Saumya Agarwal

PROBLEM

Customer Due Diligence (CDD) for corporate onboarding was slow and document-heavy, with compliance officers repeating the same judgment calls across near-identical cases.

PROCESS

Mapped the end-to-end CDD workflow with compliance stakeholders to separate genuinely judgment-based steps from repetitive pattern-matching ones that AI could safely assist with.

SOLUTION

Designed agentic UX patterns where AI surfaces a recommendation with its reasoning, and the officer approves, edits or overrides - never a silent automatic decision on a compliance case.

OUTCOME

Now piloting with the CDD team ahead of a phased rollout - early feedback points to less time spent re-verifying cases AI already flagged as low-risk, freeing officer judgment for the cases that actually need it.

PROBLEM

Corporate clients and partner banks had no self-serve way to discover or evaluate available banking APIs - every integration started with a manual sales conversation.

PROCESS

Designed the information architecture for a developer-friendly API catalog: browsable by use case, with documentation-first flows built for a technical audience evaluating integration fit.

SOLUTION

A self-serve marketplace where partners can browse, read docs and request access to APIs without waiting on a relationship manager for the first step.

OUTCOME

Rolling out to early partner banks - replacing a multi-week, relationship-manager-led onboarding with a catalog partners can evaluate on their own timeline.

PROBLEM

Advanced Analytics for corporate banking had fragmented UX patterns across features, built by different teams at different times - hurting adoption and consistency.

PROCESS

Ran a series of journey-mapping workshops with senior stakeholders to align product, design and research on one customer journey, then introduced localization and component-QA processes so new patterns stayed consistent before shipping.

SOLUTION

A shared analytics design language plus a lightweight governance process - new components get tested and localized before production, not after.

OUTCOME

The design language and QA gate shipped org-wide, and the workshop series was recognized internally with an "Appreciated for H1'2024" note from leadership for the cross-team alignment it produced.

PROBLEM

Order cancellations were hurting NMV/GMV at scale, and most support demand was voice-only - expensive to run and slow for sellers waiting on resolution.

PROCESS

Usability testing and research to find where casual, low-intent ordering was driving cancellations, alongside a competitor scan showing peers offering chat, email and self-serve support where Meesho only had voice.

SOLUTION

Meesho Balance nudged more conscious order selection at the point of intent, while a self-serve chat bot deflected the most common support queries away from voice.

OUTCOME

Shipped alongside the self-serve chat bot, giving sellers a faster resolution path than voice-only support and nudging buyers toward more deliberate ordering at the moment of intent - the kind of small intervention that compounds at marketplace scale.