Francesco Ricci | Graphic Designer

About Convo

We’ve all rehearsed difficult conversations in our heads, hoping to get them right when it matters. But what if AI could take that practice to the next level? This tool provides a safe space to try out conversations with team-mates, bosses, customers or anyone else. Here we’re using AI to provide an experience, not to shovel more content at people.

Project Overview

Having someone practice a conversation is so much more impactful than giving them a cheat-sheet or advice. It allows them to feel the emotion of a heated discussion, providing an experience as education.

As the sole Product Designer, I played a pivotal role in evolving Convo's product flow-embracing a continuous cycle of testing, learning, and refining to optimize the user journey.

This case study explores the key decisions, challenges, and insights that shaped Convo Ai experience into a seamless and engaging Coach for users.

Analysing User Behaviour

To effectively address user pain points, we first needed to uncover the underlying needs and motivations driving them. By clustering the qualitative data collected, I was able to pinpoint several key themes.

Unfamiliarity with Ai Coaching

The coaching has traditionally been an exclusive perk for high-level leadership. Since many users had never experienced Ai coaching before, they struggled to see its relevance—highlighting the need for improvements to clearly communicate the company value upfront.

Privacy Concerns

Coaching is a deeply personal and vulnerable practice. Given that users accessed Ai through their employer, many required clear assurances around privacy and confidentiality to feel comfortable engaging with the product.

Information Overload

Users often shifted between feeling overwhelmed by too much information and confused by a lack of clarity, making the coaching experience feel lengthy and complex. This underscored the need to refine information pacing and set clearer expectations.

Lack of Human Connection

Users repeatedly expressed a desire for more detailed information about their coach and background before their first session, as it would help them feel more comfortable and engaged.

Refining & Retesting

Having identified key improvement opportunities, we adopted an agile design approach through rapid ‘test-learn-refine’ cycles. Embracing short feedback loops allowed us to quickly validate concepts with users and ensured strategic, data-driven design decisions.

As part of a small design team, I was deeply involved in every stage of the process—from conceptualizing and refining ideas to leading and synthesizing user testing.

The Prototype

Designing a tool for practicing coaching conversations with AI involved several challenges. Ensuring realistic and varied interactions with AI coachees required sophisticated natural language processing. We also had to create an engaging user interface that would allow users to navigate easily while offering a comprehensive set of features like feedback and hints.

Test the Protype here→

Adding Customisation

In addition to the core coaching experience, we've also created a feature that allows users to customize AI-powered simulations. Here, users can design their own virtual coach and simulations by providing specific details about their goals and scenarios they'd like to practice. This feature offers a highly personalized, hands-on approach to learning and development, ensuring that every coaching experience is as relevant and effective as possible.

Final Outcomes

The project has been selected by Deloitte to be implemented for their upcoming cohort. Deloitte saw this solution as a powerful way to address the growing challenge of scaling internal coaching. Recognizing the need to develop confident, capable coaches, they were drawn to our innovative use of AI to simulate two roles in a classic coaching triad. From visual styling to tone of voice, the experience has been tailored to reflect Deloitte’s professional culture and learning environment, ensuring to fit within their internal development ecosystem.

Take a look at the final result right here.