Making proposal drafting faster, smarter, and less painful for KPMG employees.
TIMELINE
January - May 2026
ROLE
Lead product designer
TEAM
SKILLS
Figma, Product Design, Lovable, NoteBookLM, User Research, Cross-Functional Collaboration
SUMMARY
I led the vision for a new way for KPMG employees to write client proposals, using AI to reduce the time it takes to gather sources, draft, and review one.
Design Leadership
Owned the end-to-end design and led the team through design discussions, aligning vision across engineers, PMs, business analysts, and QA.
User Research Facilitation
Led the user testing initiative: wrote the research script, facilitated sessions, and synthesized feedback into next steps.
Accessibility Advocacy
Championed accessibility, guiding the team on screen reader compliance and WCAG-compliant color contrast.
THE PROBLEM
Writing a winning proposal is a slow, painful process.
Millions of KPMG consultants around the world write proposals to win clients. It is how KPMG grows its business, but the process is slow and painful. Pulling together a single proposal can eat up days of work.
How might we help KPMG employees build a strong proposal, backed by up-to-date sources, without the painful process?
INITIAL RESEARCH
To truly understand its flaws, I surveyed around 10 KPMG consultants across practices and experience levels, then used affinity mapping to cluster main ideas. Three pain points stood out:
Finding sources is slow
Tracking down up-to-date case studies and references ate up time before any writing even began.
Collaboration can be difficult to facilitate
Finishing a proposal meant multiple meetings (that were hard to schedule) and a lot of shoulder-tapping to keep things moving.
Writing and reviewing is tedious
Even with templates being used, it still takes time to draft content without it sounding generic or repetitive of other proposals. Getting leadership to review can also be a hassle.
Opportunities identified
Finding sources is slow
Tracking down up-to-date case studies and references ate up time before any writing even began.
Collaboration is messy
Finishing a proposal meant multiple meetings (that were hard to schedule) and a lot of shoulder-tapping to keep things moving.
Writing and reviewing is tedious
Drafting took a while, and using templates was frowned upon, so people often started from scratch.
EARLY CONCEPTS AND ITERATIONS
Initial designs were created and constant iterations occurred after as a result of cross-functional feedback over time.
With tight deadlines for design ideation, I used Lovable for support. I detailed prompts with the stakeholder requirements and iterated with it multiple times until it light a bulb in my head. I then translated rough Lovable concepts to high-fidelity components using the design system.
KEY FEATURES
Upload relevant content for AI to synthesize
Users can upload their client’s request documents and other supporting documents to aid in the proposal generation process.
Proposal section selection & storyboarding
Users can choose what the proposal needs and map out the structure before drafting, giving the proposal a clear shape early.
AI-Powered Content Generation
By analyzing uploaded documents and user formatting preferences, the AI generates tailored proposal content, significantly accelerating the proposal drafting process.
Reviewing Output with AI Agents
Before the proposal is finalized and exported, users can review their content using AI agents to ensure the response is compliant with KPMG guidelines, satisfies the client’s needs, and is of high quality.
USABILITY TESTING
I ran sessions with 6 KPMG professionals across Sales and Advisory. The group spanned across proposal managers to someone who hadn't touched a proposal in a decade, tenure from 5 to 17 years, holding titles from senior managers to the Head of AI for Advisory. I provided testers with loose scenarios that guide them through the flow to see how they instinctively interacted with the product.
What we heard
Finding sources is slow
Tracking down up-to-date case studies and references ate up time before any writing even began.
Collaboration is messy
Finishing a proposal meant multiple meetings (that were hard to schedule) and a lot of shoulder-tapping to keep things moving.
Finding sources is slow
Tracking down up-to-date case studies and references ate up time before any writing even began.
Collaboration is messy
Finishing a proposal meant multiple meetings (that were hard to schedule) and a lot of shoulder-tapping to keep things moving.
NEXT STEPS
Developing V2 based on MVP user feedback
Partnering with engineering to scope de-prioritized features back in
Working with our BA to track bugs in Jira as adoption grows
TAKEAWAYS
Communicate with engineers early and often.
With the team spread across time zones, syncing up was hard at first. Once I took the initiative to find a time that worked and host regular design meetings, iterations and actionable feedback came far more easily.
Bring accessibility in from the start.
Treating accessibility as a first-class concern early on, rather than a final pass, would have saved a lot of rework time.
Ask my design team for feedback early.
Later in the process, I began bringing my work to the design team for review and critique. They surfaced existing interaction patterns used in other KPMG products that could have accelerated my design thinking. From this experience, I learned the value of actively seeking out available resources and tapping into the knowledge of those around me.