How Retailers Scale Product Content Without Adding Headcount
Grosvenor Flooring arrived at merchi.ai with a 1,000-product backlog. Products were sitting in the catalogue without descriptions, without accurate technical attributes, and without lifestyle imagery. Each one was invisible to search, unfilterable on the site, and converting at a fraction of its potential. Clearing that backlog manually would have taken months of copywriter time. With merchi.ai, it was cleared without adding headcount. The result: 976% online revenue growth. That implementation is now part of our National AI Awards 2026 Finalist entry for “AI SME Business of the Year”.
That story is not unique to Grosvenor. A product content backlog is one of the most common operational problems in retail, and one of the least talked about in public. This post explains how they happen, what it actually takes to clear one at scale, and how AI changes the arithmetic entirely.
Why product content backlogs happen
A content backlog is the gap between the number of products in your catalogue and the number with complete, optimised content live on your website. Most retailers have one. It grows because product data creation is a human-hours problem.
The triggers are familiar: a new supplier range arrives and needs onboarding. The website platform migrates and the content format changes. A buying team adds 200 products in a season. A content team working manually might complete 20 in the same period. The backlog grows, and next season it grows again.
The cost of that gap is real and measurable. Incomplete product listings rank lower in search (thin or missing content means fewer indexed keywords). They also convert worse: shoppers cannot make purchase decisions without accurate dimensions, material information, and imagery showing the product in context. Returns rise. Customer service contacts rise. Revenue that should be there is not.
Manual remedies have always existed: offshore copywriting agencies, temporary data entry staff. The problems are cost, time, and inconsistency. The agency takes weeks to onboard. The temporary staff work from briefs that are never quite complete enough. The output varies between writers. You end up spending months of management time reviewing and correcting content that is still not to specification.
What scaling product content actually involves
There is a significant difference between “using AI to write one product description” and having a genuine product content pipeline. Most generic AI writing tools handle the former. They are useful for a single product or a quick draft. They are not built for a catalogue of thousands.
Scaling product content means all of the following happening consistently, across every product:
- Batch processing: uploading hundreds or thousands of products in a single operation, not one at a time
- Schema-driven output: every product produces the same attributes in the same format, defined by your taxonomy rather than the AI’s guess at what matters
- Taxonomy classification: each product assigned to the correct category, consistently, without human judgement calls on ambiguous cases
- Multi-modal input: working from product images, supplier data sheets, existing partial descriptions, or a combination of all three
- Multi-language output: generating content in 40+ languages without a separate translation workflow
- Attribution: the AI Provenance Protocol tags content to indicate AI origin, meeting emerging standards for responsible content disclosure
Generic AI tools handle none of these as a pipeline. They require a human to prompt individually, review output, paste into a CMS, and move on to the next product. At 1,000 products, that is still a manual process. The AI has made each individual step slightly faster, but it has not removed the bottleneck.
What genuine scale requires is a structured platform: images and data in, structured product content out, at volume, without human intervention at each step.
How Grosvenor Flooring cleared a 1,000-product backlog
When Grosvenor Flooring came to merchi.ai, the backlog was well over 1,000 products. The challenge was not just volume. Flooring products have specific technical attributes that matter to the buyer: dimensions, coverage, material type, installation method, wear rating. Those attributes had to be accurate, not approximated. And the output had to be consistent: the same attribute structure and format across every plank, tile, and sheet vinyl in the range.
merchi.ai processed the full backlog in a coordinated operation. Product images were uploaded in bulk. The schema was configured once to match Grosvenor’s taxonomy. Every product was generated to that specification, with descriptions, technical attributes, and lifestyle imagery (including the AI Room Visualiser, which lets shoppers see a floor product rendered in their own room). The team at Grosvenor did not need to handle each product individually.
The result was 976% online revenue growth. Products that had previously been invisible to search began ranking for specific, transactional queries. Products that had previously converted poorly began converting. There was no additional headcount. The bottleneck shifted from human writing time to processing queue.
This implementation forms part of the National AI Awards 2026 Finalist entry for “AI SME Business of the Year”.
