In September 2024, I began my role as a product design intern at KeepSimple.ai, where I contributed design perspectives to the tech-savvy startup.
My first project specifically focused on updating one of the company's core features on their AI agent development platform——the Knowledge Base, a library where users can upload external and their own data in various forms as "pieces of knowledge" for LLM processing.
Working alongside a product owner and developers, I led the effort to reshape the experience of how users add, view, and manage their "pieces of knowledge", transforming it into an intuitive and streamlined journey.
The Context
problem space
A design lagging behind.
KeepSimple.ai was gearing up for its next MVP launch in a few months. While the backend of the Knowledge Base feature was in place, its design had been left behind. The outdated interface felt disconnected, and needed a fresh, swift update to align with the company's AI agent product.
Yet, this wasn't a simple refresh; it came with a pile-up of constraints and challenges:
Legacy experience left unfinished. The interface was a patchwork of functions and elements—it wasn't built to last. It needed a full overhaul than merely an update.
A blurry brief from the start. No one really had a clear vision of what a "modernised" design should look like. Go figure.
Scalable design. It had to integrate smoothly with existing MVP features while being flexible enough to evolve with future product developments.
Rigid time limit. Everything was running against the clock. The project must be delivered within weeks and ready for handover ASAP.
The only "design guy". Working alongside a tech-savvy team, yet I was the sole designer in the mix. I had to navigate the design process on my own.
Design Process
Designing at a startup has been a delight, with its own unique rhythm in problem-solving. Through many struggles and lessons along the way, I was able to grasp the nuances and embrace an unorthodox design approach in this project: think and act, then retrospect.
Think
Understand the problem
With an unfinished legacy interface and a blurry vision, where do you even start going about redesigning a feature from the ground up? I had no clue either, so I did a lot of research at first, trying to understand how others were tackling the design on similar features.
Standing on the shoulders of giants was helpful. It made me realised that it was more of filling in the gap rather than inventing from scratch. The Knowledge Base was not a new concept in the world of AI products. The key was to build upon what already existed in the industry and pivot where necessary.
Understand the users
The Knowledge Base is for both tech-savvy experts and non-technical AI enthusiasts. It's a space that lets both user groups easily add, view and manage their "pieces of knowledge" as the building blocks of their AI agent development process.
The challenge was to shape the space to feel intuitive for everyone—something that just clicks and flows naturally and effortlessly. Meanwhile, it needed a touch of depth, offering geeks to explore and customise.
What it led to design
By walking through some of the most powerful AI agent platforms on the market, it has shed a light on the design direction, clearing the fog for a "modernised" Knowledge Base:
clarity
Clean and flat interface designs that deliver clear and concise information.
Flexibility
Allows for customisations and advanced control, so users can manage their knowledge more freely.
Efficiency
Informative and straight-forward workflows that streamline the experience.
Act
Pressing the fast-forward button
Once the design direction was made clear with the product owner and the team, everything started to move quickly. It's like pressing the fast-forward button: go and get it done. Working in a crowd of startup minds can really fuel you to put nose to the grindstone and keep things moving.
I started off with wireframes to map the skeleton for a new Knowledge Base, and in the next second, before I knew it, feedback from the product owner was already rolling in. With communications, I turned wireframes into prototypes, prototypes then turned into refined flows, and each iteration brought the redesigned feature closer to life. The pace was intense, but the momentum was a beauty.
Align on scope, with the team
There are countless moving pieces in a startup and sometimes you have to battle for resources. My trick? Feedback loops were immediate around here, and I made sure to leverage them to the fullest. I spearheaded meetings to regularly sync my designs with the product owner and the team, ensuring we were all aligned on the scope and direction.
I jumped between iterations and standups, shaping pixels to meet the functional specs. In the meantime, I kept updating with these go-getters and grounded my designs in what the business needed, what the development could handle.
Retrospect
Rounds of rapid prototyping have made the Knowledge Base experience more refreshing and functional than ever. But a voice within reminded me that the sprint wasn't endless. I took a step back to review my designs, and that's when the pace started to slow down, giving me the space to identify any blind spots.
Is the journey intuitive enough? Have I provided users with clear cues for interactions? Are the users empowered with flexibility of controls? Is the design scalable? Have I covered all potential use cases? Is there any more features we can add to the current designs?
It's a race against time, and a marathon for quality.
Final Design
Takeaways
Be ready to fail.
I have learned so much by actually doing this process, and grow so much as a designer and a thinker without realising it. Don't be afraid about having failures, instead dive headfirst into them and use it to learn.
Design system matters.
I will write a note and make sure I start with building a design system, if I could do the project again. A design system is a common language for everyone in your team. It provides framework for consistency while also allows enough flexibility to address task-specific design challenges. Having a unified framework from the outset would streamline the design process so much more, and help scale the product more efficiently as it evolves.