AI consulting

My own business / We’re Back Co.

Make campaign reporting
easier to work with.

Campaign names often live inside long tracking links. I built a tool with Codex that turns that information into simple labels on Shopify orders, so it is easier to find and group when reviewing sales.

What the tool does

  1. Read the campaign information

    Find the campaign and ad names recorded in the customer’s landing link. Skip missing or unfinished information.

  2. Add useful labels

    Turn those names into consistent order labels. A preview shows the proposed changes, and the tool keeps existing labels.

  3. Know what the labels can tell you

    The label tells you what the link recorded. It does not prove that an ad caused the sale, and it cannot fill in missing information.

Try an example

Choose a sample campaign link

What the link records

/products/sample?utm_source=instagram&utm_medium=paid_social&utm_campaign=sample-launch&utm_content=sample-creative
Sourceinstagram
Channelpaid_social
Campaignsample-launch
Creativesample-creative

What the tool proposes

  • ad-sample-creative
  • campaign-sample-launch
  • source-instagram-paid_social

Illustrative input and proposed output; this page does not connect to Shopify or change an order. A link label records what the link contained. It does not prove an ad caused a sale.

Built and used in my own business

This is a working WBC tool, with a preview mode and checks to avoid adding the same labels again. I reviewed its run records and checked a resulting label in Shopify. The examples above use sample information, not customer records.

Download the source excerpt
Operating evidence and limits

Reviewed against the WBC tagging source, attribution specification and September 1–13, 2026 run logs. The logs record 49 live-mode runs, 32 with tagging actions, and no logged errors in that interval. One recorded tag was independently read back from Shopify on September 13. This establishes recorded operation and a sample readback, not perfect attribution, time savings or sales lift. The source excerpt excludes customer records, credentials and store connection code.

Don’t Panic long-sleeve creative developed with Kive

Creative is part of the practice, too

A specific brief comes first.

For the Don’t Panic visual: a back view, scarlet backdrop, warm grain and a ’90s flash-photo feel. Kive supports image development; Higgsfield is part of my video workflow. Product and artwork review remain part of the job.

See the WBC placement

Concept to physical product

The image and the actual print.

Team Trough had to work as a campaign idea and as a shirt someone would receive. Drag to compare an AI-assisted campaign image with a photograph of the printed gear.

These are different photographs of the same design, not a pixel-matched before-and-after.

Stacks of finished red Team Trough tees showing the printed front and back artworkAI-assisted Team Trough scene showing the red tee beside a stadium trough
AI-assisted campaign imageFinished printed gear

Bring one recurring task.

We’ll define the inputs, build a prototype, review its output and document how your team can use it.

Discuss your workflow