designing with empathy at scale

How We Used AI to Empower Support Agents with Real-Time Sentiment Insights

COMPANY

Freshworks

ROLE

Product Designer

EXPERTISE

UX/UI Design

YEAR

2023

awards & recognition

Received the Craftsmanship Award

recognized for UX leadership in CRM product innovation and development

The Challenge

Customer service agents juggle dozens, sometimes hundreds, of conversations at a time. But not every ticket is created equal. Some customers are frustrated, some are anxious, others just want a quick answer.

The question we asked

How can we help agents prioritize and respond with empathy, instantly, and at scale?

The Goal

To design a real-time AI-powered Sentiment Analysis feature that helps support teams:

Research & Planning

Instantly detect customer emotions

Design & Prototyping

Prioritize negative sentiment tickets

Implementation

Deliver faster, more empathetic responses

Testing & Optimization

Monitor sentiment trends to improve support strategy

Discover and Research

We kicked off by shadowing support teams to understand their pain points. Key finding:

Intelligent Scheduling

Agents couldn't easily identify frustrated users until it was too late

Calendar Integration

Manual triaging was slow and error-prone

Personalization

Teams lacked insights into emotional trends across conversations

We also understood how AI models score sentiment - from 0 (most negative) to 100 (most positive) - based on language cues and tone

Designing the Experience

We kicked off by shadowing support teams to understand their pain points. Key finding:

Intelligent Scheduling

Agents couldn't easily identify frustrated users until it was too late

Calendar Integration

Manual triaging was slow and error-prone

Personalization

Teams lacked insights into emotional trends across conversations

We also understood how AI models score sentiment - from 0 (most negative) to 100 (most positive) - based on language cues and tone

real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.
real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.
real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.
real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.

real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.
real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.
real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.
real-time sentiment detection
Each conversation dynamically displays a sentiment emoji and score. Agents get immediate emotional context without digging through message history.
smart filtering
Agents can sort and filter conversations based on sentiment. Want to address the angriest customers first? Just sort by "Sentiment". Simple, powerful, intuitive.
Trend view for managers
For leads and supervisors, we introduced a dashboard to show average sentiment at the start and end of each chat, sentiment trend over time, and breakdown of sentiment across all conversations. This helped inform coaching, staffing, and process improvements.

Outcomes

25% Faster Ticket Resolution

Agents could instantly identify and prioritize customers who needed help the most

+20% Agent Satisfaction

"It feels like I know who's upset before I even start typing..," one agent told us

Higher First Contact Resolution

Empathy-driven responses reduced the need for follow-ups

Leadership-Level Insights

Managers used sentiment trends to make better staffing and training decisions

Want to see how agents use it in action?

Watch the video

Agents could instantly identify and prioritize customers who needed help the most

Read the support article

"It feels like I know who's upset before I even start typing..," one agent told us

The Challenge

The Challenge

The Challenge

The Goal

The Goal

The Goal

Discover and Research

Discover and Research

Discover and Research

Designing the Experience

Designing the Experience

Designing the Experience

real-time sentiment detection

smart filtering

Trend view for managers

real-time sentiment detection

smart filtering

Trend view for managers

Outcomes

Outcomes

Outcomes

Want to see how agents use it in action?

Want to see how agents use it in action?

Want to see how agents use it in action?