NexusGPT
Enhancing first-crack AI agent usability
with guided prompt design
Timeline
2 months (Dec '24 - Jan '25)
My role
Product designer
Team
1 PD, 1 PM, 1 DEV
Status
Under-development
Background
1/7
Problem
3/7
AI agents aren’t finished when they’re deployed.
Just like human employees, they grow through feedback.
As they interact with real users, they encounter edge cases, learn new tasks, and develop blind spots.
But improving an agent isn’t as simple as checking a box. It means updating its brain — the prompt.
Problem
3/7
Problem
3/7
In theory, Nexus supports this loop—but in practice, the system makes this incredibly difficult.
But don't just take my word for it. Here's Justine, a customer support manager, to share what her experience was really like.
Problem
3/7
Justine's experience
Agent is deployed
Week 1
Issues start cropping up
Week 2-3
Fixes attempted
Week 4
Real consequences
Week 5
The agent works well and receives positive feedback.
Confusing responses. Edge cases encountered.
Unnecessary escalations to human support.
Repetitive or robotic tone.
Justine edits the agent’s prompt.
She’s overwhelmed by a wall of text.
She makes a hesitant change.
The agent’s logic breaks the next day.
Justine loses trust and stops editing.
😄
🤔
😥
💀
Justine's feelings throughout
Justine wasn't alone. Countless managers faced this same uncertainty every day.
The gap
Justine's story is far from unique. It exposes the fundamental flaws in how Nexus handles prompt maintenance, making agent improvement unnecessarily complex and risky.
Problem
2/7
Problem
3/7
We studied how users maintained agents and where their understanding fell short.
Problem
3/7
Lack of prompt expertise
Most managers don’t understand prompt engineering, leading to guesswork and ineffective updates.
Feedback isn't actionable
Feedback from conversations and tasks isn’t mapped to the right parts of the prompt, forcing users to hunt for issues blindly.
Fear of breaking things
Even small edits can cause unexpected failures, making users hesitant to improve their agents.
Problem
3/7
Beyond user challenges, the Nexus platform itself created friction that made prompt maintenance slow, risky, and unintuitive.
Problem
3/7
Unstructured prompts
Prompts were presented as dense walls of text, with no hierarchy or modularity, making it difficult to locate or edit the right sections.
Poor interface design
The prompt editor functioned like a plain text field, offering no tools or guidance tailored to prompt maintenance.
No rollbacks or safety nets
There was no version control or rollback option if changes caused performance deterioration, discouraging experimentation.
Our goal
It was clear that Nexus needed a simpler way to maintain and improve agent prompts. Our goal was to build a tool that is structured, easy to use and safe so managers can improve their agents with confidence.
Discovery
3/7
Problem
3/7
We have text editors for communicating with people. Code editors for instructing computers. But nothing yet for AI agents.
Typing started with a simple notepad, but over time it evolved to serve new use cases like writing, messaging, and coding. Now it has evolved to prompt-writing, and the next step in that evolution is here: prompt editors.
Problem
3/7
Problem
3/7
Learning how AI reads its instructions, and how formatting and structure affect it
AI reads many formatting cues the same way humans do — lists suggest order, capitalization signals importance, quotes indicate exact wording. The more intentional the structure of a prompt, the better the AI can follow it.
Problem
3/7
Return home
Back to work