top of page
Bank Connect | Intuit
AI automation that makes connecting your bank to QuickBooks fast and trustworthy
Team: Quickbooks Accounting Automations at Intuit
Duration: 12 weeks
Role: As a product management intern, I led the redesign of QuickBooks Online's bank connection flow, increasing connection rate by +15% and launching it across 5 regions (US, CA, UK, FR, AU).
I collaborated with engineers to determine feasibility & backend logic, with accountants to map decision logic, with data analysts to create tracking/testing requirements, with global PMs to enable scalability to international regions, and with content designers on UX copy.
PROBLEM
15% of users dropped off before seeing QuickBooks' core value.
Connecting a bank unlocks QuickBooks' 10+ accounting automations at scale. However, funnel analytics showed that user
drop-off concentrated on one screen, account mapping, where 38% of users left.
RESEARCH
Business owners didn't speak accounting. Accountants didn't trust what they couldn't verify precisely.
User interviews and live testing with 10 small-business owners and 5 accountants showed two opposite needs: one group wanted speed and simplicity, one group wanted accuracy and control.
Understanding the manual experience:
Clicking 'Add new' opens pop up overlay ->
1. Two steps at once (selecting accounts, then each account type)
2. Lack of contextual guidance (unexplained steps, weak visual affordances)
3. Complex accounting terms that small business owners don't understand
4. Lack of focus with information overload & unclear CTAs
I also discovered a missing use case: Existing connected accounts were creating duplicate transactions.
SOLUTION
Automated Bank Connect
An AI-driven experience that helps small-business owners (SMBs) seamlessly connect their bank transactions to Quickbooks
1. Select accounts
Pre-selects new accounts and a suggested import date, with a note on what each choice means.
3. Existing accounts
Suggest the most likely match with evidence. Let users confirm or create a new account.
2. Account type
Pre-fill each account from bank data and allow users to verify or correct it. Show guidance when user input is needed.
4. Confirmation
A final review that educates users on what will be imported and what's reversible, with every account editable side by side.
DEFINE
Designing the logic behind the automation
I worked with accountants to map every data scenario into a decision tree:
Account selection:
-
1 : 1 or confident matches -> pre-fill suggestion
-
1 : many or partial matches -> show relevant dropdown options
-
No data -> manual options with guidance
Existing accounts:
-
Pre-fill suggestion if provider AND account type match
-
Otherwise, dropdown with 1) previously connected and 2) other existing unconnected accounts
The tree became the flow, turning one screen where 38% of users dropped off into focused steps.
DEFINE
Three tradeoffs behind automation people trust:
Suggest, don't skip
SPEED VS CONTROL
FOCUS VS TRANSPARENCY
Show the why
AI pre-fills every choice, but users still see and confirm each one.
Suggestion explains itself in simple language so users understand AI's reasoning. Otherwise, keep text minimal.
EDITABLE LATER VS IRREVERSIBLE
Match friction to stakes
Irreversible actions get evidence & a final side-by-side review.
USER
TESTING
Full automation made users faster, but less confident.
The first version automated/skipped as much as possible. In round 1 of testing, 6/10 users preferred the old design (without automation).They kept going back to check what the AI did and didn't trust steps they couldn't see, so I redesigned a more guided and transparent flow.
REDESIGN
1) Select accounts
%20select.png)
SPEED VS CONTROL
Users didn't trust actions being skipped. I made sure each pre-selected choice stayed visible and allowed users to unselect.
FOCUS VS TRANSPARENCY
Users need to understand the impact of high-consequence actions and reasoning for suggested date. Already linked accounts can be collapsed for focus.
2) Account type
%20type.png)
FOCUS VS TRANSPARENCY
Users need to understand why AI chose something or why it needs their input. Drop downs show with suggestions based on data scenario and short sub-descriptions educate users on complex types.
SPEED VS CONTROL
Confident suggestions were kept visible, and allow users to verify/change them.
%20dropdowns.png)
3) Existing accounts
%20existing.png)
EDITABLE LATER VS IRREVERSIBLE
Users accepted pre-filled suggestions without checking. For an irreversible choice, users need to consciously choose. I replaced one skimmable drop-down to two clear options. Helper text let users understand why an existing account is suggested.
4) Final review
%20confirm.png)
EDITABLE LATER VS IRREVERSIBLE
Allow users to check AI's work before it was final. A final review clearly shows how each account type maps into QuickBooks and what can be undone. It distinguishes which are new vs existing accounts and explains what will be imported.
RESULTS
9/10 users preferred the AI-driven flow
After launch, bank connection rose 15%. We rolled out gradually with A/B testing and tracked guardrails like correction rates and duplicate-transaction tickets to catch where the AI got it wrong.
1. Enabled automation coverage of 98% by using bank data to map a decision tree for data scenarios, working with developers and accountants to define rules, use cases and edge cases
2. Defined complete backend, design, tracking, testing requirements in a full product requirement document
3. A redesigned user experience increasing completion rate by 40% for SMB owners, validated through 20+ user interviews
4. A prioritized global roll out plan to UK, CA, AU, FR, with A/B testing, working with data analysts and global PMs
5. Documentation of research & data for the user problem
LEARNINGS
Trust & transparency above all (especially in Fintech & AI)
Through iterations of redesigning and testing a new user experience with users (both SMBs and accountants), I learned that what's consistently important in fintech/automation is creating a sense of transparency - about what is being automated, what is occurring and especially what is being hidden. A trustful & friendly experience is priority, even if it means automating less. User guidance & education is also key in helping users complete an automated flow. This also extends into how to best foster team collaboration, as I learned that showing up as my authentic self with my honest questions and ideas was key to building trust on my team.
Re-strategizing through information roadblocks is essential
Throughout my internship, I encountered moments where I did not have enough / the right information needed to implement, explore or build on my ideas. Especially in a large company with large databases and with products owned by multiple large teams, seeking the right information or determining whether a piece of information exists can be difficult. Learning to navigate these situations by being resourceful, creative, communicative when reaching out and reframing what I have in face of missing information was an invaluable skill I learned.
Alignment is key
In a team with various cross-functional stakeholders, I learned that collaboration & alignment is key to success - understanding each person's goals/needs in the project, involving relevant stakeholders in decision-making & ideation, setting clear objectives, regular communication, sharing documentation that ties back to the user problem, and bring a great team spirit. I am grateful for the mentorship I gained through connecting with PMs on my team on navigating problem to solution and how to learn & grow into a great PM overall.
Move quickly with purpose
With only 12 weeks in my internship, I had to prioritize based on deadlines, goals and impact - whether I pursued a side project, a new exploration or choosing to spearhead implementation vs new designs. Learning to move with velocity and start proactively, even when I don't have all the answers or testing yet, was important in achieving both impact and fulfillment.
Special thank you to: Ying (manager), Joseph (PM manager), Justin (UI/UX designer), Viral & Sandeep (Developers), Kasey & Kelsey (Content designers), Renee & Tianpei (Data analysts), Trevor & Pauline (Accountants), Eric (PM mentor)
That's me!
Let's connect!
If you like what you see, I'd love to chat!
Alternatively, here's a fancy sheet of paper:
bottom of page