AI coaching assistant

I designed the user interface and experience for the AI coaching assistant, collaborating with the product owner from benchmarking similar tools to a design approved for launch.

As an enterprise tool used across the organization, it had to meet accessibility requirements, and I accounted for accessibility throughout the design.

Prepare for the conversation. Then rehearse it.

A manager may need help deciding how to approach a performance concern, or want to rehearse the conversation before having it.

The AI coaching assistant brings both activities into one experience: ask for coaching guidance or practice a scenario with a simulated associate.

It was also built into the tools team managers already use for their one-on-one meetings, so they could prepare without leaving their usual workflow.

Another goal was for it to stand out from the other AI tools available, with an interface and experience designed specifically for team managers.

I designed the user interface and collaborated with the product owner to align the experience with the coaching use cases. Team managers are the direct audience, within a total pool of 8,000.

Learn from what exists, then stand apart.

Before designing, I used AI tools to benchmark similar tools within the company and analyze their design styles. That showed me how to make the interface intuitive and visually distinct from the other coaching simulators.

I collaborated with the product owner on a consultation basis to refine the user experience and keep it aligned with functional goals. The leadership team and the product owner reviewed the design and approved it for launch right away.

Work through a real concern.

The conversational experience lets managers ask about situations such as an associate prioritizing speed at the expense of customer-interaction quality.

The interface separates the manager’s question from the AI’s response and keeps a follow-up input available beneath the exchange. In the product example, the AI coaching assistant organizes guidance around the Situation-Behavior-Impact framework, followed by suggested ways to collaborate on improvement and provide support.

The interface also supports a dark mode.

Give the practice a structure.

The scenario-practice designs let managers choose both a situation and an associate response style. The interface explains the roles: the manager conducts the coaching conversation while the AI plays the associate.

Scenario context remains available alongside the exchange. Custom inputs let managers describe situations beyond the predefined options.

The designed flow includes an explicit action to end practice and request AI-generated coaching feedback.

Make the limitations visible.

The assistant answers from a knowledge base of the leadership development team’s own documents, so its guidance reflects how the organization approaches coaching.

The actual product screenshot places two important expectations near the conversation:

• AI can make mistakes, and important information should be checked.
• Chat history is not saved in the represented version.

These messages give managers context for using the guidance and deciding what information to retain. They describe the interface’s expectations, not a technical guarantee about safeguards or data handling.

AI was the right fit because the assistant draws on trained, customized data from the leadership development team’s knowledge center. That makes its answers far more accurate, and in line with policy, than a general-purpose tool or static guides.

The trade-off is coverage. Guardrails stop the assistant from answering certain questions or using certain words, so it can’t answer everything. That was accepted to keep its guidance within policy.

Adoption in the months after launch.

Over approximately 6 to 8 months following launch, more than 5,000 of the 8,000 eligible managers used it, with an average of over 450 daily users and sustained daily engagement.

These metrics demonstrate adoption. Measuring improvements in coaching quality or associate performance would require separate evidence.

My contribution was the user interface and experience design, in collaboration with the product owner. This case study does not attribute the underlying model or every AI-generated response to me.