Vasudha Mamtani — Portfolio

Portfolio Presentation

Vasudha Mamtani

Product Designer

Here's some of my work I'd love to walk you through.

About Me

A decade of craft. IC depth, strategic reach.

2015

Graduated

Computer Engineering

2015

Deloitte Digital

Learned to design
for everyone

2020

Amazon

Learned to design
for scale

2022

Spotify

Learned to design
for delight

2024

Sentry

Learned to design
for complexity

2025

Oracle

Learned to design
for enterprise

About Me

Cat parent · Doom scroller · Budding runner.

My cat >>>> Most people

Fantasy novels are my new jam

In my hiking and running era

UX Bootcamp Instructor
+ Amateur Design Writer

GUIDED QUOTING

Unifying fragmented legacy workflows into a single agentic experience

Oracle CPQ — Configure, Price, Quote

Role Lead Product Designer

Platform Enterprise B2B · SaaS · Web

Type Agentic AI · Legacy redesign

Oracle · Context — Where CPQ sits

What is CPQ?

Configure, Price, Quote.
The system Sales Representatives use to turn large, complex customer orders into accurately priced quotes.

The cycle

A Sales Rep receives a large order from a customer. They use CPQ to configure the right products and price them to meet both business and customer goals — then generate a formal Quote.

Oracle · How it started

The product

A legacy CPQ tool,
acquired by Oracle.

01

Design and tech debt

Built before Oracle ownership — the UX carried years of accumulated decisions nobody had gone back to fix.

02

Fully manual workflows

No automation, no intelligence — just a rep and hundreds of product lines.

03

Workarounds as muscle memory

Sales Reps had been working around its problems for years — manually, slowly, with no real alternative.

The moment

Oracle announces
an AI-first pivot.

"Every Oracle product will be redesigned around AI capabilities."

— Leadership directive

My read

This was the perfect opportunity.
CPQ's workflows were manual, and the appetite for change was finally there.
Paired up with my Product Manager to start drafting the AI-forward vision for CPQ to pitch to the leadership.

Oracle · The current workflow

A day in the life of a Sales Rep

Oracle · What data reflected

3 days

to complete a quote

Creation to submission — before the approval cycle even starts.

ONLY 1 in 6

quotes became an order

The rest were abandoned — started, never finished, or never sent.

20%

INTERNAL

/

35%

CUSTOMER

rejected or sent back

Quotes returned for rework — before they could move forward.

Customer research · What Reps told us

"I go through hundreds of products line by line — one wrong selection and I have to start over."

"Configuring products is the longest part. By the time I'm done, I've spent most of my day just on that."

"Pricing approvals are unpredictable. I never know if I'm in the right ballpark until it gets rejected."

— Sales Rep

          + Used Oracle's AI-powered Content Library to study past research documents, customer interview snippets & market data.
          

Oracle · The Strategy

Two audiences. One coherent strategy.

What users needed

Time to quote completion — reduce it

Automation of manual effort — translating documents and configuring products

Reduce rejections — with high confidence quotes

What leadership needed

Make a splash at Oracle AI World — demo natural language quoting live

Targeted client demos showcasing upcoming AI capabilities with real data

Planning · Product + Design

Oracle AI World

Natural language
quoting demo

Real product data, zero engineering cost

Client Demo 1

Medical industry
tailored quoting

Client-specific products and pricing scenarios

Built with Figma Make

Client Demo 2

Industrial sector
tailored quoting

Upcoming capabilities, real client data

Built with Figma Make

Product Launch

The real thing
ships

↳ Document processing

↳ Pricing Agent

↳ Score + Win Probability

Oracle · Using AI — Figma Make

What I tried

For Ideation

Goal

Ideate the NL input use-case using the Agent Builder DS

Result

Terribly bland ideation · Couldn't parse the DS · Burned tokens too fast

Instead

Built the concept screens myself

expand ↗

Make output

Wrong tool for this job

What I used it for

For Scaling

How

Generated filler screens · Built a fully functionational prototype

Friction

Components needed tweaking to map real data.
Some intervention needed — but the result was a configurable prototype.

expand ↗

Broken → Final

Multiplied the impact

Taught PMs to build client demo prototypes in Make

        Given no constraints — preferred stack would have been Claude Code + Figma MCP + Document connector
      

Oracle · Goals & Guardrails

Reduce time to completion

From requirements to a ready-to-send quote, significantly faster.

