---
title: "AI auto-tagging with the Avodah Intelligence taxonomy"
url: "https://kaperkins.com/lab/ai-auto-tagging-avodah-intelligence"
author: "Karl Perkins (K.A. Perkins)"
site: "Avodah Dynamics"
category: "AI"
published: "2026-06-21"
updated: "2026-06-28"
description: "Every capture gets classified across 5 dimensions — Assignment, Season, Path, Audience, Emotion — so the system understands why a piece exists, not just what it is."
---

# AI auto-tagging with the Avodah Intelligence taxonomy

## Summary

Every capture gets classified across 5 dimensions — Assignment, Season, Path, Audience, Emotion — so the system understands why a piece exists, not just what it is.

## Problem

Categories were too flat. A "blog post" tag doesn't tell you which season it belongs to, who it's for, or what assignment it serves.

## Solution

Created a taxonomy_terms table with 53 seed terms across 5 dimensions. Extended the AI pipeline so Gemini classifies each capture into all five dimensions during the first pass and writes the tags onto the source row.

## Lessons Learned

The taxonomy is the first thing that made the AI feel like it understood the mission, not just the surface text. When the model knows the season, the lede writes itself.

## Next Steps

Let the taxonomy drive feed routing — Dispatch in Launch season, Lab in Build season, Letters in Reflection season.

---
Source: https://kaperkins.com/lab/ai-auto-tagging-avodah-intelligence · Karl Perkins, Avodah Dynamics