The challenge
UNDP came to us with a brief for an online engagement platform they could use to crowdsource youth opinions on different policy issues. We had 6 months to build the MVP from 0-1. CogCo had already been running participatory surveys with UNDP, but building our own platform would enable us to incorporate our own UX and functionality improvements.
While UNDP was the client who commissioned the product, the wider goal was to enable any organisation to run discussions with its people.
The result
An desktop and mobile compatible platform that has been used by UNDP and other organisations to generate discussion across industries and countries on a variety of topics. In 2025 alone, UNDP reached over 15, 000 young people in Asia and the Pacific.
Key features include the ability to moderate statements before they enter the discussion space, the ability to ask participants questions before they join the discussion space, and cluster analysis on the results to create opinion groups.
[insert image of ekota]
The approach
Following initial discovery work which involved interviews with employers and employees, the insights from which were used to create 6 personas, I ran an ideation workshop to brainstorm features relevant to each persona’s goals, using the MoSCoW method to prioritise which features to take forward into design. Needs were viewed using the jobs-to-be-done approach to ensure features were tied to a clear action and purpose.
The reason the MoSCoW method was chosen was because it allowed us to quickly see a a visual of where personas overlapped in their needs and which features were coming up as important to all of them, which would directly inform the features built. The design team were in this workshop together (myself and 2 other designers) and ensured we had an open discussion about the features, the pros and cons of each, before narrowing them down.
After prioritisation, I moved into sketching user flows. thinking about the journey each person would take and how they would interact with the proposed features. At this stage, I also sketched iterations of what different features could look like.
[insert images of ekota sketches]
I then moved into Figma and began working on the building the design system and the high fidelity designs for desktop and mobile. For the design system, I began with typography and colours as the basis to work from. It was kept simple with a brand colour for ekota, a colour that organisations could customise with their own brand, state colours and neutrals. This simplicity was suitable as ekota was a product that would not have a large number of pages or flows and dashboards were also designed to be simple and clear because not everyone reading them would have a data background.
The components in the design system were built and iterated on as we were building the product. The reason for this approach was because it allowed me to test components as I went to see how they fit the design and what was needed along the way, as opposed to building an entire system and constraining myself to that with the design. It also meant I was able to move into high fidelity prototyping earlier because building an entire design system upfront would have meant spending more time working with sketches to figure out components needed.
[image of design system]
Following design was development. The developer we worked with built a playground site for us to use throughout the build, testing features as they were built. This made the process collaborative, iterative and interactive as I was able to play with the features in real time to provide feedback ensure the vision for the product was executed well before the full launch.
Ekota has now been used across various countries and sectors over the last 4 years, crowdsourcing opinions from tens of thousands of people on different topics.
Reflections
Ekota was built in 2022, before the boom of AI. If I were to rebuild it in 2027, here’s where I would incorporate AI into the workflow, and why.
Initial research and discovery: While I would maintain running interviews with people as I believe that human to human conversation is invaluable for nuance and insights, I would use AI to analyse and consolidate the research into jobs-to-be-done which could then be taken into the feature prioritisation workshop where human discussions would be had to make the final call.
I would also use AI in market research, prompting Claude with its deep research feature to look at existing products in this space and compile a spreadsheet for me with key information such as pricing, key features and positioning analysis.
Design: In the design phase I would use AI to help turn designs into interactive prototypes earlier on so they could be played with in the web by the rest of the team and the developer to get a better sense of the final product. I would also use Claude as a brainstorming partner to test out variations of components, feeding in the design system to quickly generate ideas I could then iterate and refine.