Why consolidation matters
If your EMR has hundreds or thousands of document types (often built up over years), the AI has to learn each one as if it’s unique. That slows learning and holds back automation. Fewer, clearer categories let the model learn quickly and perform better.
Think of this like bins in a mailroom.
If you have 1,000 small bins with slightly different labels, sorting is slow. If you have 200 clear bins, sorting is fast and reliable.
What “good” looks like
- Clear, broad categories (Referral, Labs, Imaging, Consult Notes, Pharmacy, Admin Forms).
- No duplicates or near-duplicates (avoid tiny naming changes that mean the same thing).
- Only keep a category separate if the workflow is truly different.
Example: Referral vs. Consult Note → different downstream steps, so keep them separate.
Before/after examples
| Before (many, overlapping) | After (clean, learnable) |
| Referral – Cardiology, Referral – Surgery, Referral – Dr. Smith | Referral |
| Labs – Bloodwork, Labs – Urinalysis, Labs – A1C | Labs |
| Imaging – CT, Imaging – MRI, Imaging – X-Ray | Imaging |
| Consult – Dr. Patel, Consult – Cardiology | Consult Notes |
When Document Lists Aren’t Consolidated
This real-world scenario shows why starting with a clean list is critical before training Inbox Agent.
The scenario
A clinic went live with hundreds of overlapping document types:
- “Referral – Cardiology,” “Referral – Surgery,” “Referral – Dr. Smith”
- “Consult – Endocrinology,” “Consult Note”
- “Lab Report – A1C,” “Bloodwork,” “Results,” etc.
Later, they added a second, simplified list (“Referral,” “Consult,” “Labs”) — but instead of replacing the old ones, they kept both active.
That effectively doubled their total number of document types.
What happened next
- Each type had only a few samples for the AI to learn from.
- The model treated similar faxes as unrelated.
- Confidence scores dropped sharply — even simple referrals showed 5–20% confidence instead of 70–90%.
- As a result, few documents met the automation threshold.
The lesson
Always consolidate before training.
Keep one active list — clear, consistent, and free of duplicates.
The cleaner your categories, the faster Inbox Agent learns and the higher your automation accuracy climbs.
How to approach consolidation (practical checklist)
- List out your current EMR document types.
- Group by intent/workflow (e.g., all diagnostic results → Labs).
- Merge duplicates and near-duplicates.
- Keep only what changes routing or follow-up as distinct (e.g., Referral vs. Consult Notes).
- Review edge cases together with us (we’ll advise when to keep vs. merge).
Real-world note: Clinics that reduced from thousands of types to a few hundred now see strong automation (often >70% of eligible faxes auto-processed) once training/usage ramp up.
Important rules once training starts
- Don’t remove, archive, or reorder existing document types after training.
- You can add a new type later, but it won’t be recognized until the model is retrained (or it learns gradually from usage).
- Renaming a type is okay only if the meaning doesn’t change.
- If you plan a bigger restructure, loop us in first so we can protect accuracy.
Use Dynamic Tags for Naming Precision
Once you’ve consolidated your document type list, avoid creating new document types just to capture small naming differences like physician name, test name, modality, or body part.
Instead, use Dynamic Tags to build more specific document names automatically while keeping your document type list clean and manageable.
What Dynamic Tags Do
Dynamic Tags pull specific details from each document and use them to build a more descriptive file name.
This lets you keep one clean document type while still showing helpful detail in the final document name.
You can also combine dynamic tags with static text.
That means part of the name can stay fixed, and part of it can be filled in automatically from the document.
Example:
Instead of creating:
MRI – Abdomen – Dr. Smith
MRI – Chest – Dr. Lee
Use one document type:
MRI Report
…and apply a naming convention such as:
[Modality] – [Body Part] – [Physician Name]
This keeps your list small and structured while maintaining document-level precision.
Supported dynamic tags
You can use tags such as:
Appointment Date
Appointment Status
Assigned Provider Name
Body Part
Diagnosis
Document Description
Document Name
Document Type
Encounter Date
File Received Date Time
Health Card Number
Insurance Plan Name
Insurance Request Type
Lab Test List
Lab Test Name
Legal Request Type
Medications Ordered
Modality
Order Expiration Date
Patient DOB
Patient First Name
Patient Last Name
Physician Name
Priority
Reason for Referral
Referral Request Type
Referral Status
Rx Request Type
Sender Fax Number
Sender Name
Specialty
Surgery Date
To Provider Name
Static Text + Dynamic Tags
You are not limited to tags only.
You can type regular text directly into the naming field and combine it with dynamic tags.
Example combinations
MRI Report - [Body Part] - [Physician Name][Patient Last Name], [Patient First Name], [Encounter Date], [Document Type][Document Type] - [Sender Name] - [Encounter Date]
This is useful when you want a naming format that stays consistent across your clinic.
Why This Matters
Using Dynamic Tags helps in three important ways:
Model performance
Fewer, cleaner document types improve classification accuracy.Maintainability
A shorter document type list is easier to manage, update, and troubleshoot.Searchability
Documents stay specific and easy to find without bloating your setup.
How to Configure Dynamic Tags
A short visual walkthrough is helpful here, especially for users who may not realize they need to click into the field before adding tags.
Log in to Phelix and turn the Control Panel on.
Go to Inbox Agent → Inbox Settings → Document Types.
Click into the naming field to place your cursor where you want the name to appear.
Add any static text you want included.
Click the + button to insert a Dynamic Tag.
Select the tag you want from the list.
You can also click the gear icon next to the document type you want to edit.
Repeat this process to build your preferred naming format.
Click Save.
Once saved, new incoming documents for that document type will use the naming pattern you created.
Common Mistake to Avoid
Do not create a separate document type for every variation in physician name, body part, modality, or test name.
That makes your document list harder to manage and reduces AI accuracy over time.
Use one consolidated document type, then use Dynamic Tags to add the detail.
Cleanup Recommendations
If you currently have a large or overlapping document list:
- Merge variations that differ only by provider, body part, or test name.
- Keep only clinically distinct types.
- Add your final, consolidated list to Phelix.
- Configure Dynamic Tags afterward for naming precision.
Please complete this cleanup before training or retraining your document classifier.
💬 Need help?
Contact your Phelix Account Manager — we can review your list and suggest best-fit categories before training begins.