Copywriter Agent • Ocula

Ocula Technologies is an AI platform built for e-commerce brands and retailers. The core proposition: take product data and turn it into optimised, channel-ready content at scale. For large retailers managing thousands of product detail pages, that's a time-consuming, valuable problem to solve.

The engine, GC3.0, Ocula's generative copywriting model, was already built. It could produce brand-aligned product descriptions at a quality and volume no human team could match. The problem was how customers accessed it. Every job ran through Retool and each request had to come in by email. A member of the internal team received the brief, interpreted it, ran the generation, formatted the output and sent it back.

Two weeks after I joined, the Head of Design and I flew to London for a two-day sprint with the full leadership team: CEO, CTO, Product, Engineering and Data Science. The brief coming out of that sprint was clear: design a self-service platform that put the generation capability directly in the hands of customers.

Two days in London with the team — where the project took shape

The absence of a customer-facing interface created two compounding problems. For customers, there was no autonomy. Generating copy required submitting a request, waiting for a response, reviewing an output they hadn't been directly involved in shaping, and going back-and-forth until it was right. For Ocula, every customer job was a support burden.

Moving to a product-market fit model meant breaking that dependency on Ocula staff. Customers needed to be able to request, receive and iterate on generated copy themselves.

When I joined, the company had recently pivoted and there was not much that aligned with the new design direction to inherit and no product layer to lean on in the early stages. There were few wireframes, minimal interface designs and no component library. The starting point was Retool, spreadsheets and email threads.

The team was small and the timeline was short: less than two months from sprint outcomes to shipped MVP. That meant making decisions quickly, with limited research runway, and trusting the sprint findings rather than running extended discovery of my own.

The Copywriter MVP gave customers a self-service interface to generate product copy at scale. The key design decisions across the three layers were:

The core generation flow was designed to feel lightweight and iterative, not transactional. Customers could scope a job, trigger generation and review results without needing to understand what was happening underneath. The interface made GC3.0's capability accessible without exposing its complexity to the user.

Sample and Feedback Loop replaced a previous, manual method using a Typeform survey. Customers began onboarding by reacting to real generated samples rather than filling in abstract preferences. Thumbs up, thumbs down, specific edits. The content style built from their responses and users could see it update in real time. The design goal was to move customers from uncertainty to confidence in their output before they committed to a full generation run.

Proactive Insights changed the relationship between the platform and the customer from reactive to ongoing. Rather than waiting for a customer to decide they needed to optimise copy, the feature identified the opportunity and surfaced it directly. Low-performing PDPs, outdated descriptions, content that hadn't been touched since launch: all flagged automatically with a clear prompt to act. The design kept the recommendations tightly scoped so they felt useful, not overwhelming.

The shipped Copywriter MVP — customers could now request, generate and review copy without internal support

Content style built through reaction, not a form — customers rated samples and watched their preferences take shape in real time

The most direct outcome was structural: Ocula's team stopped being the bottleneck for every customer job. The manual process that had been running through email and Retool was replaced by a platform customers could use themselves.

Sample and Feedback Loop reduced the back-and-forth that had been a persistent drain on Customer Success. Customers arrived at generation with a content style they'd had a hand in building rather than one the internal team had interpreted on their behalf and the number of revision cycles dropped significantly.

Finally, Proactive Insights gave the platform a reason to exist between campaigns. Engagement shifted from event-driven, triggered by a product launch, to ongoing. That was the commercial behaviour the business needed to move beyond a high-touch consultancy model.

Low-performing and outdated PDPs surfaced automatically, giving customers a reason to keep coming back