James Washington
writing about archive utils free game ideas replays
Automation

Report-only shadow-mode spam judge for helpdesk tickets

Watches support tickets and notes which ones look like sales spam, but only writes down what it would have done so a person can check it for a week before it is allowed to act.

built Oct 2026 updated Oct 2026 Letting AI make calls, safely

how it works

Shadow-mode trial: asks two separate property questions (is it outreach? is it a customer asking for help?) and requires high/low respectively before a would-close verdict; reads via the D1 REST query API, writes only a local ledger.

Automation: Something that runs on a schedule by itself.

The code for this one lives in a private repo. This page explains the idea so you can build your own.

helpdeskllm-judgesafetyshadow-mode

related

  • Read-only multi-tenant support KPI report Command-line tool Produces a weekly scorecard of support tickets across several brands' helpdesks.
  • Jev shadow scorer for recruiter mail Automation Tests an AI scoring of recruiter emails quietly, logging results without letting it decide anything yet.
  • Code-owns-workflow, models-supply-judgment outreach pipeline Automation A program that researches leads and writes draft sales emails for a person to approve, never sending anything itself.
  • Two-layer inbox classifier (rules first, LLM judge second) Automation Automatically labels every incoming email as belonging to one of several businesses or a personal area, so every downstream tool sees the same answer.
  • APK-safe bug-report relay Worker API Lets a TV app send bug reports to the support desk without hiding any passwords inside the app.
← everything I've built
© 2026 James Washington
github twitter rss