How to Extract and Verify Benefits from a PDF
By RossTo extract benefits from a PDF safely, keep a copy of the original, separate each plan and effective period, copy the rates and benefit terms with page references, and verify the result before calculating costs. A readable table is only the first deliverable. The reviewed plan data must still agree with the source when it reaches the client presentation.
1. Identify the documents and plan boundaries
Inventory the rate sheet, Summary of Benefits and Coverage (SBC), renewal letter and any enrollment file. Record their dates and revisions. An SBC describes benefits; it may not contain the group's negotiated premium. Keep the matching rate sheet rather than treating a missing rate as zero.
HealthCare.gov explains the SBC's comparison role. It is a summary, so unresolved limitations may require the plan document or carrier confirmation. For a scanned PDF, check that text recognition has preserved decimal points, column headings and footnotes.
2. Extract a small, explicit field set
- Full plan name, carrier, current or renewal role, network and effective dates.
- Monthly premium for each coverage tier, with currency and billing basis.
- Individual and family deductibles and out-of-pocket maxima, identifying in-network versus out-of-network values.
- Coinsurance and visit terms with qualifiers such as “after deductible.”
- Document filename and source page for each plan or field, plus unresolved questions.
Use unknown for absent values. Do not infer family deductibles from individual figures, infer a monthly rate from a total without confirming enrollment, or silently turn “20% after deductible” into “20%.” If two documents disagree, record both versions and resolve which applies.
3. Try a repeatable prompt on a fictional file
Download the two-page fictional Ridgeline packet. Page 1 is Atlas PPO 1500 Current; page 2 is Atlas PPO 1500 Renewal. It contains invented organizations, rates and enrollment, with no client information.
Use the attached PDF only. Extract each plan's full name, current or renewal role, effective dates, four monthly tier rates, deductibles, out-of-pocket limits and exact coinsurance wording. Include source pages. Use null for absent values and list unresolved items. Do not use web search or connected apps, and do not create or modify any cases.
ChatGPT supports file attachments, as described in its official usage guide. File support does not establish extraction accuracy. Availability and settings differ by account and product. Use an approved data-handling arrangement before uploading a real carrier or enrollment file; start practice runs with fictional data.
4. Verify the rates and arithmetic independently
| Tier | Current, page 1 | Renewal, page 2 | Enrollment |
|---|---|---|---|
| Employee only | 500 | 550 | 5 |
| Employee + spouse | 1,000 | 1,100 | 2 |
| Employee + child(ren) | 900 | 990 | 1 |
| Family | 1,400 | 1,540 | 2 |
Current monthly premium is 500×5 + 1,000×2 + 900×1 + 1,400×2 = $8,200. Renewal is $9,020: a 10% increase on the same census. Annualized totals are $98,400 and $108,240. The source states a 75% employer contribution to each tier. With an additional assumption of 26 deductions, renewal employee-only cost is 550×25%×12÷26 = $63.46 per paycheck, rounded at the end.
Both pages state a $1,500 individual / $3,000 family deductible, $4,500 / $9,000 out-of-pocket maximum and member coinsurance of 20% after deductible. Verify the words as well as the numbers.
5. Review, save, reopen and inspect the exported file
Keep extraction separate from approval. Confirm plan roles and dates, resolve unmatched fields, apply contribution assumptions, save, then reopen the case and inspect the actual exported proposal. A correct rate extraction can still lose a benefit qualifier or export a different cover than the preview.
Our documented AI PDF comparison separates extraction checks from dated Planlined workflow observations. The October 20 import record matched eight rates but lost “after deductible” and required manual role assignment. That single-file result is a reason to retain review steps, not a general accuracy claim.
What Planlined adds after extraction
Document parsing starts the structured review. Plan comparison and contribution modeling turn approved inputs into employer and employee views, and AI connections connect supported assistant workflows to Planlined. The final goal is a client-ready proposal that agrees with the reviewed data.
Use the quote analysis guide for a complete review checklist and the invoice reconciliation guide when the document is a bill requiring enrollment and rate checks.