Plum needed an AI assistant that could guide people through financial decisions — personally, at scale, within regulatory constraints. I led the end-to-end design, from strategy to shipped product.
Sitting on cash they could invest — a big underserved market with clear appetite for guidance.
In low-return cash — eroded by inflation and pulled out of UK capital markets.
“Lack of knowledge” is the #1 barrier to investing — yet our survey showed clear appetite for guidance.
Can an AI conversational layer turn that appetite into action?
Sources: Barclays Prosper (2024) · Social Market Foundation (2024) · FCA Cash Savings Review (2023)
We approached compliance as a design principle, not a limitation — embedding it into the architecture to unlock the rest of the solution.
We started with both market research and a clear view of our regulatory constraints, using each to inform and balance the other. This led to a financial guidance decision tree, designed to route conversations using clear, rule-based thresholds.
Stay within FCA boundaries while ensuring guidance remains clear and actionable.
Help users understand their options and make informed decisions themselves.
Adapt responses based on user input to reduce friction and cognitive load.
Deliver tailored guidance consistently using structured pathways.
Contextual cards that surface relevant goals and products within the conversation — making guidance tangible and immediately actionable.
Structured option sets that guide the conversation forward — keeping users in control while ensuring responses stay within compliant pathways.
The landing experience blends both patterns — topic cards for discovery, suggested questions for quick entry — meeting users wherever their intent sits.
Structured cards for surfacing product information and recommendations — scannable, tappable, and easy to act on without reading long text responses.
Conversational responses with clear typographic constraints — 32 characters per line max, left-aligned when wrapping — optimised for readability at speed.
Pre-defined question flows for collecting financial context through structured, compliant interactions — reducing ambiguity while keeping the experience conversational.
30 participants · February 2025
Discovery
58%
Could locate the assistant entry point in 5 seconds
30 participants · February 2025
Intent
78%
Could articulate a real question they’d actually ask
Nearly doubled year-on-year in 2024 (Public · CBInsights)
AUM tripled year-on-year, crossing the billion mark (Public · Press)
Year-on-year growth in average revenue per customer (Public · Press)
In faulty attribution signals — enabling reliable, data-driven product decisions (Internal)
Led end-to-end design across iOS and Android, partnered with Product and Data Science on AI messaging, and ran value-first guidance workshops — shipping v1 while exploring conversational AI as the primary interface.
The real craft was in the conversation architecture — the decision tree, the regulatory guardrails, the guidance-vs-advice boundary with legal. The UI was straightforward once those decisions were made.
Always open to conversations about product, design, and leadership.
Get in touch
