Sidekick

Timeline

May 2025 - Present

Platform

Internal Product (SaaS)

Roles

Lead UX Designer

Collaborators

Project Manager

AI Specialists

Data Scientists

Context

Sidekick is Highmark’s internal AI assistant designed to help employees work more efficiently through conversational AI, document ingestion, and prompt management tools. The platform offers general and specialized chat sessions — including General, Code, Medical, and Policy modes — to provide context-specific assistance across the organization.

I joined the project as lead and sole UX designer after the previous UX team requested to be pulled due to shifting requirements, unclear priorities, and constant scope changes. My role involved stabilizing the design direction, enhancing usability, and introducing new features to improve accessibility and productivity for employees — while keeping development efforts grounded in user needs and core functionality.

Outcomes

Challenges

When I stepped in, Sidekick faced:

Goals

After understanding the challenges, I created some goals that would help ensure project success.

Design Timeline

Overview of the work I completed. Throughout the sprints, weekly sprint reviews were conducted to separate essential features from nice-to-have additions.

Small Changes, Big Impact

Initial changes carried immediate impact.

Workbook History

Initially, users only had the ability to open a workbook session by clicking on the tile. After researching, we found that users often utilize the chat history feature to find specific information from the session instead of starting a new conversation. We decided to mirror that capability in workbooks along with adding other improvements like the search bar.

From Tiles to Cards

In order to accommodate for added capabilities, both the workbook and prompt library tiles needed to be changed to card components. This was due to accessibility and usability concerns.

Session Settings

Introduced a temperature gauge to visually communicate the AI’s creativity level, helping users anticipate response style and reduce unpredictability. Embedded a settings icon directly in the chat input, giving instant access to key controls (temperature, system instructions, lite grounding). These changes increased transparency and allow users to develop a clearer understanding of AI, which is key to building trust and confidence in Sidekick.

Introducing Light Mode

Following feedback from the initial re-release, we identified a strong demand for a light mode. Leveraging the existing design system as a foundation, I developed a new color palette and implemented a light mode version of Sidekick. Early responses have been overwhelmingly positive.

Testing and Iteration

Despite not having the capacity to complete user testing sessions I still was able to gather insight in the following ways:

Leverage Super Users

Sat in on superuser training sessions to hear their feedback and ongoing struggles in the platform.

Surveys

We leverage MS forms and sent them out to the super users to gauge their thoughts about new and potential features.

Design Critiques

I held design critiques with fellow designers when I was working through features. Hearing their feedback allowed me to refine my work or pivot if needed.

Impact & Results (early outcomes)

Feature Prioritization

Clear prioritization aligned with user needs prevents wasted effort and design drift.

Avoiding Superuser Bias

Relying solely on the superuser group’s insights can introduce bias, so we’re actively engaging less frequent Sidekick users to gain a more balanced perspective.

Accessibility & Adoption

Platform accessibility improvements drove higher adoption rates and greater user satisfaction.

Next Steps

Home

Next Project