← Back to work
Case study

AI content transformation

Turning a weeks-long photography and editing cycle into a same-day process, without losing consistency.

Weeks → 1 day turnaround for a full seasonal image set

The situation

The business ran its marketing calendar around seasonal peaks, several times a year. Each one needed a fresh set of product imagery: hundreds of assets shot, retouched and formatted for a launch date that couldn't move.

That pipeline took days to weeks per campaign. The lead time didn't flex when a launch moved up or a range changed late, and it didn't scale with the number of seasonal ranges launching each year.

What was missing

Speed was the obvious problem. The less obvious one was consistency.

What made a set of images feel like one campaign lived in people's judgement. Each photographer and retoucher had their own sense of how lighting, backgrounds and colour should look, so imagery made by different people on different days didn't always sit together.

That was manageable while people did the work. It wouldn't survive AI: a model produces whatever its brief describes, and a brief can't describe a standard nobody has written down.

Before/after comparison of a product image produced through the traditional photography workflow versus the AI-assisted workflow

My role

I researched and tested the tools, worked out where AI could replace or speed up each step without breaking brand standards, and built the workflow: prompts, templates, the written standards and the review stages. It was built for the whole team to use every day, not as an experiment on my own computer.

What got written down

Before the workflow scaled, I wrote the standards down: the judgement calls a reviewer would otherwise make from memory, stated plainly enough that the output stayed on-brand whoever ran it. Four rules did most of the work, and nothing shipped without a human check.

The standards the workflow ran on

  • Background and lighting must match the existing photography style for that product category: no visible seams between AI-generated and camera-shot imagery in the same set.
  • Any generated image showing a product's key selling feature (texture, size, included parts) goes to human review before use.
  • Brand colour accuracy takes priority over aesthetic preference: reject and regenerate rather than adjust in post.
  • Batch runs are checked in sets of 10 before scaling to a full range, not reviewed one-by-one at full volume.

AI-assisted workflow

From weeks of shooting and retouching to a single day

AI introduced end-to-end, with brand standards and a human check built in, not bolted on

  1. 1

    Generate & refine

    AI generates and refines product imagery from a reference set

  2. 2

    Standardise

    Backgrounds and framing standardised across the whole range in one pass

  3. 3

    Batch variants

    Every size and format each channel needs, produced in one run

  4. 4

    Human review

    Checked in batches of 10 before scaling to a full range: nothing ships unseen

  5. 5

    Launch on schedule

    Ready for the date the business wants, not the date imagery happens to be ready

Weeks to 1 day

Turnaround for a full seasonal image set, with more consistent output than the manual process it replaced

What changed

A complete seasonal image set went from days or weeks to one day, and the results were more consistent than the manual process had been. Campaigns could launch on the dates the business wanted instead of waiting for imagery, and ranges that never had the time or budget for a photo shoot could have imagery of their own.

The rule that made this reliable, checking sets of ten before scaling to a full range, is the one Batch blindness shows the absence of: what happens to an AI image batch when nobody checks it that way.

The speed came from the AI. The consistency came from writing the standards down.

---
title: AI content transformation
slug: ai-content-transformation
updated: 2026-10-03
summary: Turning a weeks-long photography and editing cycle into a same-day process,
  without losing consistency.
stat: Weeks → 1 day
stat_label: turnaround for a full seasonal image set
---

If a problem like this sounds familiar, the audit is where I'd start: one workflow, one 90-minute session, and a written brief on what's missing and what to fix first.

Book the audit

Prefer to talk first? Get in touch — no obligation.