Lightspeed powers retailers and restaurateurs worldwide with their unified point-of-sale and payments platform. Last year, Lightspeed invested in in-house AI by pivoting their data and business analysts to AI implementation. Caroline Kuflik, an AI enablement specialist since July, took the lead with HumanFirst.

The Problem 

Customer support managers at Lightspeed spend hours reviewing call transcripts manually. Support representatives lose time between calls logging summaries and sentiment assessments. Traditional analytics tools and internal reporting systems aren’t granular enough to capture meaningful data at the individual conversation level.

The AI Experiment

With HumanFirst, Caroline took on the challenge of building custom conversation analytics: scoring customer sentiment, providing personalized coaching to support representatives, and tracking call escalations to improve triage from the beginning. With no internal blueprint for this kind of AI work, Caroline was breaking new ground.

"This was the first time in my career where I couldn't just ask someone else how to do it—because nobody had done it this way before. I’ve loved that. We’re learning together, figuring out how AI can move our business forward.”

Caroline Kuflik

AI Enablement Specialist

The Solution

Caroline used HumanFirst to build a comprehensive analysis workflow that processes thousands of support conversations in minutes instead of months. She built, tested, and fine-tuned a series of custom prompts, using Google Cloud’s AI models to manage a multi-dimensional analysis of every call log.

The initial analysis of 10,000 customer interactions gave new visibility into support operations. The team can now: 

• Track call escalations and routing accuracy
• Generate personalized coaching reports for support agents on a regular basis
• Understand trending customer issues and how they correlate with business initiatives
• Measure customer sentiment across custom criteria–without bothering the customer

“This will save our Leads and Seniors so much time in manual inspection. We know today that human QA takes an average of 14 minutes per interaction evaluation. Assuming we were to QA all 10,000 chats per month (which is impossible today), this could save us 2000+ hours each month.”

Angelo Livanos

Senior Director for Global Support

From Experiment to Company-Wide Adoption

Caroline’s project saw immediate interest from parallel departments. What started as a pilot for one support segment quickly expanded to global support teams and even the sales organization.

Other teams that thought they'd have to wait months for insights could suddenly see exactly what was happening across customer engagements. And because the workflow lives in HumanFirst, new users can easily adapt it to different use cases.

Caroline’s work eliminated hundreds of monthly hours of manual review, provided more feedback to every employee, improved call routing accuracy, reduced support costs through improved triage, and replaced multiple SaaS tools. 

“The escalation topic data is now surfacing insights we were not previously aware of, such as the number of interactions that require Account Managers. This is now supporting operational discussions with those leaders to find how we can better address these. We can and will apply similar logic to other findings like this in the future.”

Angelo Livanos

Senior Director for Global Support

‍

Beyond analytics, the project accelerated Caroline’s transition from data analyst to AI specialist, and demonstrated for Lightspeed the power of investing in internal AI capabilities. Caroline’s team is now working on automated solutions for call routing and in-depth sales call analytics.

"HumanFirst changed what I thought was possible with AI. Their team helped me build prompts that could extract exactly what we needed. These capabilities are now central to my work, and I'm just getting started."

Caroline Kuflik

AI Enablement Specialist