Unlocking Faster, More Confident Call Handling with an AI Summary Assistant

Problem

Call center representatives spent 2–3 minutes per call searching through dense, legacy work order pages to answer basic status and delay questions. With a target of 80 work orders per day, this friction directly impacted call efficiency and response quality.

my role

UX Designer — Owned research, AI summary definition, system prompt design, and the AI chat experience on the work order page. Partnered closely with Product and Engineering.

timeline

May 2025 – July 2025

project goals

Reduce time-to-context for work order understanding. Enable faster, accurate responses during live calls without replatforming legacy UI

constraints

Legacy codebase limited structural page changes. Work orders serve many roles (dispatch, service, invoicing, validation). Sensitive internal and vendor data could not be exposed externally.

primary outcome

A strategically scoped AI summary and chat experience launched internally and externally, improving access to critical work order context without redesigning the core page.

Problem

Call center representatives spent 2–3 minutes per call searching through dense, legacy work order pages to answer basic status and delay questions. With a target of 80 work orders per day, this friction directly impacted call efficiency and response quality.

my role

UX Designer — Owned research, AI summary definition, system prompt design, and the AI chat experience on the work order page. Partnered closely with Product and Engineering.

timeline

May 2025 – July 2025

project goals

Reduce time-to-context for work order understanding. Enable faster, accurate responses during live calls without replatforming legacy UI

constraints

Legacy codebase limited structural page changes. Work orders serve many roles (dispatch, service, invoicing, validation). Sensitive internal and vendor data could not be exposed externally.

primary outcome

A strategically scoped AI summary and chat experience launched internally and externally, improving access to critical work order context without redesigning the core page.

How I Approached This

Research & Design Process

How I Approached This

Research & Design Process

01

Identify the Highest-Leverage Bottleneck

Shadowed call center reps to understand where calls slowed down. Rather than auditing the entire page, I focused on moments where reps paused, searched, or cross-referenced information during live calls.

lens used

Cognitive load, workflow efficiency

intent

Pinpoint the specific friction that most directly impacted call duration and quality

KEY QUESTION

Where is time being lost during a live call, and why?

outcome

Confirmed that understanding work order history, delays, and related work was the primary bottleneck.

01

Identify the Highest-Leverage Bottleneck

Shadowed call center reps to understand where calls slowed down. Rather than auditing the entire page, I focused on moments where reps paused, searched, or cross-referenced information during live calls.

lens used

Cognitive load, workflow efficiency

intent

Pinpoint the specific friction that most directly impacted call duration and quality

KEY QUESTION

Where is time being lost during a live call, and why?

outcome

Confirmed that understanding work order history, delays, and related work was the primary bottleneck.

02

Choose Augmentation Over Replacement

Designed an AI layer instead of redesigning the legacy page.

02

Choose Augmentation Over Replacement

Designed an AI layer instead of redesigning the legacy page.

03

Define What “Useful” Means for AI Output

Designed the AI summary around decision-making, not completeness.

03

Define What “Useful” Means for AI Output

Designed the AI summary around decision-making, not completeness.

04

Pivot from Passive Summary to Active Call Support

Shifted from review-oriented summaries to question-driven interaction.

04

Pivot from Passive Summary to Active Call Support

Shifted from review-oriented summaries to question-driven interaction.

05

Expand Value Without Expanding Risk

Adapted the experience for external clients safely.

05

Expand Value Without Expanding Risk

Adapted the experience for external clients safely.

Key Decisions

Solution & Reasoning

Key Decisions

Solution & Reasoning

decision made

Solve for time-to-context with AI, not page cleanliness

decision made

Ship a Standardized Payroll Summary Report

decision made

Prioritize common questions over exhaustive summaries

decision made

Design Reporting Around Post–Pay Run Workflows

decision made

Add verification guidance for AI responses

decision made

Make All Costs Traceable to Totals

decision made

Enable shared understanding before scaling to clients

decision made

Make All Costs Traceable to Totals

My Results

Outcomes & Impacts

My Results

Outcomes & Impacts

Time-to-Context

Reduced

Reps no longer need to sift through logs and related work orders to understand status and history.

Time-to-Context

Reduced

Reps no longer need to sift through logs and related work orders to understand status and history.

Call Readiness

Immediate

Common caller questions are answered directly through AI during live calls.

Call Readiness

Immediate

Common caller questions are answered directly through AI during live calls.

Cognitive Load

Lowered

Work order history, delays, and related work are summarized into a single, interpretable view.

Cognitive Load

Lowered

Work order history, delays, and related work are summarized into a single, interpretable view.

Question Coverage

3/3

Highest-frequency caller questions are addressed by default.

Question Coverage

3/3

Highest-frequency caller questions are addressed by default.

Adoption Signal

Tracked

Engagement data and qualitative feedback confirm real-world usage and perceived value.

Adoption Signal

Tracked

Engagement data and qualitative feedback confirm real-world usage and perceived value.

UX Researcher & Designer

2026 Designed by Emma Blackwell. All rights reserved.

2026 Designed by Emma Blackwell. All rights reserved.

Emma Blackwell