- YesPerson decides: The clinical team reviews and gives direction
- NoPerson decides: Staff resolve routine administrative questions
- System or team: The appointment proceeds under the clinical workflow
- System or team: The service team handles administrative follow-up
What the agents do, and what people decide
Agents prepare, on their own
- Summarize the request, contact history and outstanding tasks for the patient access representative.
- Check referral documents against the approved administrative checklist, with source references.
- Explain the evidence behind a possible duplicate record.
- Retrieve approved administrative answers and route every other question.
- Highlight unresolved requirements before the appointment.
A named person decides
- Patient access staff resolve exceptions and confirm the next action.
- Authorized staff resolve ambiguous identity matches before records are linked.
- Clinical teams assess the medical relevance and adequacy of a referral.
- Clinicians answer questions about medication, devices, symptoms and anything else that needs medical judgment.
- Scheduling and clinical teams decide whether an appointment proceeds or changes.
An agent never
- gives clinical direction or answers a clinical question,
- links patient records,
- decides whether an appointment proceeds, or
- sends preparation material that has not been approved.
Where the agents and people fit
| Workflow | Agent contribution and platform | Human responsibility and why |
|---|---|---|
| Administrative coordination | A Fusion agent summarizes the request, contact history, and outstanding tasks. | Patient access staff resolve exceptions and confirm the next action. |
| Referral review | A Bedrock agent finds information required by the approved intake checklist and highlights missing or inconsistent fields. | Staff verify findings. Clinical teams assess medical relevance and adequacy. |
| Patient identity | Data Quality presents potential duplicates, and an agent explains matching evidence. | Authorized staff resolve ambiguity because incorrect linking can attach information to the wrong person. |
| Preparation support | A Fusion workflow delivers approved material for the confirmed appointment. An agent retrieves approved administrative answers and routes questions. | Clinicians address medication, devices, symptoms, and other clinical questions requiring medical judgment. |
| Appointment exceptions | A Fusion agent highlights unresolved requirements and summarizes outreach. | Scheduling and clinical teams determine whether the appointment can proceed or needs to change. |
Bedrock is selected where the existing AWS document pipeline and content controls make reuse practical. Fusion manages the service workflow and employee tasks. Deployment requires validation of the specific services, contractual arrangements, access controls, and patient-information flows. Clinical questions follow the organization’s established care process.
What success would look like
The baselines and targets below are planning assumptions to validate in a pilot.
| Business measure | Assumed baseline → modeled target | Why it matters |
|---|---|---|
| Referral received to administratively ready | 4 → 2.5 business days | Measures resolution of information needed for scheduling. |
| Repeat requests for information already supplied | 25% → 10% | Measures avoidable effort for patients, providers, and staff. |
| Administrative effort per eligible referral | 24 → 16 minutes | Tests whether review and summaries save time after verification. |
| Eligible administrative inquiry handling time | 12 → 8 minutes | Measures speed of resolving routine questions. |
| Late cancellations caused by missing administrative information | 4% → 3% | Shows whether coordination preserves appointment capacity. |
| Patient-reported ease of scheduling | 70% → 80% favorable | Measures whether the experience feels easier to patients. |
The network also reviews inappropriate routing, missed clinical escalations, incorrect preparation material, and identity-linking errors. Appointment and completion metrics come from the EHR and scheduling systems, combined with Fusion service records in approved operational reporting. Clinical and appointment data require explicit integration and reporting design.
The financial case Modeled estimate
At 12,000 eligible referrals and 15,000 eligible administrative inquiries a year, the proposed reductions would release 2,600 staff hours. All amounts are modeled, in US dollars.
| Annual financial impact | Calculation and assumption | Value |
|---|---|---|
| Referral-processing cost avoided | 12,000 × 8 minutes ÷ 60 × $40 an hour × 50% converted to avoided spend | $32,000 |
| Inquiry-handling cost avoided | 15,000 × 4 minutes ÷ 60 × $40 an hour × 50% converted to avoided spend | $20,000 |
| Contribution from recovered appointment capacity | 12,000 appointments × 1 percentage point fewer administrative cancellations × 70% yielding additional completed visits × $300 contribution per visit | $25,200 |
| Total annual benefit | Avoided spend plus additional visit contribution | $77,200 |
| Incremental annual running cost | Platform usage, integration operations, evaluation, and support | ($30,000) |
| Annual operating improvement | Before tax and one-time implementation cost | $47,200 |
With a $65,000 implementation investment, modeled payback is about 16.5 months after reaching full operating volume. Three years at full volume would produce $76,600 net benefit, a 49.4% cumulative return on $155,000 of total implementation and running costs.
Additional visits count only when they create recovered capacity. Rescheduling without increasing total completed volume creates no incremental contribution in this calculation. Referral effort and inquiry handling must be measured as distinct work to avoid counting the same minutes twice.
If recovered capacity contributes nothing, annual operating improvement falls to $22,000, and payback extends to about 35.5 months. Avoided labor costs require evidence of reduced spend or avoided hiring.
The estimate assumes core applications and integration foundations already exist. New licensing, substantial infrastructure changes, deployment time, and adoption ramp require recalculation.
- Volumes assumed
- 12,000 eligible referrals and 15,000 eligible administrative inquiries a year; 12,000 appointments
- Time saved assumed
- 8 minutes per referral and 4 minutes per inquiry at $40 an hour; 50% converted to avoided spend
- Capacity assumed
- 1 percentage point fewer administrative cancellations; 70% become additional completed visits at $300 contribution each
- Costs assumed
- $65,000 to implement; $30,000 a year to run
- Downside shown
- No contribution from recovered capacity: $22,000 a year and payback in about 35.5 months instead of 16.5
For the patient, the referral is acknowledged, missing information is requested once, and the question about her implanted device reaches a clinician before she arrives. For the network, fewer administrative surprises protect staff time and appointment capacity.