AI Assistant for Self-Care

PayFit Copilot is an LLM based assistant that aims providing instant and qualitative support to customers queries.

About the project

The project was about providing a fluid and smooth experience to customers in their journey to ask support. The goal was to replace an existing chatbot with tentacular rules and branches, hard to maintain, by an assistant that answers level basic legal and HR questions and redirects to customer support when it needs more expertise.

Company

PayFit EU

🔍 Context

The objective was to create a strong first impression with a fast, clear, and trustworthy experience, especially that an existing experience was already settled. The journey needed to convey reliability while also embracing a friendly tech aesthetic.

🧐How I've proceeded

Organization optimization : Split the team into Experience and Quality. I was mainly in charge of experience.
Discovery & Research: Conducted user interviews with user research lead and end-users to understand support journey pain points. Designed and delivered with the data teams research on the existing usage.
AI Strategy & Design: Collaborated with engineering to evaluate LLM capabilities and limitations. Designed conversation flows that felt natural while handling complex HR/payroll scenarios. Created fallback mechanisms experience for AI limitations including a live chat solution integration.
Cross-functional Execution: Led weekly alignment sessions between AI engineering, customer success, and design teams. Established success metrics combining technical performance (response accuracy) with user experience (satisfaction scores).
Iterative Validation: Implemented phased rollout with continuous user feedback collection.
International Rollout while considering local specificities
Feedback loops between customer success and product development for continuous improvement.

📈 Results

25% reduction in support costs through automated responses.
10,000 new users adopted the tool within 6 months.
Long-term AI roadmap validated by the executive committee.

AI Assistant for Self-Care

PayFit Copilot is an LLM based assistant that aims providing instant and qualitative support to customers queries.

About the project

Company

🔍 Context

The objective was to create a strong first impression with a fast, clear, and trustworthy experience, especially that an existing experience was already settled. The journey needed to convey reliability while also embracing a friendly tech aesthetic.

🧐How I've proceeded

Organization optimization : Split the team into Experience and Quality. I was mainly in charge of experience.

Discovery & Research: Conducted user interviews with user research lead and end-users to understand support journey pain points. Designed and delivered with the data teams research on the existing usage.

AI Strategy & Design: Collaborated with engineering to evaluate LLM capabilities and limitations. Designed conversation flows that felt natural while handling complex HR/payroll scenarios. Created fallback mechanisms experience for AI limitations including a live chat solution integration.

Cross-functional Execution: Led weekly alignment sessions between AI engineering, customer success, and design teams. Established success metrics combining technical performance (response accuracy) with user experience (satisfaction scores).

Iterative Validation: Implemented phased rollout with continuous user feedback collection.

International Rollout while considering local specificities

Feedback loops between customer success and product development for continuous improvement.

📈 Results

25% reduction in support costs through automated responses.

10,000 new users adopted the tool within 6 months.

Long-term AI roadmap validated by the executive committee.

← Previous Post

← Previous Post