Strategy · AI Design · IBM
Generating Insights from Customer Reviews
Strategy · AI Design · IBM × British Financial Institution
Timeline
2 months · 2026
Role
Team
IBM Strategy, Design & Build teams
Tools
IBM watsonx Orchestrate · IBM watson BI · Figma · Mural · Adobe AE · Adobe PR
01The Brief
How might we...
How can banks proactively act on customer feedback while keeping a real-time view of competitor innovation?
02Context
A team drowning in a process of their own making
A major UK bank's Channel Insights team was drowning in a process of their own making — 30 minutes every day manually copying app store reviews into spreadsheets, two hours to answer a single insight request, and no visibility of what competitors were doing. They had the data. They had the team. What they lacked was the infrastructure to use either at pace. With 75% of UK adults using mobile banking and 49% having opened a digital-only bank account, app store reviews are one of the most direct signals a bank has into how its product is performing — and how competitors are moving. The client had 30,000 of them sitting unread.
03Insights Discovered
Three compounding problems
Three compounding problems were consuming the team's capacity before any strategic work could begin. Manual processes meant 30 minutes a day to collate reviews and two hours per insight request — leaving no room for the recommendations the product team actually needed. Decisions were made on internal data alone, with no systematic way to benchmark the client's app against Barclays, Monzo, or any other market player. And with the product team entirely dependent on the insights team to access data, the culture had become one of reactive firefighting rather than proactive roadmap input.
04Process
Discovery to delivery in two months
I was brought in as Lead Designer, Facilitator, and Project Lead to scope and shape an AI-powered solution from first client conversation to signed deal — working across IBM strategy, design, and build teams to take the opportunity from discovery to delivery in two months.
Discovery & Scoping
Mapped the current state. Surfaced pain points, opportunities, and the central tension between data availability and usability. From there we defined the pilot statement, value drivers, and measures of success — aligning on a visualised self-serve dashboard as the core solution concept.
Solution Design
Worked with IBM build teams to translate the scoped solution into a deliverable architecture — automated collection, sentiment analysis, thematic categorisation, natural language querying.
Use Case Video
Produced and directed the use case video — translating a complex AI architecture into a human story for non-technical stakeholders. The primary asset for internal sign-off.
Stakeholder Presentation
Presented the Business Value Assessment to senior client stakeholders. The engagement secured sign-off and progressed to a signed deal within the client's Enterprise Licence Agreement.
05Solution
Always-on review intelligence
An always-on review intelligence platform built on IBM watsonx Orchestrate. Any member of the product or insights team can ask a plain English question — what are customers saying about the new card feature? how do our reviews compare to Monzo this month? — and receive an instant, structured, actionable response. No analyst required. No spreadsheet. No wait.
The Product
A self-serve dashboard giving the Channel Insights team real-time access to categorised, sentiment-analysed app store reviews — queryable in plain English without analyst involvement.
How It Works
Automated collection feeds into sentiment analysis and thematic categorisation, stored in a structured database. A natural language interface sits on top — any question, instant structured response.
06Outcome
Scoped in two workshops. Delivered in two months. Built to last.
The pilot secured full stakeholder sign-off and a signed deal within the client's Enterprise Licence Agreement for IBM technology. The engagement unlocked further AI work with new financial services clients. What started as a 30-minute daily manual task became a reusable, queryable intelligence asset the team could operate independently — freeing the insights team to focus entirely on strategic recommendations rather than data collection.
30,000+
app store reviews now processed in minutes
~99%
time saved compared to the previous manual process*
