CASE / CORA BANK / AI OPERATIONS
Turning high-volume customer emails into reviewed AI-assisted responses.
Cora’s customer service operation handled a high volume of email requests through manual response workflows. Repetitive tasks consumed agent time, limited scalability and increased operational costs.
I helped design Gepetto Skills, an internal AI product that generated contextual response drafts from customer messages and account data, while keeping people responsible for validation, quality and final delivery.
92%
Operational requests
Semi-automated within three months
Seconds
To generate a response draft
Preserving quality and accountability
COMPANY / CORA BANK
PRODUCT / GEPETTO SKILLS
ROLE / PRODUCT DESIGNER
AI OPERATIONS / GEPETTO SKILLS
FROM MANUAL EMAIL HANDLING TO REVIEWED AI RESPONSES
PROBLEM / A REPETITIVE OPERATIONAL LOOP
Every new email restarted work the operation had already solved before.
Analysts repeatedly interpreted similar requests, searched for account context and assembled responses manually. As volume increased, capacity remained tied to repetitive writing and additional BPO support.
INCOMING EMAIL
An unstructured customer request enters the service queue
CONTEXT LOOKUP
The analyst searches for account and operational information
RESPONSE ASSEMBLY
Similar answers are written again from the beginning
HUMAN CHECK
Accuracy, tone and customer context still require validation
SCALE LIMIT
More volume requires more manual capacity and operational cost
STRATEGIC DECISION / AUTOMATE THE DRAFT, KEEP THE DECISION HUMAN
Design for assisted judgment before full automation.
TRADE-OFF / The first release kept human validation mandatory, prioritizing safety, traceability and operational learning before expanding automation.
Generative AI could reduce repetitive writing, but customer service responses involved personal information, account context and operational risk. The workflow required control over accuracy, tone and final delivery.
We designed Gepetto as a semi-automated experience. AI generated a contextual draft, while an analyst reviewed, edited and approved every response before it was sent.
To make this reliable, we first structured service tags, skill ownership, testing and performance tracking. This created a foundation for measuring each skill and improving it over time.
An internal hackathon, Corathon, helped align Design, Engineering, Data, Cybersecurity, TechOps and CX around the capabilities and constraints of a recent technology.
EVIDENCE → 5 IN-DEPTH INTERVIEWS → CROSS-FUNCTIONAL IDEATION → IMPACT-EFFORT PRIORITIZATION → MARKET & TECH RESEARCH
SOLUTION / THREE CONNECTED EXPERIENCES
One operational loop for creating, reviewing and improving AI skills.
Gepetto connected three needs that are often separated in AI initiatives: building reusable skills, embedding response generation in the service workflow and measuring whether each skill was useful.
Managers could create and assign skills by service topic. Analysts received contextual drafts to review. Accuracy and response-time data supported continued improvement.
HUMAN-IN-THE-LOOP - INTERNAL DATA GOVERNANCE - SKILL OWNERSHIP - ACCURACY + TIME TRACKING
IMPACT / THREE OPERATIONAL SHIFTS
AI became a governed operational capability.
PRODUCT PRINCIPLE
AUTOMATE REPETITION. PRESERVE HUMAN ACCOUNTABILITY.
Gepetto created operational value by combining fast response generation with clear ownership, measurable quality and mandatory human review.
01
MANUAL RESPONSE WRITING
→ 92% OF OPERATIONAL REQUESTS SEMI-AUTOMATED
Reusable skills covered most recurring service demands while analysts remained responsible for final delivery.
02
AI USE OUTSIDE THE WORKFLOW
→ CONTROLLED INTERNAL GOVERNANCE
Response generation, customer context, skill ownership and testing moved into an internal platform with traceability.
03
CAPACITY TIED TO REPETITIVE TICKETS
→ LOWER BPO DEPENDENCE AND MORE AGENT CAPACITY
Reduced response effort and BPO costs allowed agents to dedicate more time to complex requests and professional development.
RESPONSE DRAFTS IN SECONDS
Recurring emails became ready-to-review drafts inside the service workflow.
REDUCED BPO DEPENDENCE
Increased internal capacity reduced the need to scale repetitive work through external contracts.
Gepetto turned recurring emails into ready-to-review drafts in seconds, with traceability and human oversight built into the workflow.