Ashley Kim · Design Lead · J.P. Morgan Global Private Bank Design Studio

The team had one stubborn problem.

Can a private bank verify who is calling in seconds, without making the client feel interrogated?

Nine months, three applications merged into one platform, and 837 client service users later: authentication success rose 88%, and secure-method usage rose 357%. Live in production since 2023.

The goal in one line: raise authentication success. Users were not failing because of risk rules. They were failing because the application made authenticating hard.

Client Service Platform: case dashboard with incoming call and authentication panel

+88%

authentication success

+357%

secure-method usage

-38%

reliance on less secure methods

837

client service users

Try the prototype See the numbers

The numbers The problem The journey The decisions Scale Reflection

RoleDesign lead, and the only designer. Research, IA, interaction patterns, rollout, and the learning program.

TeamMe, product owners, and tech. The platform strategy was set jointly by design, product, and tech.

WhenNovember 2022 to July 2023 build · Houston pilot October 2023 · all locations January 2024 · internal product

StatusLaunched 2023 for U.S. Private Bank · International Private Bank planned 2027

Impact

Every metric below includes how it was measured.

+88%authentication success rate

Measured: authentication metrics before launch, compared against the same metrics one year after launch, tracked on the performance dashboard we shipped as part of the platform.

+357%

secure-method usage

Measured: share of authentications using secure methods (Voice ID, mobile app), before launch versus one year after.

-38%

reliance on less secure methods

Measured: drop in fallbacks to security questions and weaker checks, before launch versus one year after.

837

client service users

Measured: the full U.S. Private Bank client service population on the platform after rollout.

This was never just a UX project for the bank. Every authentication that moves to Voice ID is fraud risk reduced, a compliance posture strengthened, and a client who was not interrogated. That is why it was funded, and why the dashboard reports the numbers to leadership.

01 · The problem

Three tools, one phone call, and a client waiting on the line.

Advisors juggled separate applications while a client waited, and authentication was the step everyone dreaded. The insight that ended up steering everything came later: security behavior follows interface behavior. People do not pick the safe method. They pick the near one.

What advisors were living with

1Authentication success was low, and the flow created a poor experience on both ends of the call.

2Caller identification was unclear. Clients can hold multiple accounts, and authorized related parties can initiate calls too.

3Client preferences were not centralized or easily accessible, so personalization depended on memory.

4Client details were disconnected, scattered across multiple applications, so the advisor swiveled between screens mid-call.

What client service actually prioritizes

QualityThe quality of the client conversation itself.

SpeedDelivering the client's request on time, during the call.

TrustAuthenticating the client during the call, without it feeling like an interrogation. This is where the design had to win.

02 · The journey

Nine months, from swivel-chair chaos to one platform.

Nov 2022 · the starting state

Three separate applications for one phone call. Advisors authenticated clients with the slowest, least secure methods because those were the ones at hand.

The strategy

Design, product, and tech set the sequence together: build the client information view first, inside the case app, and bring the phone and authentication in last. Prove the space is useful before merging the riskiest piece into it.

Define and ideate

User interviews on one question: what do you actually need on screen during a call? Personas, data analysis, priority against effort, and low-fi testing killed weak directions before they got expensive. The population turned out to be six coverage types across six US locations, and a tenured-versus-new learning gap. One-size-fits-all was off the table.

The pushback

Stakeholders pushed back on authentication sitting at the top of the client view. I believed the call starts with trust, so the trust step belongs where the eye lands first.

The test settled it

User testing settled it: authentication stayed on top. And the design went further: after the first method succeeds, the view changes state, with a button to add another method, and we tested exactly how advisors would layer a second authentication after the first.

Develop

Design iteration with QA and pilot testing. The counter-clockwise pattern around the merged screen survived because it matched the real order of a call, tested as intuitive, and built muscle memory fast for both tenured and new advisors.

Launch · 2023

Launch was designed like a product too: a user learning program, onsite training, a town hall, and a performance dashboard so the impact numbers above had a source from day one.

Next · 2027

International Private Bank rollout is planned for 2027, carrying the scalability work below.

03 · Research

Question, method, insight, and what changed in the design.

Research at a glance

The sequence: a survey across the 837-person client service population on how calls actually go and where they break, interviews built on the survey analysis, a 12-person brainstorm with product, engineering, and executive directors, impact-versus-effort prioritization into a two-phase roadmap, a low-fi prototype to make the project concrete, user testing at every phase, and a pilot before launch. Six coverage types, six US locations, tenured and new advisors.

