The challenge
robo_Folks could already help users with technical tasks and, in some cases, perform actions on their behalf. But its conventional chat interface did not make those capabilities visible.
Users had to know what to ask before they could benefit from the assistant. The experience felt static, easy to overlook, and gave little indication that robo_Folks could help users get work done — not just answer questions.
The challenge was not to make the AI smarter.
It was to make its capabilities visible.
From chatbot to AI agent
I redesigned robo_Folks to give it a stronger presence inside the Cyber_Folks product.
My focus was the experience around the AI: visual direction, interface hierarchy, interaction patterns, motion, personality, and capability discovery.
I didn’t design what the AI could do. I designed how people discovered that it could do it.
Don’t make users guess what to ask
An empty chat puts the entire burden on the user:
“What can I actually ask it?”
This was especially limiting because robo_Folks could do more than answer questions. It could help users with technical tasks and actions.
I introduced short action pills as direct entry points into the experience.
Instead of asking users to understand the system first, the interface surfaced useful starting points directly in the chat.
The goal was simple:
Show, don’t explain.
The first interaction should already answer:
“What can robo_Folks do for me?”
Making an AI feel alive
Animated eyes that follow the cursor
This small interaction gave robo_Folks a sense of presence. Instead of a static widget, it felt like a responsive character inside the product.
The interaction was designed to attract attention without interrupting the user, make the assistant approachable, and create a reason to engage.
Quick-action system
A grid of categorised action tiles — hosting configuration, domain management, SSL, email setup, and more — let users tap into a topic without writing a single word. This removed the blank-input friction that made the old widget feel uninviting.
The goal wasn’t decoration.
It was to create a reason to notice the assistant.
A visual language for AI
At the time, the wider Cyber_Folks interface was restrained, flat, and functional.
For robo_Folks, I deliberately introduced a brighter, more expressive direction. A vivid gradient, motion, and character made the assistant distinct from the surrounding product.
Colour + motion + character became the foundation of robo_Folks’ identity
The result was an assistant that felt less like a generic support widget and more like a visible product feature with its own personality.
Designing the agent, not just the UI
AI products can easily feel interchangeable. A chat window and message bubbles alone do not create a memorable experience.
I designed how robo_Folks appears, reacts, attracts attention, introduces its capabilities, and communicates through motion and visual hierarchy.
My role included
- UX/UI design
- Visual direction
- Interaction design
- Motion
- AI personality
- Capability discovery
- Gradient and colour-system direction
One assistant, a broader visual direction
The visual language created for robo_Folks later influenced work beyond the assistant itself.
I shared the direction with other Cyber_Folks teams, including Shoper. Rather than copying one exact gradient, teams could create their own variations based on their product and group colours.
What started as an AI assistant redesign became part of a broader, more expressive visual direction across the Cyber_Folks group
Outcomes after launch
Following the November 2025 launch, robo_Folks became a meaningful part of the support experience.
3.2 → 4.7
Trustpilot rating after launch
53%
of all chat conversations handled by robo_Folks in the latest measured quarter
−33%
support conversations across the measured period
11,959 → 8,032
These are product-level outcomes observed after launch and should not be attributed solely to the redesign.
What I learned
AI UX is not only about the intelligence behind the model. It is also about how clearly the product communicates what that intelligence can do.
An assistant can have powerful capabilities and still feel limited if the interface does not expose them.
For robo_Folks, the design challenge was creating a visible relationship between:
Capability → interaction → trust → action
If I continued the project, I would explore proactive, contextual capability discovery — surfacing useful actions based on what the user is doing, instead of waiting for them to start a conversation.
Don’t make users learn what the AI can do. Make the product show them.