Project Overview
Apromore is an enterprise process mining platform that enables analysts to visualize, analyze, and optimize business processes using real operational data.
The Problem
Apromore is an enterprise process mining platform that enables analysts to visualize, analyze, and optimize business processes using real operational data.
While the platform already empowered users to explore complex models and uncover inefficiencies, doing so still required significant time, technical expertise, and domain knowledge.
As generative AI became more accessible, the business saw an opportunity to augment analyst workflows with a smart assistant, one that could answer questions like:
- "Why is this step taking so long?"
- "How can I improve this process?"
This led to the need of an AI chatbot called Apromore AI.
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
The Scope & The Squad
The problem wasn't just poor UX. Many user frustrations were rooted in the limitations of the Gen AI infrastructure itself. This meant we had to design the UX in a way that worked with those constraints, not against them.
- High User Expectations – Users expected ChatGPT-like responsiveness and contextuality.
- LLM Infrastructure Constraints – The backend lacked streaming and had long response latency (~30s).
- Accuracy Limitations – AI responses were only accurate ~30% of the time, mostly for repeated queries.
- Tight Release Timeline – The project was roadmap-locked, with limited room for backend changes.
This project was a close collaboration between product, design, and AI engineering. I contributed as the Lead Product Designer in the product trio, reporting to the Head of Product Design. I collaborated closely with the Product Manager and Lead AI Engineer to shape the user experience within real technical constraints. Together, we took a lean, cross-functional approach consitituting 4 sprints. Our shared focus on feasibility and user value helped bring Copilot to life on schedule.
The Design Sprint
The project involved designing an intuitive, conversational AI interface within the architectural limitations. We followed an approach to translate backend constraints into opportunities for UX innovation.
• Manual Context Selection – Users could select a category or define a context scope before submitting a query, improving precision.
• Guided Prompt Templates – We introduced structured query options to steer users toward high-confidence answers.
• Fallback Patterns & Feedback Loops – Designed graceful error states with helpful messaging and built-in feedback capture.
This approach significantly reduced dependency on generative complexity, allowing us to ship usable, intuitive flows on time.
The Outcome
Copilot launched in 2025 as a core part of Apromore's product strategy, positioning the platform at the forefront of AI-powered process improvement.
Key Results
- Prototype UX fully approved after user testing for development and delivered on time
- Increased task completion rate and reduced query retries in test sessions
- Guided prompts led to reduction in LLM load
- Positive user feedback calling it "an elegant solution to a frustrating problem"
