JP Morgan Private Bank: Onboarding

Planning and estimation

Generative research: user interviews

Competitor review

AI vibe coding

Strategy/Architecture

Service Design

Interaction design

Evaluative research: user testing

Developer handoff and accessibility

An international onboarding and KYC redesign that reduced onboarding from 30 days to 4.



I unified fragmented systems, embedded AI patterns, and reduced operational risk while navigating cross-programme coordination.

Planning and estimation

The Product Owner needed a plan to roll out the new app. Although it was intended as a multi-year project, leadership wanted visible impact within a few months.

I drew up this plan to analyse the existing functionality, cross-reference it with the preliminary designs, and identify features we could release early without disrupting existing user journeys or data structures.

This incremental approach would demonstrate progress quickly while reducing risk by delivering working code as soon as possible.

Generative research

We had access to a pool of bankers and operations staff who had volunteered to help with the redesign. We acknowledged that not using random sampling would skew our findings, and proceeded because it allowed us to iterate rapidly.

In collaboration with researchers, I developed research plans, and conducted interviews and panel discussions to validate assumptions and challenge our brief.

I was at odds with the researchers about using panels due to the HiPPO effect, but ultimately the panels:

• increased the volume of data
• increased buy-in and decreased the risk of rejection by involving more Advisors

Later we would create an affinity map/clusters and used the needs and motivations we discovered to prioritise design features.

Competitor and predecessor review

Fenergo is third party onboarding and KYC solution. I assessed its features and journey and found much to like:

• Highly visual experience: progress meters, tree hierarchies and colour coding abound, in contrast to the austere aesthetic of the existing system.
• Assigning document properties at the point of upload prevents duplication and access permission problems later.
• Architecture: 52 tasks were split into 14 categories. I felt that 14 top level categories was too many for a user to easily absorb, so I made a note to test it later.

The existing COB (Client OnBoarding) system was generally poor, but had some interesting takeaways:

• By virtue or being built over many years by many teams, there were some accidental gaps and overlaps which created "power user" features. We noted that these abilities should be carried through to the new design.
• The sprawling design meant that training new users was lengthy and frustrating. We determined that intelligibility should be a metric we tested our new design upon.

AI vibe coding/generative designs

I used Figma Make and Lovable outside the firewall to discover any blind spots, and to generate new ideas.

An AI generated specification was used to protect IP, and after a few iterations both Figma Make and Lovable made a reasonable generic design.

Their designs raised some interesting questions:

• What is the best way to express risk level?
• How deep vs. wide a navigation is optimal for our audiences?
• Given that first-time failure is is more common than success, should our "happy path" be ready to accomodate frequent setbacks? 
  

Neither app was able to handle complex branching logic or subtle variations in language such as may/could/should/must, that we collaborated with our KYC SME to develop.

Architecture through Object Oriented UX (OOUX)

Years of un-coordinated growth created a Frankenstein's monster of bolted-together apps and architecture that made the lives of staff and clients painful.

I led an exercise to create an Object Oriented UX architecture: a collection of objects, actions on objects, and relationships between objects that I validated and iterated with SME's.

This radically simplified data structures and paved the way to slim down the amount of UI required. The greatest achievement was the sigh of relief when staff looked at it and said "this makes sense".

A detailed view can be seen here.

Service Design

My team and I worked as the connective layer between the U.S. and International lines of business, which operated on disparate technology stacks, workflows, and expectations.

The role required unifying processes across these environments while shifting responsibility from Operations teams to Advisors (many of whom saw the work as outside their remit).

The project demanded not only technical coordination but careful navigation of organisational politics to secure buy-in and deliver a sustainable, shared way of working.

View a high resolution version of the Onboarding Service Blueprint.

Interaction design

I always start with sketching to shake out issues quickly and maximise creative freedom.

Progressing to digital, I've found that mid-fi doesn't save much time given how easy design systems are to use. Client expectations are also now more mature: they know that a crisp looking static design doesn't mean it's ready to deploy.

In this project, seven designers needed to create novel patterns quickly, which caused duplication across common components. We solved this by:

• Discovering who needed a new component most urgently and assigning the work to them.
• Agreeing as a group what functionality each component needed to have.
• Creating a new design system to pull from.

Evaluative research

Interactive prototypes and written propositions were put before user testing subjects to gather feedback.

As onboarding is a multi-step, multi-role process, we weren't able to gather any end to end metrics. We did experiment with other metrics:

• satisfaction scores out of five required more samples to be meaningful than the circa 8 we took at each round of testing.
• Net Promoter Score (NPS) was suggested to be able to compare with industry averages, but didn't get traction with senior leaders.

A point of failure was communication between the product team, design triad (experience design, content and research) and users: sometimes features and propositions were tested that were outdated, irrelevant or known to be unwanted.

Developer handoff and accessibility

We developed the new design over dozens of sprints; we were cooperating closely with developers in daily stand-ups.

When a specification document's explanation of conditional logic was too complex , I made visuals to aid understanding.

All Figma work was associated with a JIRA ticket. This additional overhead ultimately saved UX time by ensuring all work was budgeted and recognised.

I specified Greenlines for accessibility:

• Landmarks
• Headings
• Alt text
• Tab order

We slashed average onboarding waiting from 30 days down to 4 through a new service design.

Our new visual identity and interaction design transformed client and internal user experiences.