
🏠 Evolving a Mortgage Payment IVR System
💡 Overview
When a regional credit union saw low self-service rates and frequent call transfers, I led the full redesign of their mortgage payment IVR system. The new voice experience is faster, more natural, and purpose-built to increase call containment, reduce friction, and improve user trust.
Project type
Rules-based IVR for a Financial Services firm
Team, Role and Responsibility
Team: 1 PM, 1 Data Scientist, 2 Engineers, 1 Customer Success Manager
My Role: Lead Conversation Designer
Tools Used


IDE
🚧 Problem
The legacy system frustrated users:
-
Rigid, repetitive authentication logic
-
Verbose and unclear prompts
-
High agent handoff rate (keyword "Agent" was included in the Welcome menu)
-
Completion rate for payments hovered around 45%
Key learning: Users wanted self-service. The IVR wasn’t letting them do it.
🛠 My Contributions
As Lead Conversation Designer, I:
-
Analyzed real call transcripts and drop-off points
-
Rewrote prompts with clearer confirmation and tone
-
Designed a new authentication + payment flow
-
Partnered closely with PM, Engineering, and UX
-
Validated improvements post-launch with KPIs
🎯 Goals
-
Increase the utility of the mortgage payment feature
-
Reduce agent escalation
-
Increase self-service and CSAT
🧠 Key Improvements
-
Smarter disambiguation/ confirmation: Used recent account activity to suggest defaults
-
Natural payment phrasing: "Pay my usual" and “Full balance” recognized
-
Robust error handling: Graceful fallback + triage to live support
-
Clear confirmations: Voice feedback + SMS with confirmation code
-
Streamlined flow: Shorter handle times and fewer steps
💡 Key Learnings
-
Transcript analysis reveals usability pain points others miss.
-
Deterministic IVRs can feel smart! Logic + tone go a long way.
-
Fallbacks can be trust-builders, not just error handlers.
🧠 My Mental Model
Q: Why did I approach it this way?
-
I focused on a single, high-impact journey: making a payment. Improving one critical path would benefit the highest number of callers.
-
I simplified and restructured the IVR prompts to reduce friction, eliminate dead ends, and surface the most valuable options up front.
-
I treated the IVR as a system, not a script. I aligned conversation design with backend logic, business rules, and escalation paths.
Q: What did I learn from the solution?
-
Simplicity scales. Reducing decision points and increasing clarity led to better outcomes across metrics.
-
In an IVR, how you say something matters as much as what you say! Tone, pacing, and perceived helpfulness build trust.
-
Designing with real data (e.g., call drop-off points, intent misroutes) turns assumptions into action and drives measurable improvement.
V1 Mortgage Payment Call Flow




V2 Mortgage Payment Call Flow


