Train QuickBooks
I designed a customer experience to showcase the power of logistic regression models and conducted 100+ customer interviews to nail the customer experience. This level of new automation in QuickBooks aimed to minimize manual bookkeeping and accounting.
Building the first logistic regression model in QuickBooks to reduce manual categorization of common expenses.
Duration
- February-June 2024
- Product designer & researcher & 10 weeks as a pseudo PM
Customer problem
- I am a small business owner using QBO and am faced with a wall of transactions that keep me from running my business. This leaves me disappointed because I expected QBO to do my bookkeeping and help save me time on manual review.
Business impact
- Today, only 30% of DIY users have completed 95%+ of their monthly transactions.
- QBOs first attempt to use statistical models to do work on bank transactions with no review required by users
Why is this interesting?
- This marks the first time QBO has introduced advanced AI within the transaction domain.
- This sophisticated AI has been in development for years, and I had the privilege of shaping the customer experience.
- I needed to translate our logistic regression models into an intuitive customer journey.
Pioneering QBO Automation
- I needed to determine the appropriate level of customer engagement when assessing our automation.
- Teams before me experimented with various basic automation features in QBO, setting the groundwork for more advanced solutions in the future.
- My challenge was to pinpoint the most critical steps for customers to engage with.
100+ Customer Interviews
Trust & control principles I developed:
- Prioritize trust & control from the customer's perspective
This principle will forever be our guiding star. We must earn our customers' trust before encouraging them to opt into automation. - Establish trust early, introduce automation gradually
Not all customers are ready to embrace new automation from day one. How can we effectively prepare initial users while also accommodating those who adopt later? - Build a robust foundation before expanding
Customers desire a seamless, end-to-end experience. We must integrate legacy automation with our new system to eliminate inconsistencies and ensure a cohesive journey.
First, we calibrate the model using a set of expenses that best represent the aggregate data from all linked financial accounts.
Customers will first preview the flow, followed by a preview of expenses set for immediate automation.
After the personal model has achieved the baseline training threshold, we prompt the customer to opt-into automation.
Content during opt-in is crucial due to the diverse interpretations of automation comprehension.
My favorite project to date!
Wow, what a wild ride. I never imagined working in an even more technical space than before. Accounting was challenging enough, but designing for logistic regression models took my work to another level. On top of that, I only had about a month to onboard and start crafting designs solutions.
I adore my PM, but when he went on paternity leave for 8 weeks during testing, I had to make product decisions with my engineers. This experience was a significant learning moment for me in this project and in my career as a senior designer. I was left to make product decisions and work closely with our AI team to ensure I was making the right choices for our customers.
Every day is day 1
Although we can’t reinvent the product and are always constrained by technology. This project was an excellent opportunity to embrace these constraints and focus on mastering the tech that underpins everything, ensuring success for both the team and myself.
This project reinforced the value of asking questions and allowed me to be openly vulnerable. Despite the final solution seeming straightforward, it involved many iterations. I found myself making product decisions typically handled by PMs or engineers. By the end of it I discovered adeep interest in designing for AI and would not have traded it for any other project.
San Francisco, CA