How merchi.ai handles product content at scale
The platform is built specifically for retail merchandising at volume. Key capabilities:
- ZIP upload: upload a folder of raw supplier product images directly; the platform extracts, processes, and generates content without manual file handling for each image
- Spreadsheet import: if you have existing structured data (supplier data sheets, existing PIM exports), import it alongside the images and the platform merges the inputs into the output schema
- Configurable schema: define exactly which attributes each product type needs. A flooring product has different fields from a lighting product. The schema is set once; every batch run inherits it automatically
- API and webhook integration: for retailers with existing PIM, ERP, or ecommerce platforms, merchi.ai connects directly via API. New products trigger content generation automatically as they enter your system, without a manual upload step
- Writing assets: the platform’s output is shaped by writing assets you control: brand voice guidelines, category-specific tone, product naming conventions, and any constraints specific to your range. These are configured once and applied consistently across every batch run, so the content reads like your team wrote it
- 40+ languages: content is generated in any supported language without a separate translation workflow, useful for retailers selling across multiple markets
- AI Provenance Protocol: every piece of generated content is tagged with attribution metadata, making the AI origin transparent to search engines and compliant with emerging disclosure standards
If you are thinking about generating individual product descriptions, the scale capabilities sit on top of that same foundation. And if you want to understand the real cost of manual product data entry before you can make the case internally for a change, the numbers are instructive.
Manual vs AI product content: the comparison
| Manual approach | AI approach (merchi.ai) | |
|---|---|---|
| Speed | 5-20 products per person per day | Hundreds per batch run |
| Time to clear 1,000-product backlog | 10-40 weeks | Minutes to hours |
| Consistency | Varies by writer, brief, and day | Schema-enforced: identical structure for every product. Context-grounded: writes in your tone of voice following your writing rules |
| New range onboarding | Brief the team, quality-review output, correct errors | Upload images / data, run batch, review exceptions |
| Ongoing resource | Permanent or agency headcount | Platform subscription |
See how this works for your catalogue
If your product catalogue is growing faster than your content team can keep up, we can show you how this works in practice. Book a 20-minute walkthrough and we will walk through the Grosvenor Flooring setup and apply it to your specific catalogue and schema.
Frequently asked questions
How long does it take to generate product content for 1,000 SKUs with AI?
With merchi.ai, a batch of 1,000 products can be processed in minutes to hours rather than weeks. The exact time depends on the content specification: number of attributes, description length, whether lifestyle imagery is included. The constraint shifts from human writing time to processing queue rather than calendar weeks.
Can AI generate product content consistently across a large catalogue?
Consistency is one of the core advantages of AI-generated content over manual writing. merchi.ai uses a configurable schema: you define exactly which attributes each product type needs, and every product in a batch is generated to that same specification. That means no missing attributes and no format variations between copywriters. The platform is also context-grounded: writing assets define your brand voice, tone of voice, and category-specific writing rules, so the output reads like your team wrote it rather than generic AI copy.
What is a product content backlog and why does it matter for ecommerce?
A product content backlog is the gap between the number of products in your catalogue and the number with complete, optimised content live on your site. Incomplete product listings rank lower in search (missing keywords, thin content) and convert worse because customers cannot make purchase decisions without accurate information.
Does merchi.ai work with existing product data, or does it need to start from scratch?
merchi.ai works from what you already have. You can import via spreadsheet with your existing product data fields, upload product images directly, or use a ZIP of raw supplier images. The platform maps your existing data to the output schema. It does not require a clean-slate rebuild of your catalogue.
How does AI product content keep pace with seasonal range changes?
Because the schema and writing knowledge base are configured once per product type, adding a new season’s range is a matter of uploading the new products and triggering a batch run. You do not need to brief a copywriter on the brand voice each season. That knowledge is already built into the platform configuration.
Is AI-generated product content good for SEO?
Yes, when it is structured correctly. merchi.ai generates content with consistent keyword placement, complete attribute coverage, and the AI Provenance Protocol tag so search engines can understand the content’s origin. The Grosvenor Flooring implementation produced 976% online revenue growth; search visibility was a central part of that outcome.