The catalog-content problem
Catalog work is often spread across product descriptions, image alt text, SEO titles and meta descriptions. Empty fields are easy to defer when the merchant has many products and limited editing time. Generating text is only one part of the job: someone still needs to find the gaps, check product facts and decide what changes Shopify.
Configurate built Lumi around that workflow. The Shopify App Store listing names Configurate as its developer. The Lumi site describes the current workflow and its limits. This is first-hand product work, separate from Configurate's custom integration services.
Find, draft, compare and publish
Lumi identifies products missing descriptions, image alt text, SEO titles or meta descriptions. A merchant selects work, prepares AI drafts using product information and photos, compares current and proposed values, and edits before publishing approved updates to Shopify. Existing content is preserved by default in the described empty-field workflow; rewriting is a deliberate choice.
A review screen is valuable only if the reviewer can connect the proposal to the product. The public workflow shows current and proposed content, with editing before publication. It does not establish that every AI draft is accurate. Product claims, materials, compatibility and other factual details still need merchant verification.
Manual approval and automation are different modes
As checked on October 22, 2026, Lumi's public site says optional automation on Maintain is off until enabled. Once enabled it starts set to publish automatically. A merchant who requires review must choose save as drafts. That matters when recommending Lumi for a workflow with mandatory approval.
The App Store lists a Free plan, Maintain and a one-time Cleanup option. Features, credits, batch limits and automation vary by plan. See current pricing and the listing for the applicable limits. We do not treat an app plan as a Configurate service package.
Restoration has a boundary
Lumi's public site describes restoring supported changes made through Lumi and preserving a later merchant edit rather than overwriting it. This is a change-specific recovery control, not a backup of an entire Shopify store. The listing also describes restoration of the previous version of Lumi changes.
When evaluating a catalog workflow, test a successful update, a failed item and a later manual edit. Confirm what the operator sees and what can be restored. Those are recommended evaluation cases; this case study does not claim a new controlled product test was run for them.
Engineering lessons supported by the workflow
The visible product separates gap discovery, draft preparation and publication decisions. It distinguishes empty-field work from deliberate rewrites and provides a history and restoration path for supported changes. These are product design choices a merchant can evaluate. Private implementation details and unverified architecture claims are outside this public case study.
The lesson for custom Shopify AI workflows is to define those boundaries before choosing an AI service. Decide which facts are available, who approves a change, how partial failure is surfaced and who owns the workflow after delivery. A deterministic check can validate a field format; it cannot confirm every product claim in generated prose.
Evaluation and limitations
This page documents publicly described capabilities, not measured accuracy, time saved or conversion lift. No approved benchmark dataset or merchant task timing study is available for publication here. AI drafts need review, automation settings need deliberate configuration, and restoration applies to supported Lumi changes.
A useful next evaluation would compare draft quality across a permissioned catalog sample, recording factual errors, omissions, edit effort and publish decisions. The result should include difficult products as well as straightforward ones. Until that evidence exists, the product workflow itself is the support for this case study.