🚀 What is Summary(k)?
Summary(k) reads a complete 401(k) plan document and produces a structured, provision-by-provision summary — organized for TPA onboarding specialists.
Key difference from Adopt(k): Adopt(k) parses checkbox adoption agreements. Summary(k) reads full legal plan documents written in prose — typically 40-150 pages.
📋 Steps
1
Select AI Model
GPT-5.4 recommended — handles large documents with 1M context window
2
Upload Plan Document PDF
Full plan documents, restated plans, and amendments are all supported
3
Click Summarize Document
Two-pass AI analysis takes 3-5 minutes for a typical document
4
Review & Export
Expand provision cards, check evidence, and export in your preferred format
5
Analyze Provisions (Optional)
Click the AI Analyze button for cross-provision intelligence — flags interactions, unusual items, and onboarding action items in ~15 seconds
🔍 How Two-Pass Works
🗺️
Pass 1: Reconnaissance
Maps the document structure — identifies articles, pages, and key provisions
📝
Pass 2: Extraction
Summarizes each provision using the map as a guide — knows exactly where to look
Why two passes? Plan documents vary wildly by provider (Boyarsky, Relius, FIS, ASC). A single-pass approach can't reliably navigate 60+ pages of legal prose. The recon pass orients the AI before it extracts — like a human flipping through the document first.
⏱️ Timing
Pass 1 typically takes 60-90 seconds. Pass 2 takes 2-4 minutes depending on document length. The elapsed timer shows progress in real-time.
🛡️ Graceful Fallback
If Pass 1 fails, the system automatically falls back to single-pass extraction. You still get results — just without the quality boost of targeted summarization.
📊 Reading the Results
Results are organized into 12 provision categories. Each card shows structured label:value pairs for quick scanning.
Value Formatting
YES
Green Badge
Provision is enabled with details following
NO
Gray Badge
Provision is not permitted or not elected
DISC
Yellow Badge
Discretionary — employer determines annually
📋
Source Evidence
Click to see the exact section and quote from the document
💡 Pro Tip: Multi-part values (like entry dates that differ by contribution source) render as bulleted sub-lists. Vesting schedules render in monospace for easy reading.
⚡ Tips for Best Results
💡
Use GPT-5.4 for large documents
Plan documents over 40 pages may exceed Claude's 200K token limit. GPT-5.4 handles up to 1M tokens.
💡
Check amendments at the bottom
Many plan documents have SECURE Act, CARES Act, or disaster relief addenda that modify core provisions.
💡
Use Copy or CSV for onboarding
The Copy export gives you a clean text summary. CSV gives you structured data for spreadsheet workflows.
💡
Compare with Adopt(k)
Run the adoption agreement through Adopt(k) and the plan document through Summary(k) to cross-verify provisions.