Zainab Hussain is a seasoned e-commerce strategist who has spent the last year redefining how customer insights drive product development at Online Plastics Group (OPG). Managing an expansive digital footprint across twelve webshops in nine European markets, she has navigated the complex intersection of customer engagement and operational efficiency. Her work focuses on closing the “Decision Gap”—the distance between what an organization knows about its customers and what it actually prioritizes in its development sprints. By integrating conversational AI as a structural layer of organizational learning, she has transformed the typical agile backlog from a stagnant list of features into a dynamic learning system that generates measurable business value and prevents costly strategic errors.
Traditional customer research often fails to reach product owners at the moment they prioritize a sprint, turning many backlogs into what you call a “graveyard for customer value.” In your experience at Online Plastics Group, what is the fundamental breakdown that prevents these insights from being used effectively?
The breakdown isn’t usually a lack of effort; it is an architectural failure in how we handle information. We often see a “translation problem” where the rich, qualitative world of customer experience research—like NPS scores or complex journey maps—simply does not slot neatly into the rigid RICE scoring systems or Jira tickets that Product Owners use to run their day. In many organizations, customer insights live in separate silos: the marketing team has their A/B test results, the customer service team has a database of complaints, and the SEO team has their own experimental data. When the moment of decision arrives during sprint planning, these disparate pieces of knowledge are tucked away in different heads or hidden in separate spreadsheets, making them inaccessible to the person holding the prioritization pen. Personal expertise doesn’t scale, and without a centralized system to capture and surface these learnings, even the best research is eventually ignored in favor of the ticket marked “URGENT.”
To bridge this gap, you implemented a system at OPG that uses AI as an interface rather than a standalone tool. How did you structure the relationship between tools like Notion, Airtable, and Claude to ensure that knowledge becomes a default reference point for the team?
We built our experimentation model around three specific roles to ensure that product development and customer insights remained connected rather than parallel. Notion serves as our primary knowledge bank, housing every experiment type from sequential pricing tests to AI initiatives, because if only marketing tests are documented, the Decision Gap remains wide open for every other department. Airtable handles the planning layer—tracking status and projected business impact—but the real “unlock” was using Claude AI as a conversational interface between these databases and the humans making decisions. We developed custom skills for Claude that classify experiments as they occur and analyze results through a one-year revenue decay model, making retrieval effortless. Instead of a Product Owner needing to launch a weeks-long research project to see what has been tested before, they can now ask the system a question and receive a concise summary with linked data in seconds.
There is a lot of skepticism regarding AI in product strategy, with many fearing it might replace human judgment or provide “hallucinated” advice. How do you maintain the integrity of your decision-making while using AI as a primary interface?
Our approach is built on the principle that AI should be an interface, never an oracle; we purposefully designed the system so that every output generated by Claude must be validated by a human in the source systems before it can influence a roadmap. Trust in AI isn’t built by delegating judgment, but through a rigorous and consistent validation process that ensures the team maintains the “muscle memory” required to recognize when an automated insight might be off-base. We also simplified the process by making documentation a byproduct of the work rather than a separate “tax” that team members hate to pay. By allowing documentation to happen through a conversation with the AI as the experiment runs, we significantly reduced friction, which naturally led to higher-quality records. This human-in-the-loop architecture ensures that while the AI handles the heavy lifting of data retrieval and classification, the final ownership of the decision remains firmly with the experts.
When looking at the operational shifts at OPG between 2024 and 2025, the numbers suggest a massive increase in output and value. Could you walk us through the specific impact this system had on your experimentation volume and the bottom line?
The shift was quite dramatic, as our experimentation output jumped from just 22 experiments in 2024 to 56 in 2025, supported by seven continuously running tests across our various European markets. This increased volume wasn’t just noise; it generated 31 roadmap-impacting insights that contributed approximately €900K in projected business value. Perhaps more importantly, we started tracking “preventive loss,” which accounted for an estimated €100K in revenue preserved by identifying and stopping experiments that would have caused damage if they had been fully shipped. It is easy for a CFO to look at the wins, but making the “stopped losers” visible proves that the system is functioning as a safety net for the business. This dual focus on upside and loss prevention has helped us treat customer value as a validated, mathematical input into every major decision we make.
One of the most interesting statistics from your work at OPG is that 21 percent of experiment ideas now come from non-marketing teams. How did this decentralized approach change the culture of product development across your twelve webshops?
Moving toward a decentralized model of optimization meant that experimentation became a shared capability rather than a specialist function tucked away in a marketing corner. We saw significant contributions from customer service, pricing, and even the development teams themselves, which is what closing the Decision Gap actually looks like in practice. For instance, during a major replatforming initiative, this centralized learning system allowed us to avoid the trap of simply rebuilding suboptimal features just because “that’s how it was before.” Because we could pull up prior experiment data in seconds, we had the evidence to justify leaving out certain legacy functions that we already knew didn’t provide value. This cultural shift means the entire organization is now empowered to use validated customer input as a compass, ensuring that the engine of our agile process is always pointed in the right direction.
For a CX leader looking to implement this Agile Customer Experience framework, what are the most critical principles they should focus on to avoid common pitfalls?
The first priority must be building a retrieval system before you focus on increasing experiment volume, because more data without a way to find it just creates more noise. You have to ask yourself: can a Product Owner find a specific learning in five seconds? If the answer is no, your infrastructure is broken. Secondly, you must extend the system beyond A/B testing from day one to include SEO, pricing, and content trials; otherwise, you are just building another silo rather than a comprehensive solution. It is also vital to measure the value of the experiments you don’t ship—counting only the winners hides half the value of the program and leads to an undervalued team. Finally, always start with the decision, not the tooling, by identifying which parts of your current roadmap are based on assumptions rather than validated input, and make that gap your real backlog.
What is your forecast for the future of Agile Product Management and AI-driven customer insights?
I believe we are moving toward a reality where the “backlog as a graveyard” will be a relic of the past, replaced by living learning systems that function as an organizational nervous system. In the next few years, the competitive advantage for e-commerce groups won’t be how fast they can ship code, but how quickly they can retrieve and apply institutional knowledge to avoid repeating old mistakes. We will see a shift where AI handles the entire lifecycle of an experiment’s documentation and cross-referencing, allowing product teams to focus purely on creative problem-solving and strategic judgment. Organizations that fail to build these retrieval layers will find themselves running faster and faster in the wrong direction, while those that treat customer value as a structured, retrievable input will be the ones that actually scale. Ultimately, assumptions will remain part of the process, but they will no longer be the only thing in the room when the most important decisions are being made.
