Automatically file, route, and chart incoming documents once the system is confident they’re ready.
1. What Auto-Processing Is
Auto-Processing is what turns Inbox Agent from a sorting tool into a hands-free filing system.
Once your faxed or uploaded documents are confidently recognized and matched, Phelix can:
- File them directly into the right patient chart (Auto-Save)
- Create new patient charts when none exist (Auto-Chart)
- Generate referral records and route them automatically (Auto-Referral)
These automations free staff from repetitive filing while ensuring documents end up in the right place.
2. Why It Matters
For most clinics, manual fax management is the single biggest administrative bottleneck.
Auto-Processing reduces that by automating every routine filing step that can safely be trusted to AI:
| Manual Step | Automated Equivalent |
| Sorting faxes by type | Classification confidence |
| Searching for the right patient | Patient-match logic |
| Drag-and-dropping into EMR charts | Auto-Save trigger |
| Creating new charts for new patients | Auto-Chart trigger |
| Logging new referrals | Auto-Referral trigger |
Over time, these features can handle 70–90% of routine faxes — leaving staff to focus only on edge cases or low-confidence items.
3. How Auto-Processing Works with Document Training
Document training teaches Phelix how to recognize your document types.
Auto-Processing uses those recognitions to act automatically.
- If a document type has been trained and frequently used, confidence is high → auto-processing is likely to run.
- If a document type is new or inconsistently labeled, confidence is low → the fax stays in the worklist until confirmed manually.
Think of it as a chain:
Training → Confidence → Matching → Automation
When the confidence and match scores meet the parameters you set, the automation is triggered. Otherwise, it holds back for review — ensuring control and safety.
4. How the Three Layers Work Together
The three automation layers can work independently or in sequence:
- Auto-Save files documents when a patient match already exists.
- Auto-Chart creates a chart when no match is found, enabling the document to be filed or routed.
- Auto-Referral builds the referral record and directs it to the right provider.
When used together, this creates a fully automated intake loop — from fax to chart to referral — while leaving uncertain cases in the worklist for human review.
5. Safe Rollout Checklist
To introduce automation safely:
- Start with high-confidence document types (e.g., labs, imaging).
- Set conservative thresholds (e.g., 80–90%) and expand gradually.
- Review worklist results daily for the first week.
- Adjust match rules and thresholds based on observed holdbacks.
- Add more document types once you’re comfortable with the results.
9. Troubleshooting Holdbacks
When auto-processing doesn’t trigger, check these first:
| Area | Common Holdback | Resolution |
| Document | Confidence below threshold | Review training data; consider consolidation or retraining |
| Patient | No match found or multiple matches | Relax matching rule or enable Auto-Chart |
| Provider | Missing or restricted | Add default provider or expand inclusion list |
| Referral | Procedure or referring provider not confident | Lower threshold slightly or add examples |
| System | Duplicate already filed | No action needed |
6. Maintaining Accuracy
- Avoid deleting or re-ordering document types after training — each one has a unique internal identifier tied to model performance.
- Add new document types only when they’re consistently used and can accumulate examples for training.
- Periodic retraining (every few months) helps the model adapt to new templates and layouts.
- Review automation reports to see confidence trends and identify which types could safely move from manual to automatic.