ENTERPRISE SAAS • B2B • AI
NICE Copilot for Agents
Agents need live coaching mid-call (sentiment, next-best-action) without breaking focus on the customer in front of them. Manual QA only reaches a tiny fraction of interactions.
ROLE
Senior UX Designer
TEAM
Senior UX Designer, 1 UXR, 1 PM, 1 Salesforce PM
PLATFORM
B2B Enterprise SaaS, Salesforce
SCOPE
AI product design, UX research, Salesforce Lightning Design Systems, AI tooling
OVERVIEW
Real-Time Interaction Guidance (RTIG) is an AI-powered tool built into NICE CXone that surfaces live coaching, sentiment signals, and next-best-action prompts to agents during active customer interactions.
I led the 0 to 1 design, embedding it directly into Salesforce design system. The product has since evolved into what NICE now calls Copilot for Agents.
The core design challenge wasn't just what information to surface, it was determining what mattered in a given moment, when it should appear, and how to present it.
Getting this wrong meant either agents ignoring the tool entirely, the tool becoming another source of cognitive load during an already high-stakes interaction.
Adding to the complexity, this guidance needed to live inside Salesforce, a CRM environment, governed by its own design language and constraints.
CHALLENGE
PROCESS
UX Research
Both on-site and remote user research was conducted with agents and supervisors across client organizations, combining quantitative user data with moderated usability testing to understand real behavior under real conditions.
This research shaped decisions about timing, visual hierarchy, and interaction patterns, like when a sentiment shift warranted a prompt, how much guidance was too much, and how to design for a glance without demanding an agent's full attention away from the customer.
Salesforce Collaboration
I worked closely with the Salesforce team and product managers to understand the Lightning Design System, designing components that felt native to the CRM agents already trusted.
I also evolved the CXone design system and component library used across the product, focused on WCAG accessibility and consistency at enterprise scale, and mentored junior designers on interaction patterns and design standards along the way.
SOLUTION
The result was an AI guidance layer that reads context in real time, contact sentiment, conversation history, and relevant knowledge from the knowledge base, and surfaces it as live coaching cues, suggested next steps, and response guidance directly inside the agent's existing Salesforce workflow.
Today, as Copilot for Agents, that same foundation extends across voice and digital channels, combining real-time sentiment detection, knowledge base answers, and AI-guided responses in one unified experience.
Learn more about the the product here.
Maye’s Contribution
Shipped a 0 to 1 AI product now used in production across NICE's enterprise client base
Led on-site and remote UX research, including moderated usability testing with agents and supervisors, to determine what guidance mattered in the moment and how to present it without disrupting agent focus
Established design patterns for real-time AI guidance that balanced usefulness against cognitive load.
Built and scaled a design system and component library that improved accessibility and consistency across the platform
Deepened a design leadership practice built on bringing clarity to ambiguous, high-stakes systems through systems thinking and close cross-functional collaboration with Salesforce, PM, and engineering partners
OUTCOMES
Copilot for Agents is used by Carnival UK, Bamboo, and World of Hyatt.
55%
Reduction in average handle time
23%
Agent attrition drop (42% to 19%)
100%
Interaction QA, vs 1-3%
Featured Work