NOMI Forecast
Incorporating AI into digital banking and empowering people to make better financial decisions.
Vision
Provide RBC clients with a forecasting tool to help them track payments, manage transfers, and balance income and expenses over time.
Team
Lead Product Designer, Visual Designer, Content Designer, Product Owner, BA, ML Engineer, AI Project Manager, System Architect
Objectives
V1 - Be aware of upcoming payments.
V2 - Balance their income and expenses over time so that they are never short of cash (i.e. never in overdraft.
V3 - Anticipate future situations, especially predicted shortfalls.
Problem Statement
RBC clients are forgetting to pay their bills. ending up in overdraft and then paying the NSF fees.
Clients don’t have a clear picture of immediate upcoming payments to manage their finances.
Business Problem: Number of calls to customer centre about NSF fees.
Desired Outcomes
Reduce: anxiety, effort to stay on top of finances, frequency and amount of overdrafts, and number of times the client doesn’t have enough to cover expenses.
Increase: automated transactions and bill payments, participation in Savings, overall financial well-being, sense of control
Best use of AI for customer experience
More than 2 million RBC clients have used NOMI Forecast since the 2021 launch. NOMI Forecast was developed alongside Borealis AI, a world-class AI research centre backed by RBC.
Product Canvas
Research
We started by running an empathy mapping and assumption slam workshop with other members of the bigger design team. The main goal was to understand the mental model of our RBC users. We ran 4 different phases of the user test. Started with the initial test to understand the user’s behaviour, then the second UT to test content, the third usability, and the last one again to validate the design solution.
Design Solution
We have developed a custom machine learning (ML) algorithm that learns from users' patterns and behaviour to predict their future bills and payments. Currently, for the minimum viable product (MVP), the system accurately predicts automated payments with 98% accuracy (+-2 days) and manual payments with 88% accuracy(+-2 days). Our approach aims to educate users gradually about this feature and how our artificial intelligence system works in the background to enhance their app experience. We are encouraging users to explore this new upcoming payment feature by incorporating carefully crafted content across different parts of the app that avoids technical jargon. Whether it's reminding clients when bills are due, helping manage cash flow, or anticipating needs, machine learning makes banking easier.
Early Sketches
Automated Payment
Manual Payment
Designed to Feed
I worked on ML design sessions with the engineers to understand the algorithm and explore ways to make it more intelligent. We tested different designs using user data sets to optimize the user experience. I aimed to design a feature that could learn from user behaviour. Additionally, we tested a payment value system by introducing ⭐️ to prioritize important bills and payments.
Missed Payments
The app notifies the user when they miss a payment by predicting the usual payment date. There are a few nuances in the features. This feature was thoroughly researched during the user testing sessions to get it right. If done wrong, the user might not be able to trust the predictions. Therefore, we have tried and tested it multiple times with a multitude of our app users. This is also the first step in understanding why users miss their payments, so the bank can help them avoid incurring any late fees in the future.
Blue-sky vision
We have established a long-term goal of creating a blue-sky vision to guide us along the way. We aim to reach the future of financial apps, which involves some level of autonomy, as quickly as possible.
Feature Release
Fall 2021