How we combined AI and human-centred design to shape the future of online marketplaces
AI by Design at OLX
Introduction
OLX is one of the world's leading online classifieds platforms, connecting millions of users to buy and sell secondhand goods. But scaling trust, relevance and simplicity across global peer-to-peer marketplaces is no small feat.
To tackle this, OLX² - OLX's global innovation team - partnered with Koos to create a design-led experimentation hub focused on the intersection of AI and human-centred design.
The Challenge
How might we use AI to enhance safety, trust, and user experience across the platform, without losing the human touch?
My role
As Research & Design Lead (Koos), I shaped and led the discovery and experimentation process:
Conducted user research and concept validation across multiple domains
Developed a scalable experimentation framework in collaboration with OLX²
Facilitated co-creation with product, data, research, and design stakeholders
Helped embed AI-by-Design innovation capabilities across the organisation
AI-powered video sales
Together, we explored and tested three strategic AI-driven solutions that addressed key marketplace challenges. Each concept was brought to life through full-bleed visuals, mock-ups and video, making the experience as tangible as the results.
1.
AI-powered video sales
To build trust and reduce uncertainty, we explored a video-first listing format. AI helped sellers create listings that conveyed product authenticity and condition better.
The result: Buyers could better assess product condition, especially in categories such as electronics and fashion. This increased their confidence and reduced friction.
2.
Hyper-personalised recommendations
We prototyped a personalised product feed powered by AI. Instead of relying on keywords or filters, it adapted to user behaviour and intent. It matched people with the right products, even when they didn't know exactly what they were looking for.
The result: Buyers were matched more effectively with what they were really looking for. Even without the exact right search terms, the AI-powered feed helped connect people to products they didn't know how to search for.
3.
Self-inspection for cars
In high-risk categories such as automotive, we created a self-service inspection flow powered by AI and image recognition, enabling more transparent and secure car listings.
The result: Buyers could rely on the inspection flow, not just the seller's claims. Even non-experts gained confidence in remotely assessing a car's condition.
Our experimentation framework
We set up a practical experimentation model to help OLX² explore bold ideas quickly and responsibly. We used real user feedback and lightweight testing to reduce risk and build confidence.
1.
Define & prioritise hypotheses
We mapped all the assumptions around our ideas, from desirability to feasibility, and prioritised the most important but least validated. These became our starting points.
2.
Design & run the experiment
We used our experiment canvas to define what we wanted to learn, how to measure it and how long it would take. It was tightly scoped, controlled where necessary, and focused on getting clear, unbiased results.
3.
Analyse & turn insight into action
We translated evidence into decisions, adapting concepts, killing weak ideas early or doubling down where we saw traction. This closed the loop and fuelled the next iteration.
Working Across Disciplines
Collaboration was key to our success. I worked closely with:
Data scientists and ML engineers to align on AI feasibility
Product teams to connect ideas to commercial goals
Researchers and designers to ground concepts in real user needs
We ran co-creation sprints and facilitated working sessions to keep momentum and alignment high.
What I Learned
This project taught me how to lead product discovery in fast-paced, ambiguous environments. I learned how to design AI-driven product concepts in close collaboration with data scientists and ML engineers. Together, we aligned user value with what was technically possible.
I also developed a deeper understanding of how to de-risk business ideas through structured experimentation. Using Strategyzer's 'Testing Business Ideas' as a foundation, I helped OLX² frame assumptions, design lean experiments, and run tests from start to finish. We turned evidence into confident product decisions.
It's an experience I continue to build on today, helping teams bring clarity and momentum to early-stage product innovation.