My role:
Product and experience strategy
End-to-end workflow definition
Design leadership and coaching
Certain details have been omitted to respect confidentiality.
Led the design of AI-driven customer care experiences that improved containment and enabled new commerce capabilities within chat. Delivered a 5% increase in containment within the first month of launch, reducing the need for escalation to live agents. Established the foundation for end-to-end transactions in chat, introducing new interaction patterns that allow users to move from inquiry to purchase within a single conversation.
*UI details obscured due to feature in active development.
T-Mobile’s AI assistant handles hundreds of thousands of weekly interactions, many involving complex customer needs. These often escalated due to unclear responses, limited workflow support, and low trust in automation. We evolved it into an AI-driven experience capable of resolving issues, supporting decisions, and enabling transactions.
I led design across multiple initiatives transforming chat into an end-to-end AI-driven experience.
Problem
Customers need a fast, trustworthy explanation for bill changes.
Customers frequently ask:
“Why did my bill change?”
When the answer isn’t immediately clear, they escalate to customer support.
Opportunity
Provide a clear, instant explanation inside chat — without requiring users to dig through details or perform their own calculations.
Approach → Structuring complex information for speed and clarity
The core challenge was structural:
How do we make complex billing changes instantly understandable?
We evaluated three models:
Table-based explanations — structured and scannable
Visual comparisons — tables supported by charts
Guided flows — step-by-step, progressively disclosed explanations
KEY DECISION PRINCIPLE
We avoided combining multiple models, as this increases cognitive load and slows comprehension. This allowed us to isolate which structure delivers the fastest path to understanding.
KEY DECISION i owned
Shifted billing explanations from visual-heavy representations to structured, table-based responses to improve clarity and reduce user confusion.
The final direction focused on clarity, trust and control.
Table-first design for fast, reliable understanding
Category grouping to simplify scanning
Clear totals that reconcile all changes
Progressive disclosure to support deeper exploration without clutter
This approach creates a straightforward, trustworthy explanation experience, reducing confusion and helping users resolve questions without contacting support.
Problem
Add-on purchases are high intent, but too slow and fragmented. Customers often need to quickly purchase add-ons like data passes, roaming, or device protection.
However, the current experience requires:
navigating away from context
multiple steps to configure options
unnecessary friction before checkout
Opportunity
Enable instant, in-context purchasing — allowing users to complete add-on purchases the moment intent is detected.
CORE PRINCIPLES
One-tap checkout
Uses saved payment methods to eliminate friction
Context-aware configuration
Adapts dynamically based on:selected line
destination
coverage dates
Inline, not interruptive
Purchase happens within the existing flow — no navigation awaySystem-level scalability
Designed to support multiple add-on types (data, protection, subscriptions)Privacy, accessibility, and performance by default
This established a reusable purchase pattern, not a one-off flow.
*Details obscured due to feature in active development.
Rapid strategy → validated interaction model
This direction was:
defined and aligned in under 48 hours under executive urgency
approved at the SVP level
translated into a set of testable interaction patterns
We've have validated three initial flows (currently under development Q1-2026)
International data pass
Device protection (insurance)
T-Satellite add-on
Each flow applied the same core interaction model, validating its flexibility across use cases.
Testing focused on speed, clarity, and confidence
We tested lightweight, embedded purchase flows to determine which pattern best balances:
Speed of action
Clarity of options (reducing doubt)
Confidence before purchase
Result
Across all flows:
6.2 / 7 average comprehension
Strong user understanding despite:
being a new interaction pattern
users having no prior experience purchasing in chat
Users were able to:
understand the offer
make a decision quickly
complete the flow without confusion
*Details obscured due to feature in active development.
*Details obscured due to feature in active development.
*Details obscured due to feature in active development.