Designing AI-enhanced client onboarding that’s 40% faster
In the wealth management division of a global bank, we found that a major slowdown in the onboarding process came from mismatch data. Most of the back-and-forth between the person creating the workflow (maker) and the person reviewing it (checker) was just about fixing incorrect or incomplete data.
This not only slowed down onboarding but directly impacted the business goal of opening more new accounts per month.
How might we use AI to proactively improve data inconsistencies while preserving user trust in a high-stakes and highly regulated environment?
I led design for this initiative, shaping the end-to-end strategy for integrating AI responsibly:
Defined how AI should appear and behave in workflows.
Designed new patterns and components for AI interactions.
Collaborated with product, engineering, and compliance to align system reliability with user trust.
1. Building trust in AI
Explored what data the AI model is build upon on and how reliable its outputs were.
Design principles:
User should have autonomy: AI should never block submission and progress the ticket.
All suggestions must be explainable and transparent. For instance, adding source and confidence score.
Recommendations should only appear at high impact error points.
2. Designing new AI patterns
Since no such AI components existed in our design system, built new ones from scratch.
Patterns were stress-tested via A/B testing, evolving from UI elements into foundational principles for AI integration.
These principles seeded a broader AI design playbook now applied across other tools.
3. Defining success early
Partnered with product to set clear OKRs:
Increase first-time-right submissions.
Reduce onboarding duration.
Metrics ensured we could track operational impact post-launch.
4. Navigating strategic decisions
Early AI versions offered multiple data suggestions which made users felt confused. We pivoted to one high confidence recommendation.
Engineering wanted an “Apply” button due to backend save issues → I challenged this. Instead, we designed a passive confirmation pattern to preserve flow integrity.
40% reduction in onboarding time.
30% increase in first-time-right submissions.
More importantly, we defined an approach for AI design in regulated environments, turning a feature into a strategic entry point for scaling AI across the bank’s onboarding roadmap.
NDA Protected
This project is under NDA,
please reach out to me at ashitajain666@gmail.com to know the full case-study.
Thanks for stopping by!
I'd ♡ to chat with you :)