Ella Asks

Experimentation with using Ella to ask clients questions in order to better target users with personalized insights.
SketchUserTesting.comInvision DSMInvisionProtopie

Who is Ella?

Ella is our virtual agent — a digital financial coach — that interacts with clients on the secure Sun Life website. She helps users better understand their finances by surfacing messages based on a decision engine. When she asks the client a question and provides a personalized insight on their next sign-in, we see a lift of over 300% in engagement.

My task was to take those initial findings, and experiment with how Ella interacts with clients — how we ask questions and provide insights. I needed to keep in mind the technical limitations of our current website but also look ahead to our future-state vision. My team sought a solution that would incorporate contextual messaging and positive reinforcement, while pushing the boundaries of Ella in terms of her personality, content, and tone of voice.

Miro board showcasing user interview insights and a jobs-to-be-done frameworkMiro board showcasing insights from the original pilot in 2020Miro board showcasing a design brainstorming activityMiro board showcasing early sketches, team brainstorming, and mid-fi designs with notes


To learn how current Sun Life clients perceive Ella, we conducted a set of moderated user interviews. We heard that clients were open to sharing data with Ella, equating it to how they would interact with a financial advisor. The biggest design challenge was that — in essence — asking the user questions to serve a business goal was not user-centric and could be a major point of frustration along their journey. Regardless, that was the ask. Conscious of creating more friction, I pushed for a design that was scaleable, intuitive, and pioneered a new interaction, delivering value and continuing to build the client's awareness of Ella.


The first iterations of the design used a modal pattern. This zone is surfaced every time a user signs into Sun Life and was used to pilot the initial concept last year. With that said, it is an in-demand zone and so our team opted to explore new types of interactions. We were excited by badge-based virtual agents, located in the lower right-hand corners of the screen (such as Workplace or AsiaMiles). This would bridge our current state to a future state and enable us to create a modular zone that could ask questions, provide insights, and host our chatbot.

Defining an MVP

The above vision excited our stakeholders and tested well with clients. Users responded best when there was a balance of being asked questions but also receiving help and value. As is often the case, our team had to remove some of our favourite features in order to create a bite-sized MVP to launch and learn with, including the modular nature of the design.

I encourage you to play around with our prototype. I built this for a senior-level presentation showcasing our MVP  design — it was built using Protopie, which has been a blast to learn.

Where it goes next...

The project is currently in development and the MVP is slated to launch with a small batch of users later this Fall. We will be tracking metrics such as how many questions a client answers per session and what the lift to engagement is on their next sign-in. These can then be compared to the original pilot, enabling us to better iterate on the strategy and design.

Back to Projects