The most consequential customer browsing an online shop may soon have no eyes, no mouse and no patience for a poetic description of a toaster. It may be an AI agent, dispatched to compare specifications and return with a shortlist. For retailers, this creates an awkward new assignment. The product page still has to persuade a person. Now it also has to be cleanly parsed, compared and trusted by a machine.
That tension is the market Ocula has chosen. Founded in 2021 by Thomas McKenna and Gregory Fletcher, the London-headquartered company makes software for the unglamorous middle of ecommerce: the catalogue. Its agents collect missing attributes, turn data into copy, police that copy against brand rules and prepare it for websites, Google Shopping, Amazon, TikTok feeds and AI answers. Ocula is not trying to be the shop window. It wants to organize the stockroom behind every digital window at once.
This matters because a retailer with 100,000 products has a different problem from someone staring at a blank document. A general chatbot can draft a description in seconds. It cannot, by itself, discover which 12,000 products are missing a material, preserve five brand voices, respect marketplace character limits, cite where a new attribute came from and stop a failed output before export. At catalogue scale, writing is the visible final inch of a much longer production line.
The long tail is where products disappear
Retail content teams naturally lavish attention on launches and bestsellers. The long tail receives whatever arrived in a supplier spreadsheet, sometimes with sparse attributes, stale phrasing or the same description as a competitor. Seasonal refreshes pile up. Agencies add capacity but not necessarily a durable system. Product-information management software can store the facts without supplying the context that helps a product answer a shopper's oddly specific question.
Ocula's Enrichment Agent is built for that gap. The company says it can extract facts from images, supplier PDFs, retailer pages, reviews and the open web, attaching a confidence score and source citation to each value. It also scores a SKU on four less-than-romantic but useful qualities: completeness, consistency, conformity and currency. The resulting data is not copy yet. It is the material from which trustworthy copy can be made.
A writer with a critic standing behind it
The Copywriter Agent turns enriched data into titles, descriptions, metadata and schema, with variations for each destination. The QA Agent then checks the result rule by rule. Length, terminology, blocked words, required elements, brand voice and formatting can each produce a pass, flag or error. A vague warning that the copy feels “bad” would be useless to a merchandiser reviewing thousands of records; Ocula says each flag identifies the rule and the reason it failed.
“Your team shifts from writing copy to approving it.”Ocula's argument for catalogue-scale automation
The distinction is important. Ocula is selling controlled throughput, not a synthetic novelist. Its Orchestrator Agent receives a plain-language objective and coordinates the steps, while people retain the approval role. That puts the product in a crowded but unusually broad competitive set. It vies with internal teams and agencies for labor, with PIM vendors for a place in the data stack, with specialist copy platforms for generation and with ChatGPT or Copilot for the executive who wonders why any of this needs another contract.
Ocula's answer is that generic prompting breaks at scale. Brand requirements need to be encoded once, outputs need to fit multiple schemas, keyword choices need evidence, and a second system must check what the first system wrote. Its platform sits alongside a PIM rather than simply replacing it. A PIM remains the system of record; Ocula tries to make the record useful for discovery.
The proof is in pages, not prompts
The company publishes a collection of retailer results, which should be read as customer-reported outcomes rather than universal guarantees. Ocula says clients have optimized more than five million product pages. The Sole Supplier reported 84 percent more organic traffic on pages using Ocula copy. Blain's Farm & Fleet reported a 95 percent improvement in workflow efficiency and a 15 percent conversion increase. Across the case studies, Ocula advertises an average 15 percent traffic lift.
The customer list explains the appeal. AO and Boots carry broad inventories. Hornby Hobbies sells products with details that matter deeply to enthusiasts. Marks & Spencer must preserve a familiar voice while accelerating search work. Sportswear brand Castore has deployed Ocula across 35 Shopify storefronts worldwide. Travel Chapter gives the system thousands of properties and several distinct brands to describe. These are not businesses short of copy. They are businesses where variation, freshness and governance multiply faster than headcount.
A company that changed with the question
Ocula did not begin with this exact pitch. Incorporated in January 2021 as Intelabs Analytics Limited, it adopted the Ocula name that December. McKenna and Fletcher came from strategy, analytics and data science, and their first public products reflected that background. Ocula Boost surfaced prioritized actions for ecommerce performance; a price-optimization module addressed another familiar retail problem. AO adopted Boost in 2023.
The company then sharpened around generative product content. By 2024, investors were willing to fund the transition: a Series A led by Praetura Ventures reached £4 million, with existing backers Castelnau Group and Lloyds Banking Group participating. The stated plan was to expand sales and technology teams and develop the product range. Ocula's current LinkedIn profile places it in the 11-to-50 employee band and shows staff across London, Belfast and New York.
The progression from dashboard to copywriter to coordinated agents follows the industry's changing question. Retailers first asked how analytics could tell them what to fix. Then they asked AI to write the fix. Now they are asking how a system can discover the missing data, produce the change, test it and repeat when the market moves. Ocula's “enrich, generate, activate, refresh” loop is a product roadmap disguised as four verbs.
Where the bet could break
No retailer should confuse volume with truth. Extracting an attribute from a photograph or review can introduce ambiguity. A source can be current and still wrong. Automated QA can catch a forbidden phrase while missing a subtly misleading implication. Search platforms can change how they rank or cite products, and AI-answer visibility remains harder to measure than an ordinary click. The platform's value therefore rests as much on provenance, review controls and integrations as on the fluency of its generated prose.
There is also a strategic squeeze. PIM companies can add stronger generation and enrichment. Large retailers can build internal agents around their proprietary data. General models keep getting cheaper and more capable. Agencies can pair software with human judgment. Ocula must show that its specialized workflow compounds: that every approved output, failed rule and performance result makes the next catalogue decision better, not merely faster.
Yet the underlying problem is durable. Product discovery is fragmenting across conventional search, onsite search, retailer marketplaces, social commerce and conversational interfaces. Each channel wants the same product expressed in a different structure. The human shopper still wants confidence and a reason to care. The machine intermediary wants explicit facts it can retrieve and compare. A thin supplier description satisfies neither.
The commercial model is familiar enterprise software: a demo-led SaaS sale to brands and retailers, with pricing kept private. The more revealing economics belong to the buyer. Ocula estimates that automating 10,000 SKUs can avoid more than £250,000 in manual costs, though each catalogue will carry its own labor rates and review burden. The decision is therefore not “AI or no AI.” It is whether a retailer pays people to repeat the same preparation across every channel, pays an agency to absorb the queue, builds a bespoke system, or rents a specialized production line and keeps its staff on strategy, merchandising and approval.
The product page is no longer only a page. It is becoming an interface between a retailer's stock and every machine that might recommend it.
That is where Ocula fits: above the raw product record, below the customer-facing experience, and across the channels between them. Its customers can use it to clear copy backlogs, launch assortments faster, localize content, expose overlooked use cases and prepare product data for queries nobody thought to place in a title. The work is quiet. If it succeeds, the result is simply that the right product appears with the right facts when somebody - or something acting for somebody - asks.