💡 AI-structured email archive format for knowledge-base consumption

[github-actions[bot]]

Summary

Enhance gmail-to-drive-by-labels' Google Doc output format to include structured metadata blocks — participant lists, date ranges, subject lines, and optional AI-generated topic tags — that make email archives directly consumable by AI knowledge tools like NotebookLM, Gemini, and RAG pipelines. This transforms raw email dumps into searchable, AI-ready knowledge documents without changing the core archiving workflow. The structured format also improves human scanability of archived threads.

Market Signal

NotebookLM integration was highlighted at Google Cloud Next 2026 for Workspace Studio workflows, signaling Google's investment in structured documents as AI input sources. The PRD already identifies email archives as "source documents for tools like NotebookLM" (Innovation section). Google's September 2026 feature drop added cross-app content creation capabilities and Google Vids can now transform Docs into video summaries — both features work better with structured input. The broader industry trend: documents are increasingly consumed by AI systems, not just humans. Structured formats with explicit metadata dramatically improve AI reasoning quality and citation accuracy. Inbox Zero (open-source competitor) already structures email data for AI consumption with categories, draft replies, and CRM integration.

User Signal

gmail-to-drive-by-labels currently produces Google Docs with raw email text separated by thread/message markers (============================== and ------------------------------[THREAD:id]). These docs are human-readable but lack the structured metadata that AI tools need to reason about them: who participated, what topics were discussed, what the date range covers. Users who archive emails for later reference — the core use case — benefit from both better human scanability AND AI-query readiness. The existing gmail-ai-classifier script in the codebase already demonstrates the team's capability and interest in AI-powered email processing.

Technical Opportunity

The existing archiving pipeline already processes thread metadata (participants via getFrom()/getTo(), dates via getDate(), subjects via getSubject()) but discards most of it during Doc writing, keeping only the message body text. The architecture supports this enhancement directly: insertParagraph(0, ...) can prepend structured header blocks, and the dual-file pattern (code.gs + src/index.js) enables fully testable formatting logic. Optional Gemini integration (for topic tags and one-line summaries) can be gated behind a config.gs flag (e.g., enableAiEnrichment: true), keeping the base script Gemini-free for users who don't want AI processing. The getCleanBody() utility in src/gas-utils.js already handles content normalization.

Assessment

Dimension	Score	Rationale
Feasibility	high	Metadata is already available in the pipeline; this is a formatting/output change with optional AI enrichment gated behind a config flag
Impact	med	Improves both human scanability and AI readiness of archives; positions the project for the NotebookLM/RAG trend Google is investing in
Urgency	med	NotebookLM adoption is growing but not universal; the base format improvement (metadata headers) has standalone value regardless

Adversarial Review

Strongest objection: How many users actually pipe their email archives into NotebookLM or RAG systems? This seems like a niche use case optimizing for a future that may not arrive.

Rebuttal: The structured format benefits ALL users, not just AI pipeline users. Adding a participant list, date range, and topic header to each archived thread makes the Doc better for human scanning too — instead of scrolling through raw email text, users see "Thread: Q3 Budget Review | Participants: Alice, Bob, Carol | Oct 3-7, 2026" at the top of each section. The AI-optimization is a zero-marginal-cost bonus on top of a genuine UX improvement. The PRD explicitly calls out NotebookLM as an emerging use case. And with Google's cross-app content creation (Sep 2026), structured Docs are increasingly the interchange format between Workspace apps and AI tools.

Suggested Next Step

Add structured metadata blocks to the gmail-to-drive-by-labels Doc output: thread header section (participants, date range, subject line) prepended before each thread's content. Implement as a formatting enhancement in src/index.js with full test coverage. Preserve backward compatibility with existing separator-based dedup markers ([THREAD:id]). Phase 2: add optional Gemini-generated topic tags and one-line summary, gated behind a config.gs flag.