Workflows that instill confidence

Surface the appropriate information and meet users where they are.

Minimal HitL intervention

Aim for a very low error rate when Agent adds products to build user confidence.

Prioritising document-driven quoting

Moving away from leadership's directive — focusing on solving real user problems first.

Limitations

"We could experiment with the workflow — but not with the patterns."

Oracle has strict guidelines around designing experiences.
We had room to reimagine the workflow and information architecture, but were bound by Oracle's Agent Builder design system for interaction patterns, components, and visual language.

Oracle · Where teams were at

Strategizing with Engineering

Requirements

email / call

Select Products

add to quote

Configure

products

Set Pricing

discounts

Manager Review

approval needed

Send Quote

to customer

Product Extraction from Documents

Reads requirement docs and extracts relevant products automatically.

Smart Configuration Agent

Helps reps configure the right products for a quote.

Oracle · The vision

Unify

One conversation.
Start to quote.

Bring the two siloed agents into a single, coherent experience the Rep never has to stitch together.

Extend

Add a
Pricing Agent.

Address the biggest remaining pain — pricing was a shot in the dark. The agent brings clarity.

Signal

Confidence,
not guesswork.

Adding Quote Insights. Surface approval likelihood and pricing confidence inline — so the Rep always knows where they stand before submitting.

Oracle · The work

1.1

1.2

1.3

Discovery · Entry Points

Goal was to nudge feature discovery across the product lifecycle, without restricting access to the original flow

Pages targeted were — Existing Quote page, All Quotes page, Within the Sales App, many more...

expand ↗

Oracle · Beyond Design

With Engineering

Confidence Score

Drafted the concept of a confidence score for extracted products. Documented logic for automation vs HitL using score.

Trust-building

Designed patterns so reps could verify agent decisions progressively — building confidence in the system over time.

Stress Testing the Agent

Generated industry-specific documents to QA capabilities — product spec PDFs, requirement sheets, hand-written notes, natural language.

↗

↗

↗

With Partner Teams

With Sales

Proposals to build data gathering workflows to incorporate Sales needs and build end-to-end quotes from the get-go.

With Agent Builder Design System

Added Document Processing workflows to make the HITL pattern universal across all agentic features on Oracle.

With Orchestrate

Created contextual entry-points into the Guided Quoting workflow from sister applications.

Oracle · Impact

An Agentic workflow that instilled trust and confidence

12%

Customer rejection rate — down from 35%

< 1 day

Time to completion — including internal approval

65%

Return rate — users came back within 30 days of first use

Retrospective

What I learnt · I

Prototypes over engineering demos

Figma Make let us build client-ready prototypes without touching the codebase.

What I learnt · II

Rollout Strategies

Learned how PMs select clients for Beta and manage rollout risk.

What I'd do differently

Cross-team collaboration from day one

Customers use a suite of products — sharing data across systems earlier would have built better solutions.

That's all for this one.

Oracle · Guided Quoting

Sentry ›

Sentry

Dashboards Revamp

Late-stage startup · Designing for engineers

Sentry · Dashboards Revamp

Research

Planning & Prioritisation

Re-design + User Testing

Small-win Features

Sentry · Dashboards Revamp

Before & after

Before

12k DAU — down 40% since 2021

Only 1 in 6 dashboards actually used

18.5% week-4 retention — lowest in product

50+ open GitHub tickets from engineers

Broken widgets, 4 templates, no catalogue

After

↑ 12% daily active users in first 30 days

42% of Business orgs favourited a dashboard

> 60% of product pages updated for consistency

> 30 GitHub tickets resolved

8 new components in Design Library