Designing the Future of Audit Workflows: From Ambiguity to Agentic AI

Context & Project Overview

TL;DR: I led the design strategy and interaction model for DataSnipper's Agentic AI workflow, taking the product from an ambiguous platform opportunity to a validated AI-native B2B workflow concept. The work reduced setup friction, clarified how users should review and trust AI-generated work, and helped the team shape a feasible MVP with 12 partner firms.

What is DataSnipper? DataSnipper is a platform that helps audit and finance teams work faster by automating repetitive document and data tasks. It started as an Excel-based productivity tool, but today it is evolving into a broader platform that supports end‑to‑end audit workflows. The core idea is simple: reduce manual copying, checking, and cross‑referencing so professionals can focus on analysis and judgment instead of mechanical work.

Why did we start this project? DataSnipper had strong product-market pull, but the core experience was still centered around manual execution. We needed to define the next major platform opportunity: how emerging AI could help audit teams move from individual document tasks to repeatable, end-to-end workflows.

My Role: I started this initiative as Senior Product Designer and was promoted to Staff Product Designer as the scope expanded. Across the project, I led product design strategy, research synthesis, workflow systems design, interaction design, prototyping, and partner validation in close collaboration with PM, Engineering, and customer-facing teams.

Timeline: Feb 2025 – Present

What this case demonstrates

• Turning strategic ambiguity into a concrete AI-native product direction
• Designing AI behavior around user intent, system constraints, human review, and traceable output
• Translating complex B2B workflows into clear interaction patterns across roles and expertise levels
• Using partner validation to make scope trade-offs between desirability, feasibility, trust, and launch readiness

Initiative Impact

As the project evolved, it delivered four outcomes that pushed the DataSnipper platform forward:

• Immediate business signal: Three design partners proactively asked about pricing before launch, with one customer requesting to expand from 40 to 1,000 seats.
• Validated desirability: Repeated prototype cycles confirmed strong appetite for streamlined workflows and revealed the setup friction that would block adoption.
• AI-native interaction model: The team moved from manual workflow creation toward an intent-led agent experience where users could state goals, review plans, and inspect traceable outputs.
• Real-world MVP scope: The Design Partner Program helped the team define a first release grounded in real audit workflows, not theoretical AI demos.