JENNIFER BECKER
CASE STUDY
Empowering
Agents with Real-Time AI Support
Role: UX Lead (Team: 1 Junior Designer, 1 UX Researcher)
Project Overview
Bringing AI into the Agent Experience
Agents are on the front lines of customer experience — but without access to accurate, timely information, even the best agents struggle. Our legacy call center tools lacked intelligent, in-the-moment support. As competitors began to differentiate with real-time agent assistance, we saw an opportunity to do more than catch up — we could lead.
As the UX Lead, I was responsible for designing a scalable Agent Assist experience that delivered relevant answers in real time, integrated seamlessly with existing agent workflows, and allowed agents to guide and correct the AI — not just passively receive suggestions.
The Challenge: No Context, No Confidence
During calls, agents had no centralized view of what had already been asked by the customer or answered by our automated bot. They lacked:
-
Fast access to accurate, validated knowledge
-
A way to guide or correct the AI if it surfaced the wrong suggestion
-
Summarized context when picking up an escalated or bot-deflected call
-
Without this support, agents frequently paused, guessed, or asked customers to repeat themselves — eroding trust and extending resolution times.
Process & Collaboration
Research
-
Interviewed agents, supervisors, and Sales engineers to understand breakdowns in information access
-
Identified friction during escalations, bot deflections, and high-volume call days
Define
-
Mapped user journeys for calls with/without bot deflection
-
Documented agent goals: speed, trust, clarity, and confidence
-
Aligned feature scope with Sales feedback and competitive benchmarks
Ideate
-
Led brainstorming on voice recognition, AI prompts, sentiment alerts, and real-time action cards
-
Proposed AI correction flows to allow agents to “coach” the system
Prototype
-
Created Figma prototypes for real-time suggestions, summaries, and sentiment overlays
-
Designed adaptive UI blocks for different customer segments (e.g., hotels, entertainment)
-
Delivered demo flows and scripts used successfully by Sales with live customers
Test
-
Partnered with Sales to present and validate demos in real-world customer meetings
-
Collected agent feedback to refine checklist prompts, next-best actions, and AI summaries
-
Iterated features to support multiple verticals and product suites
Design Goals
-
Integrate Agent Assist (Google CCAI, Verint) into our platform without disrupting agent workflows
-
Provide real-time voice-triggered AI suggestions, summaries, and next best actions
-
Allow agents to train and improve the AI through quick corrections or feedback
-
Offer pre-call summaries and in-call live support, especially for bot-escalated interactions
-
Maintain accessibility and clarity under high-pressure conditions
Key Features Delivered
-
AI Summary Panel: Highlights key customer info, what’s been asked, and known issues
-
Real-Time Suggestions: Voice-triggered knowledge suggestions based on active conversation
-
Agent Feedback Loop: Agents can approve, reject, or modify AI recommendations to train the model
-
Sentiment & Emotion Cues: Live indicators help agents adjust tone and pacing
-
Scenario-Based Demos: Custom scripts and visuals tailored to industries like entertainment and hospitality
Outcome & Impact
-
Enabled Sales teams to demo Agent Assist with confidence, leading to strong interest from multiple verticals
-
Improved agent speed and trust by providing in-the-moment support
-
Enhanced customer experience through faster, more accurate responses
-
Positioned our platform to compete directly with Google CCAI and Verint, while aligning with our own design standards
-
Created a scalable framework for embedding AI across the product suite — from voice recognition to next-best-action checklists
Reflection
Agent Assist wasn’t just about plugging in AI. It was about building the right interaction model — one that respected agents as decision-makers, not passive end users. By embedding clarity, feedback loops, and context into the workflow, we helped agents feel supported, not replaced.
This project strengthened our product’s competitive edge and showed that when AI is designed with empathy, it becomes a trusted partner — not just a flashy feature.