←Back/Operations/Claude/Airtable
IntermediateOperations

How to Create a Custom Product Classification System with AI and Airtable

Design a stable classification system for a collection, test it against representative items and ambiguous cases, then use Airtable AI fields to propose codes while people resolve uncertainty and protect the taxonomy from drift.

How to Create a Custom Product Classification System with AI and Airtable

Source episode · 32:30

How I AI: Building a Real-World Business with Andrew Mason & Nabil Hyatt

with Claire Vo

Andrew and Nabeel use a Claude designed library style classification for board games, store the codes in Airtable, and apply built in AI fields to categorize a large inventory and generate mutually exclusive collections that would have been impractical to maintain by hand.

Read the full episodeWatch on YouTubeListen on SpotifyListen on Apple

Before you start

What you need

  • A defined purpose and audience for browsing the collection
  • An inventory with stable item IDs and trusted descriptive metadata
  • A hierarchical taxonomy with code definitions, examples, and version
  • A labeled test set containing common, rare, hybrid, and ambiguous items
  • An Airtable base with permissions, views, change history, and AI fields

What you’ll make

A searchable catalog whose codes are explainable, reproducible, and versioned, with low confidence or multi category items routed to review rather than silently forced into the wrong shelf.

Tools used

Step by step

The workflow

Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.

4 steps

Step01

Design the Classification Framework with an AI

Define why the collection needs classification and how people will browse it. Draft a hierarchy with stable codes, names, definitions, examples, exclusions, and rules for items that span several categories.

Example prompt
Design a classification system for [collection] used by [audience and task]. Propose stable hierarchical codes with definition, parent, examples, exclusions, and ambiguous cases. Distinguish primary shelf placement from optional facets. Do not classify the inventory yet.
Step02

Set Up Your Airtable Database

Create Airtable tables for items, taxonomy versions, codes, assignments, evidence, confidence, reviewer, and status. Keep source metadata separate from AI suggestions and protect accepted fields from bulk overwrite.

Step03

Automate Categorization with an AI Model

Give the AI field the locked taxonomy and one record at a time. Require an allowed code, rationale from the item metadata, confidence, and review status, with an unresolved result when evidence is insufficient.

Example prompt
Classify this item using only taxonomy version [version] below. Return allowed primary code or UNRESOLVED, optional facet codes, rationale tied to supplied metadata, confidence, and missing information. Never invent a code. Taxonomy: [codes]. Item: [record].
Step04

Curate Collections with AI-Powered Fields

Use accepted codes to build shelf, discovery, and event collections. Audit category coverage and confusion, review low confidence items, and change the taxonomy through a versioned proposal rather than editing labels in place.

What good looks like

  • Each code has a unique definition, parent, examples, exclusions, and stable identifier.
  • Independent reviewers classify the test set consistently enough for the intended use.
  • AI suggestions include rationale and confidence, and ambiguous records remain visibly unresolved.
  • Taxonomy changes are versioned and migrated without silently changing the meaning of existing codes.

Build your next product with ChatPRD

Turn an idea into a PRD, user stories, and a plan.

Try ChatPRD free

After the steps

Runbook notes

How to recover when the loop fails and where human judgment helps.

Recover

If it goes sideways

Several categories plausibly fit the same item and results vary by prompt
Clarify inclusion and exclusion rules, allow facets where appropriate, and route unresolved primary placement to review.
The model invents new labels or changes code meaning across runs
Provide the versioned taxonomy as the only allowed code list and reject unknown codes.
Weak or inconsistent item descriptions produce unreliable classifications
Normalize trusted metadata, preserve source fields, and mark missing attributes before classification.
A bulk automation overwrites curated codes or makes errors hard to reverse
Write suggestions to separate fields, review in batches, keep history, and promote accepted values explicitly.

Keep building

Related workflows

Create an Automated Workout Tracker with a Simple Claude Voice Note
beginnerPersonalClaude

Create an Automated Workout Tracker with a Simple Claude Voice Note

Log your workouts effortlessly without ever opening a spreadsheet. This workflow uses the Claude mobile app to parse a simple voice note and automatically update your workout tracking spreadsheet in the background.

Open
How to Rebuild Your Gmail Inbox Inside Claude to Manage Email
advancedOperationsClaude

How to Rebuild Your Gmail Inbox Inside Claude to Manage Email

Create a custom email management interface within Claude to read, triage, and draft replies without opening the Gmail tab. This workflow helps you reclaim focus and compound your learnings by training your AI on every interaction.

Open
Automate Client Proposals and Onboarding with a Claude 'Pipeline Operator'
intermediateSalesClaude

Automate Client Proposals and Onboarding with a Claude 'Pipeline Operator'

Build a fully automated system in Claude to generate personalized, interactive HTML proposals and client onboarding materials, saving hours on administrative work and creating a high-touch experience.

Open

Start shipping
better products.

Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.

Start building freeRequest a demo
Free to start
No credit card
SOC 2 certified
Enterprise ready