Enter password to view case study

QBank

Internal tool

AI workflow

2025

Turning an overloaded inbox into a system that remembers

Taking Investor Relations' fund Q&A from an unsearchable inbox to a shared knowledge system over ~6 months, from discovery through launch and post-launch iteration.

QBank

Internal tool

AI workflow

2025

Turning an overloaded inbox into a system that remembers

Taking Investor Relations' fund Q&A from an unsearchable inbox to a shared knowledge system over ~6 months, from discovery through launch and post-launch iteration.

QBank

Internal tool

AI workflow

2025

Turning an overloaded inbox into a system that remembers

Taking Investor Relations' fund Q&A from an unsearchable inbox to a shared knowledge system over ~6 months, from discovery through launch and post-launch iteration.

cms image

My role

Lead designer

Timeline

~6 months

Team

Peer designers, 1 PM 1 Eng pod (2 FE / 4 BE)

Scope

0→1, concept to ship, user research

My role

Lead designer

Timeline

~6 months

Team

Peer designers, 1 PM 1 Eng pod (2 FE / 4 BE)

Scope

0→1, concept to ship, user research

Challenge

A single fund question passed through three people before it was answered, and nothing was kept for next time. IR spent 5,000 hours a year re-researching answers that already existed but were trapped in individual inboxes no one could search.

Strategy

QBank auto-captures every thread and surfaces precedent, but the client-facing answer stays written and owned by a person.

Result

40% of advisor questions now self-serve without reaching IR, reclaiming 58% of the time once lost to manual search, across 4,000+ queries a year.

01

Background

Players in a fund Q&A flow

In alternative investments, a single fund question passes through three roles: an Investor asks it, an Advisor who owns the client relationship fields it, and an Investor Relations (IR) analyst researches the answer. This project lives between the two who do the work:

Players in a fund Q&A flow

In alternative investments, a single fund question passes through three roles: an Investor asks it, an Advisor who owns the client relationship fields it, and an Investor Relations (IR) analyst researches the answer. This project lives between the two who do the work:

High Net-worth Client

Ask the question

The end client reviewing a fund report. Wants a clear, fast answer

Financial Advisor

Owns the relationship

Fields the question, escalates the hard ones to IR

Investor Relations

Holds the exprertise

Fund expert. Researches, drafts, and supplies the answer

Fund Q&A today

One question moves through several hands before it's answered, and nothing is kept for the next time.

Fund Q&A today

One question moves through several hands before it's answered, and nothing is kept for the next time.

Step 1

Question reaches IR

An advisor escalates a client's question by email. An inbox manager finds the right owner in a coverage tracker and assigns it.

Step 2

IR searches for precedent

The owner looks for a past answer, but Outlook search can't surface it, so research restarts from scratch.

Step 3

IR drafts the response

The reply is written by hand with no prior wording to reuse; quality depends on who answers.

Step 4

Logged but never saved

The query is counted in a spreadsheet. The answer is stored nowhere reusable, so the next identical one repeats the cycle.

02

Research

What we heard from the users

What we heard from the users

12+

Interviews

With the IR team and advisors, mapping the real Q&A workflow end to end.

2 cities

New York & Mumbai

Full-day shadowing on both sides of the workflow, across the time zone.

Marcus Chen

Private bank advisor

Manhattan, New York

EST

"My client wants an answer today. I shouldn't have to wait overnight for IR just to explain a number on a report."

Goal

Give clients confident, same-day answers without becoming the fund expert.

Frustration

Simple questions bounce to IR and come back hours later, across a time zone.

Need

To self-serve answers that already exist, instead of escalating everything.

5,000 hours a year spent answering 4-5K advisor questions over email

40% of the Advisor questions could have been self-served

Alex Menon

Investor Relations specialist (IR)

Mumbai, India

IST

"I know we've answered this before. I just can't find where, so I research it from scratch."

Goal

Answer accurately and reuse past work instead of redoing it.

Frustration

Weak search and no tagging mean precedent is effectively lost.

5,000 hours a year spent answering 4-5K advisor questions over email

30%

Searching past Q&A

25%

Drafting response

20%

Managing in mailbox

10%

Searching fund docs

15%

Reaching out to partners

Need

Searchable memory of how the team has answered, with context attached.

The overnight gap in every Q&A

Advisor and IR sit 9.5 hours apart. Their days barely overlap, so questions get tossed over email and answered the next morning.

The overnight gap in every Q&A

Advisor and IR sit 9.5 hours apart. Their days barely overlap, so questions get tossed over email and answered the next morning.

03

Solution

A shared memory of how we answer, not just what the answer is.

QBank captures every communication, surfaces past precedent, and helps IR draft a response faster

A shared memory of how we answer, not just what the answer is.

QBank captures every communication, surfaces past precedent, and helps IR draft a response faster

Designed for

AI structures the thread and surfaces past precedent so IR can see how the team has answered before. IR reads, judges, and writes the reply.

Designed against

Letting AI write the client-facing answer on its own. In high-consequence finance, the response that reaches the client has to be owned by an advisor.

Step 01

Capture at the source

Capture the email into a thread automatically

An advisor asks over email as always. Rather than a tool the team has to remember to fill in, I made capture happen at the source. The moment a message arrives it becomes a structured thread. I chose invisible capture over a cleaner manual form because the memory only builds if every question reaches it.

Step 02

Capture at the source

Every communication lands in one searchable place

Instead of a shared mailbox no one can search, every thread is collected and auto-tagged by fund, client, and owner. IR opens a repository instead of an inbox. It is the first place the team's answers can be found again.

Step 03

Read the precedent

The thread restructured into a Q&A summary

IR sees an AI summary up top and the email chain rebuilt into a clean Q&A panel with key figures and attachments pulled out. The AI structures so IR can judge fast but never writes the client-facing answer. In high-consequence finance that judgment stays with a person.

Step 04

Ask Coach

Coach searches the team's memory with sources attached

For anything complex, IR asks Coach in plain language. It searches the whole repository, fund docs, and partner FAQs. Every answer links back to its source so precedent is traceable instead of taken on faith.

Step 05

Human-owned response

Reply in your own words with the precedent in hand

IR drafts the reply, drops in the precedent they lifted, and closes the case. QBank did the finding and structuring but the answer is written and owned by a person. The one-off email is now team memory for next time.

04

Impact

Every answer becomes a reusable asset instead of a one-off email.

Institutional knowledge compounds instead of leaking.

Every answer becomes a reusable asset instead of a one-off email.

Institutional knowledge compounds instead of leaking.

0%

Time saved

Reclaimed from manual mailbox and document searches

0%

Self-serviced

Advisor questions that no longer need IR

0+

Queries / year

Annual Q&A volume QBank now streamlines

Advisor

Before · escalated almost everything to IR, then waited across a time zone.

Now · self-serves 40% of questions, with no overnight lag.

IR Specialist

Before · re-researched existing answers, a third of the week on Q&A.

Now · reuses the team's memory and reclaims 58% of that time.

Copyright © 2026 Ting Fang Cheng

Copyright © 2026 Ting Fang Cheng

Copyright © 2026 Ting Fang Cheng