Who actually uses this, and how differently?

Method: user interviews, personas, and behavioral data analysis

Six coverage types with distinct behaviors, plus a tenured-versus-new learning gap.

→ One platform, designed to flex by group instead of averaging them

Which authentication flow will advisors really use mid-call?

Method: low-fi user testing, then A/B testing in development

The authentication screen kept getting pushed behind other screens, like the ones for finding client info.

→ Merge authentication into the service desktop itself, so the secure path is the near path

Why do secure methods go unused?

Method: authentication metrics analysis

Going in, we assumed advisors avoided Voice ID because they did not trust it. The metrics said the opposite. They trusted it fine. It just lived in another app, and mid-call, another app might as well be another building. The problem was never resistance. It was friction.

→ That flip changed the whole direction: put the secure options one click from the incoming call

04 · Major decisions

Three calls that made the numbers move.

DECISION 1

Build the client view first. Bring authentication in last.

Alternatives

Merge everything at once, or lead with the flashiest piece, the phone. We chose neither. The sequence was strategized jointly by design, product, and tech.

Why

Advisors had to find the client information space useful on its own before the riskiest step moved into it.

The price

With limited time and tech capacity, v1 had to cut: the after-call view of call and email history, authenticating additional callers, and editing client information and system preferences all waited for v2.

Outcome

By the time the phone merged in, the space around it had already earned trust, and advisors said so unprompted.

DECISION 2

Authentication lives at the top of the client view.

The pushback

The top position was my call, and stakeholders pushed back on it. There was real risk in holding the line: if testing had gone the other way, I had spent my credibility on a layout. I brought the test instead of a better argument, and the test decided it.

The detail

After the first successful method, the panel changes state and offers a button to add another. We tested exactly how advisors layer a second authentication on top of the first.

Outcome

"I love how the authentication option changes depending on the client's preferences." Trust became the first thing on screen, and the numbers followed.

DECISION 3

The counter-clockwise pattern.

Why

The counter-clockwise pattern was mine, and it won for three reasons, in order of weight: it matches the real order of a call (phone rings, identify, authenticate, act), users tested it as intuitive, and it builds muscle memory fast, which mattered with a tenured-versus-new learning gap. Teams designing later screens now inherit the pattern instead of inventing one.

Outcome

One flow that six coverage types could share without six trainings.

SMALLER CALLS

Identification filters, and preferences inside the call.

Why

Recent-match filters make the likely client appear the moment the call comes in, and the client's preferred method sits on the call panel with an audit line for who changed what. Personalization at zero clicks.

Outcome

"I mostly use the recent matches."

The two screens those decisions live on

Authentication panel on top of the client view

The top position, held: Voice ID and mobile authentication above the client's details, changing state after the first success.

Caller identification panel with recent and caller matches

Identification at ring time: most selected, recent selected, and caller matches, with a caution banner when carrier porting is reported.

07 · Reflection

What this project proved.

1Security is a UX problem packaged as compliance. I spent a while thinking we had a trust problem, and what we actually had was a distance problem. Nobody chose the insecure method because they liked it. It was simply closer. We moved the secure path nearer, and the numbers moved.

2Launch is part of the design. The learning program, the training, and the dashboard did as much for adoption as any screen — the jump from 42% to 79% was this rollout design at work, not the screen alone.

3A population is not a user. Six coverage types, six US locations, two experience levels. Averages design for nobody.

4Bring the test, not the argument. The debate over authentication’s top position ended the day users settled it.

08 · What came next

The next step we wrote down was AI helping advisors during calls.

That question would not leave me alone. It became my own project: an AI operating model where the human keeps control at two gates.

Read SPOT OM, the AI operating system case study →

If you are designing systems people have to trust with their money or their name, that is the work I want to do.

Ask me about any decision in this platform. I can walk you through the why in twenty minutes.

Ashley Kim · ashleykim1229@gmail.com

Can a private bank verify who is calling in seconds, without making the client feel interrogated?

Every metric below includes how it was measured.

Three tools, one phone call, and a client waiting on the line.

Nine months, from swivel-chair chaos to one platform.

Question, method, insight, and what changed in the design.

Three calls that made the numbers move.

Build the client view first. Bring authentication in last.

Authentication lives at the top of the client view.

The counter-clockwise pattern.

Identification filters, and preferences inside the call.

Designed for 837 users, built to survive the next 8,000.

What users said, in their words.

What this project proved.

The next step we wrote down was AI helping advisors during calls.