[WSJF-7] Boundary-test: identify emails that don't need AI #7

New issue

Open

opened 2026-06-27 10:45:28 +02:00 by bullitt · 0 comments

bullitt commented

2026-06-27 10:45:28 +02:00

Owner

Context

Article: 'Evaluator worth the cost when task exceeds model's solo capabilities.'

Problem

Many emails from DHL/Hermes have identical formats. Once a parser exists, AI is never called again — but we don't know which email types are 'solved' vs. which still need AI.

Proposal

Add /api/parsers/stats showing: emails processed by parser vs. by AI (last 30 days)
Identify high-volume parsers (>10 uses) as 'solved' patterns
Identify emails that consistently fall through to AI as patterns needing manual parser creation
Goal: reduce AI costs by knowing where deterministic parsing already works

WSJF: 7

## Context Article: 'Evaluator worth the cost when task exceeds model's solo capabilities.' ## Problem Many emails from DHL/Hermes have identical formats. Once a parser exists, AI is never called again — but we don't know which email types are 'solved' vs. which still need AI. ## Proposal - Add `/api/parsers/stats` showing: emails processed by parser vs. by AI (last 30 days) - Identify high-volume parsers (>10 uses) as 'solved' patterns - Identify emails that consistently fall through to AI as patterns needing manual parser creation - Goal: reduce AI costs by knowing where deterministic parsing already works ## WSJF: 7