A global ticketing platform processing over 500 million tickets annually manages millions of fan interactions across voice, chat, and email channels. Behind each conversation is a fan trying to solve something that often matters in the moment: a family trying to transfer concert tickets after a loss, a sports fan looking for answers about a postponed game, or a buyer unsure whether a payment went through.
Before partnering with Level AI, supervisors evaluated just 2% of those customer contacts using a legacy tool called Playvox. The remaining 98% passed without review. Leaders were managing thousands of agents without a clear view of what was happening in most customer conversations, leaving supervisors unable to confirm whether agents resolved issues on the first attempt or why some calls took much longer than others. Turning to Level AI, the organization established complete operational coverage across all global support channels.
Scattered spreadsheets and unverified data stalled operational growth
The ticketing powerhouse manages a high volume of interactions, and to manage this volume, internal teams work alongside global BPO partners, including 200 agents at Concentrix in the United Kingdom, 900 agents across North America, and contact center teams in Mexico, Brazil, and Peru.
These interactions generate plenty of data, but it lived across disconnected systems and does not always tell a reliable story. Legacy evaluation tools went unused, while quality supervisors still had to manually export voice call data from Five9 and email & chat data from Zendesk and Databricks into tools such as Domo and Power BI. Even the way calls were categorized was built more for product reporting than for helping supervisors understand agent performance.
As the Sr. CX Director explained, "Our dispositions were created with an external point of view of how do we get this data to product to resale versus exchange versus transfer. That's great for external, but it's not really powerful for internal."
Quality supervisors maintained isolated Excel spreadsheets to log coaching sessions. This leaves gaps in two places
1. The lack of interaction visibility created friction between team leads and frontline agents. Reviews based on two calls per month frequently produced subjective feedback, causing high-performing agents to challenge compliance scores. When supervisors attempted to address complex processes like ticket exchanges or event cancellations, they lacked complete conversation records to support their coaching.
2. Coaching notes also lived in separate spreadsheets, so managers had no single place to see what had been discussed or whether performance improved afterward.
The visibility gap also had a real financial cost. Internal estimates showed that cutting just 30 seconds from average handle time across global queues could save nearly $1 million a year. But without a clear view of the conversations themselves, supervisors could not see where time was being lost or coach agents to resolve issues more efficiently.
The team had heard promises about automated QA before, but previous tools struggled with its multiple channels, complex workflows, and multilingual operation. When evaluating new partners, Level AI stood out by being clear about automation limits and working directly with frontline supervisors to build workflows around how teams actually worked.
Building one clearer view across quality, coaching, and customer conversations
Level AI integrated with Five9 for voice transcription with its native integration, establishing SFTP pipelines for Zendesk data flowing through Databricks, and deploying workstation screen recording across global contact center desks. Technical teams configured pipelines to separate automated system messages from human responses and properly attribute ticket ownership when issues moved between departments.
Operations teams integrated four primary product capability areas across global workflows:
Automated quality management & centralized coaching
Level AI now evaluates 100% of voice, chat, and email interactions, with scoring calibrated to more than 85% accuracy across English, Mexican Spanish, and French Canadian. Localized workspaces allow Spanish- and French-speaking supervisors to operate in their native languages, and evaluators use custom metric tags to flag first-contact resolution failures when fans send multiple follow-up emails on a single ticket.
Coaching now happens in one place, with relevant transcript moments pulled directly into each review. Managers can see what was coached, how often it happened, and whether performance improved over time.
AI Workers
Operations teams use natural language queries via Level AI’s AI Workers to analyze conversation trends, retrieve specific call histories, and generate executive summaries without waiting for a data analyst. This capability provides instant answers to complex operational questions through real-world scenarios:
Resolving high-priority escalation: A grieving parent posted on social media about their son, who had passed away from cancer, after struggling to transfer his four tickets for a rock concert to close friends so they could attend in his memory. A contact center manager copied the post into AI Workers, which surfaced the customer’s previous calls within seconds. The team confirmed what had happened and resolved the transfer.
Evaluating mass email impact: After a local hockey game was canceled, client support managers sent refund emails to 2,000 fans. AI Workers analyzed responses in real time and supported A/B testing on email phrasing, giving teams an early read on resolution rates without waiting for survey results.
Diagnosing regional delivery issues: Regional leaders in Europe used AI Workers to investigate reports of SMS passcode failures in Israel and Turkey. The query surfaced 30 matching interactions instantly.
Uncovering policy rules: AI Workers compared successful and unsuccessful ticket exchange calls to identify the four criteria tied to a successful exchange. Teams could use those findings directly instead of searching external policy documents.
Screen Recording
Level AI Screen Recording provides visibility into the activities on agent desktops during a customer conversation. Supervisors can see exactly where a process breaks down, whether an agent misses a system prompt, or where extra time is being spent, showing where agents get stuck or miss something important on screen.
Concrete operational scenarios highlight how teams use screen recordings to diagnose frontline execution:
Identifying unflagged marketplace orders: Screen recordings showed that agents were missing a red order-source tag during exchange requests. Supervisors used that evidence to retrain agents on when to check marketplace flags.
Troubleshooting telephony timestamps: Screen recordings helped engineering teams trace audio-video timestamp mismatches to device configuration issues and restore sync across regional sites.
Voice of the Customer
With Level AI’s Voice of the Customer, this company is able to analyze every interaction to show which issues are growing, how fans are feeling, and where new friction is emerging without requiring agents to manually tag call dispositions.
Operations leads structure these insights into dynamic topic hierarchies that align with operational priorities and seasonal demand:
Aligning feedback with event calendars: Customer friction points shift dramatically between winter arena seasons and summer stadium tours. Operations teams configure topic hierarchies that automatically isolate seasonal trends and track issues such as venue entry pass errors or mobile wallet transfer glitches as major concert tours move from city to city.
Preparing for product launch rollouts: When product teams prepared to release two new digital ticketing features—an "Activate" ticket button and an "Assign" feature for group purchasing—operations leads set up pre-built conversational topic categories in Voice of the Customer. The moment the features went live, management monitored incoming fan sentiment and downstream support friction in real time, giving teams an early view of customer reactions rather than waiting weeks for survey results.
Standardizing CSAT across channels: Because voice, email, and chat previously used separate feedback metrics, technical teams built a unified customer satisfaction model inside Level AI, mapping multi-channel ratings into a single score to evaluate global fan sentiment across all lines of business.
Achieving their highest customer satisfaction score with Level AI
Moving from a 2% sample to 100% conversation coverage changed the way teams understood performance. Agents and supervisors could finally work from the same evidence instead of debating what a small sample might mean.
Reaching a six-year customer satisfaction peak
In May 2026, the company recorded its highest global CSAT score in six years. With every interaction visible, teams could spot recurring issues earlier and give agents more specific feedback across voice, chat, and email.
Reducing reliance on external analytics tools
As confidence in Level AI data grew, contact center leaders began reducing their reliance on tools such as Domo and Power BI. Operations teams could answer more questions directly instead of waiting on separate BI requests.
Finding handle-time savings
Internal calculations showed that reducing average handle time by 30 seconds could save nearly $1 million annually. Supervisors used transcripts and screen recordings to identify where calls slowed down and coach agents on more efficient processes.
Centralizing coaching
Coaching moved out of separate Excel files and into one shared record. Managers can see what was coached, how agents responded, and whether performance improved over time.
Creating a clearer view across global teams
With every interaction visible, leaders have a clearer view of what fans and frontline teams experience every day. QA now gives teams a consistent way to identify issues, coach agents, and track customer sentiment across the global operation.
