The challenge

Aviary AI helps financial institutions deploy custom AI voice agents. Previously, staff had to manually walk clients through complex prompt creation. I redesigned the experience into an intuitive 6-step wizard that uses conversational UI to automate onboarding.

Before the redesign, creating an AI voice agent required manual onboarding from Aviary’s internal team. Users (primarily bank and financial managers) were expected to write raw prompts without much context or guidance, which led to confusion and inconsistent results. The interface lacked structure, clarity, and real-time feedback, making the process feel overwhelming for non-technical users. As a result, onboarding sessions were long, and support teams were routinely pulled in to walk users through the process one-on-one.

‍

Research & analysis

We started with interviews across three groups: bank managers (end users), Aviary’s customer success team, and internal AI trainers. Common themes included confusion around prompt structure and uncertainty about how inputs affected output.

We then analyzed onboarding sessions and support tickets from the previous 6 months. Data showed that over 70% of users stalled within the first two steps, and more than half needed help phrasing their agent’s purpose correctly.

‍

Concept & strategy

I designed a conversational wizard where users progress through six short steps, each framed as a mini-chat with the AI. This broke down complex concepts into bite-sized inputs and simulated the natural flow of a guided conversation.

Key Features:

• Conversational UI for input collection
• Visual progress tracking
• Pre-filled examples and contextual tips
• Live preview of the agent’s evolving responses
• Option to revise steps before finalizing
• CSV mapping key for importing unique data types

‍

‍

Reflection

This project demonstrated the power of conversational UX in demystifying complex, technical tasks. By translating dense AI concepts into short, guided dialogues, we made the product more accessible for financial professionals—without sacrificing depth or control.

The project was successful, with a 78% decrease in onboarding time, 89% reduction in onboarding help requests, and a fully self-service agent creation flow.