Oak Grove

The brand

Authentic Japanese stencils, built into a home goods brand

Oak Grove transforms authentic 19th-century Japanese stencils into woven jacquard blankets, throw pillows, and other home decor goods. Each pattern begins as katagami: a hand-carved stencil used for dyeing traditional textiles like kimono, furoshiki, and nōren.

I built the brand identity, product line, US and Japanese Shopify storefronts, and a sourcing and curation pipeline for stencil artwork. Fulfillment partners handled production and shipping; I directed the product, commercial, creative, and operational work around them.

Business decisions

Building products that could work commercially

I compared production methods, suppliers, materials, formats, costs, margins, shipping, and pricing to decide which products to offer, when to launch them, and where to make them available. Those decisions shaped the catalog, assortment, storefronts, and release plans.

Product strategy, merchandising, localization, marketing, analytics, and order operations were parts of the same business system. The work was not finished when a design looked good; the product also had to be viable to produce, price, ship, explain, and sell.

2022–2026

The division of work changed as AI capabilities expanded

The answer changed as ChatGPT, Claude, and Gemini gained new capabilities.

I worked with the models to research, explore alternatives, critique ideas, surface tradeoffs, develop plans, create content, and rethink decisions as new information appeared. What I could hand off expanded from conversation into files, images, connected data, and selected structured workflows.

Operations and review

Airtable held the catalog and the review states

As the catalog expanded, I moved product records, source patterns, colorways, storefront content, and creative assets into Airtable so the catalog had one place to live. Records were linked where the connection mattered, including which pattern and colorway became which product and what was still waiting on review.

Airtable became the control point once workflows arrived. Records triggered work, generated outputs landed back with the prompt that produced them, and nothing moved toward a storefront until I approved it there. Designing where those review points sat was most of the actual work.

Human–AI boundary

Some work stayed manual by design

Selecting historical source material and cleaning the artwork remained manual. Source selection determined each product's provenance, while cleanup choices affected the visual integrity of the final pattern. Those decisions carried brand risk that automated screening could not reliably absorb.

The goal was not to automate as much as possible. I assigned work according to capability, risk, and quality, and retained final authority over consequential product and publishing decisions.

2026

Later formalization: selected recurring work

By 2026, model and tool capabilities had expanded from chat to file generation, image editing, direct Airtable access through MCP, and API-connected workflows. I formalized selected recurring tasks in n8n, connecting Airtable, Shopify, DeepL, and Gemini, Anthropic, and OpenAI models. These workflows were a late extension of the way I had already run Oak Grove with AI, not the origin of the business.

I designed and evaluated workflows for familiar tasks and new possibilities, including staging products into rooms from inspiration photos and applying palettes from reference images to stencil designs. The important design work was not just connecting APIs. It was defining constraints, making intermediate outputs reviewable, and keeping human approval at consequential points.

The two examples below represent different kinds of applied AI work: multimodal creative direction and a connected decision-and-review pipeline.

Workflow 01

Inspiration Stager

Before

Writing staging prompts meant starting from scratch each time, describing lighting, furniture, geometry, and mood in text, with no guarantee the output matched what I'd imagined.

After

The workflow accepted two images: a room photo and a product photo. Gemini used the room to interpret lighting, furniture, and layout, then placed the product into a new scene while preserving its pattern and colors.

I designed the workflow around two sequential Gemini calls rather than one. The first analyzed the inspiration photo for lighting, furniture, and geometry but was explicitly told to ignore the product. The second placed the product into that scene. Splitting the calls reduced the risk that Gemini would redesign the product to match the room.

Inspiration Stager Workflow

Queued in Airtable

→

Gemini: read scene (ignore product)

→

Gemini: place product into scene

→

Airtable record

Dropbox

▸ Human review gate: image QA before use in listings or ads

Workflow 02

Colorway Pipeline: Three Linked Workflows

I designed a linked sequence to capture color inspiration from product photos, apply it to a stencil, generate up to 6 colorway mockups, and route the results through two review apps before anything could move toward production.

Before

Chose colors by instinct, like using a color picker as a Ouija board, guided by gut feeling with no record of what I'd tried and no way to compare options.

After

1. A photo of a promising color scheme entered the workflow from desktop or mobile.
2. Gemini extracted its colors into hex codes in Airtable.
3. Those colors drove pillow mockups in up to 6 colorways.
4. Promising options could be sent to reviewers before anything moved toward production. (not illustrated above)

Step 1 of 3 · Palette Lab

Capture color inspiration from anywhere

I built this workflow to save color inspiration from photos as reusable palette records. Gemini analyzed images submitted from desktop or mobile, extracted hex codes, and wrote them to Airtable for use by the Colorway Generator.

Palette capture workflow

Desktop bookmarklet

Mobile photo upload

→

Gemini: analyze image

→

Color palette saved to Airtable

Grid view

Gallery view

Step 2 of 3 · Colorway Generator

Apply a palette to a stencil and generate up to 6 variants

The Colorway Generator combined a validated Palette Lab record with a black-and-white stencil and generated up to 6 mockups for a first pass on which combinations were worth pursuing. Exact color still happened in Photoshop: Gemini could get close enough to evaluate, not close enough to ship.

Colorway generation workflow

B&W stencil

Palette Lab record

→

Gemini: generate armchair mockup

→

Gemini: generate up to 6 color mockups

→

Colorway URLs saved to Airtable

Step 3 of 3 · Design Review

Curate a batch, send to reviewers, collect yes / maybe / no votes

Two small web apps I built with Claude Code. In the First Cut web app, I reviewed finished colorways and selected which were worth showing. The second app, Design Review, was designed for family reviewers to vote Yes / Maybe / No and leave notes. Votes were saved to Airtable so I could compare feedback before committing anything to production.

Feedback loop workflow

Airtable: completed colorways

→

First Cut: I select keepers

→

Airtable: Design Reviews table

→

Reviewer app: Yes / Maybe / No

→

Votes saved to Airtable

Curator view: select colorways to send for review

Reviewer view: Yes / Maybe / No per design

▸ Human review gate: curate selection before publishing to listings or ads

Supporting workflows

Additional workflow designs across content, publishing, and analysis.

What this demonstrates

What almost 4 years of solo operation taught me about working with AI

Over four years, Oak Grove became a production environment for studying where AI can support business workflows, where it introduces risk, and where human review is necessary to protect quality, accuracy, and brand trust.

These judgments were grounded in real failures. AI guidance on paid advertising could be confidently wrong, while near-real product imagery risked creating a gap between what customers saw and what they would receive. Those risks informed the approval states and review gates I designed into the later workflows.