Let's talk

Designing an AI-Assisted Workflow for Enterprise Users

A case study illustrating my process, decisions, and impact.

Low-fidelity wireframes showing the AI workflow redesign

Project Snapshot: I redesigned a workflow tool to help enterprise teams quickly generate AI-assisted insights, improving task completion time and user confidence.

Snapshot

The "too long, didn't read"

Company

bp plc

Role

Experience Designer

Timeline

Jan – Apr 2025

Team

PM, 2 Engineers, Data Scientist

Impact

Validated with 8 users

Tools

Figma, FigJam

Methods

Interviews, Prototyping, Usability Testing

Domain

AI-assisted enterprise workflows

The Problem

Context.

Lack of visibility created mistrust

Enterprise users needed a way to use AI confidently within existing workflows. Initial testing showed users were unsure how AI results were generated and how reliable they were.

Low-fidelity wireframes showing the AI workflow redesign

Manual review slowed productivity

Enterprise users needed a way to use AI confidently within existing workflows. Initial testing showed users were unsure how AI results were generated and how reliable they were.

Low-fidelity wireframes showing the AI workflow redesign

Teams needed accountability & auditability

Enterprise users needed a way to use AI confidently within existing workflows. Initial testing showed users were unsure how AI results were generated and how reliable they were.

Low-fidelity wireframes showing the AI workflow redesign

Goals & Success Measures

{Insert subtitle here}

  • Reduce time to complete a task by 20%
  • Increase user confidence with AI outcomes
  • Ensure workflows met compliance needs

My approach

{Insert subtitle here}

User Research

I conducted interviews with 8 enterprise users and mapped common workflows to identify friction points and trust gaps.

Systems Thinking

I analyzed dependencies across policy, data, and UI touchpoints to ensure consistency across the ecosystem.

Iterative Prototyping

I tested 3 design iterations to validate clarity, speed, and usability.

Low-fidelity wireframes showing the AI workflow redesign
Early wireframes tested with users to identify clarity issues.

The solution

{Insert subtitle here}

The final design made AI actions transparent and traceable, with clear feedback and contextual guidance.

  • Clear status & feedback messages
  • Visible audit trails
  • Inline guidance to set expectations

Impact & outcomes

What feedback said

While not all metrics were quantifiable at this stage, qualitative feedback clearly showed improved clarity and efficiency.

  • 🕒 Estimated 25% faster task completion
  • 👍 6/8 users reported improved trust
  • 🔍 Compliance reviewers felt more confident signing off
  • 🕒 Estimated 25% faster task completion
  • 👍 6/8 users reported improved trust
  • 🔍 Compliance reviewers felt more confident signing off
  • 🕒 Estimated 25% faster task completion
  • 👍 6/8 users reported improved trust
  • 🔍 Compliance reviewers felt more confident signing off
  • 🕒 Estimated 25% faster task completion
  • 👍 6/8 users reported improved trust
  • 🔍 Compliance reviewers felt more confident signing off

Reflection

What I learnt and key takeaways

Designing for AI requires trust, clarity, and accountability. Users value transparency even more than speed — and involving them early reduces risk later.

Positives

Negatives