Maven is changing the way RBC employees work, using GenAI.

The opportunity

Generative AI (GenAI) has transformed how we search for information and use generated content to make informed decisions more efficiently.

RBC has an opportunity to use the benefits of this emerging technology to streamline work processes and simplify daily tasks for employees.

How can we use GenAI to change the way employees work at RBC?

Initial team

👩‍🎨 1 Designer (me), 1 Senior VP, 3 Senior Directors.

Current team

👩‍🎨 I’m leading the design team with 5 designers. 75 employees for project.

Timeline

August 2024 to now

Result

✅ Leading the design for one of RBC’s largest GenAI initiatives to improve workflows for all employees. My prototypes got buy-in from several executives for this project.

✅ Reduced traditional support inquiries by 30% during the product's pilot launch for 60,000+ employees.

✅ Developing UI/UX standards for AI-powered internal tools to enable efficient scaling and seamless integration with legacy systems. Created 20+ templates to date being actively used by 5 teams.

✅ Created the 3 year vision and transitions for this project. Currently executing on the first few iterations.

The biggest challenge

 
ChatGPT vs. Maven

Commercial GenAI tools like ChatGPT are good at providing summarized results for a wide range of topics. But, the results become less reliable as users ask more specific or complex questions. The burden is on the user to check for accuracy.

When creating an internal GenAI tool, the company is responsible for accuracy instead of the user. If an incorrect response causes an employee to perform a task improperly, the bank is liable for the consequences.

So, when developing our product, we had to rigorously test each model with target users to minimize incorrect responses, support the expanding use cases across different departments, meaningfully improve employee workflows based on thorough research, and make it easier for all employees to find general company wide information.

My vision for the product design was driven by these objectives.

“To create a simple, trustworthy, and personalized experience that helps employees work smarter, find information faster, and confidently use AI for work.”

Key Design Component #1: Innovative features

Innovative features are broad use cases that are applicable to everyone at RBC (e.g. People Search is used by employees to find information about their colleagues).

Incorrect answers in these cases won't have severe consequences but these interactions are often employees' first experience with our tool and incorrect responses can make them lose trust in using it for their work.

I included sample questions, controls, citations, feedback options and other governors to make sure employees could understand how the AI currently works and what to expect from it.

Sample questions to get users started
Results with sources and feedback options
Controls when there are multiple results

Key Design Component #2: Business Functions

Our overall product strategy was to start improving the workflows within a business function, expand to interdepartmental workflows, and finally merge all cases relevant to an employee into a unified search experience. Pilot program participants would have access to all business functions and models applicable for their roles. They can report feedback for each response as they try to use it to do their daily tasks. This feedback helps us improve our responses and reveals any gaps in their workflow processes.

User selects a business function (T&O) to see relevant models & use cases (Enterprise Identity Access Management)
All models available for a user via New Chat

Users can also easily access these models through a chatbot integrated into their everyday websites. This allows users to test our models and technology without switching to a separate platform.

Benefits chatbot placed in existing HR websites.

Key Design Component #3: Building fast with templates and a comprehensive UI kit

Our team has to support new use cases every month. Each use case could consist of the following:

  1. Multiple AI models

  2. Search, chat, and conversational mode for each model

  3. Version control for data and documents

I created a comprehensive UI kit featuring standardized templates for each use case type, developed with foundational design elements and components.

The templates offer a standard level of support and understanding for a new model so people can start using it immediately. Then, we refine the UI based on user feedback.

Example 1: Standard way to select a different model and mode within a business function
Example 2: Standard "People card" when employee name was associated with a result
Example 3: Standard way of displaying example questions

Results

✅ Leading the design for one of RBC’s largest GenAI initiatives to improve workflows for all employees. My prototypes got buy-in from several executives for this project.

✅ Reduced traditional support inquiries by 30% during the product's pilot launch for 60,000+ employees.

✅ Developing UI/UX standards for AI-powered internal tools to enable efficient scaling and seamless integration with legacy systems. Created 20+ templates to date being actively used by 5 teams.

✅ Created the 3 year vision and transitions for this project. Currently executing on the first few iterations.