STATEMENTS | 10 MIN READ

How Speech Analytics and Automated Agent Scorecards Improve Agent Results and Reduce Compliance Risks

What is an Automated Scorecard?

Automated scorecards allow technology to do what’s traditionally been done manually by supervisors one call at a time. Automated scorecards use keywords and keyword lists to auto-answer crucial questions about every call.

• Did the agent read the “mini-miranda” to the consumer?
• Did they tell the consumer that call was being recorded?
• Did the agent, or the consumer, use profanity or threatening language?
• Did the agent verify right party contact before discussing a debt?
• Did the agent communicate any additional information required by the business?

Instead of relying on subjective judgment, automated scorecards assess how an agent performs into specific, objective measurements.

• FDCPA and CFPB compliance
• Script adherence
• Consumer-agent sentiment analysis
• Customer satisfaction and verification of client SLAs
• Agent training and performance monitoring at scale

Not only do automated scorecards provide managers with clear, actionable information about how an agent is doing; they also allow managers to quickly isolate and flag problematic calls across a contact center.

For example, an automated scorecard can be used to flag any call where profanity or other prohibited language is used.

Manual Scorecards vs. Automated Scorecards

Using scorecards to assess agent performance has historically been a manual process where a supervisor manually answers a series of questions based on reveiwing a recording of a call.

This process is tedious and time-consuming. In fact, manual scoring is often done overnight using offshore services. Outsourcing this process is costly and oftentimes untrustworthy.

Automated scorecards also eliminate the potential for personal bias. The technology assesses each call and agent based upon the same criteria.

Monitoring Agent Activity By Exception

Automated scorecards allow supervisors to monitor calls by exception. Instead of listening to every call looking for a “needle in a haystack, a supervisor is free to only review flagged calls.

This strategy of managing by exception is really only possible through the speech analytics technology and the use of automated scoring.

Understanding the Role of AI and Machine Learning

Building effective automated scorecard starts by understanding the basics of how speech analytics works.

Speech analytics uses AI and machine-learning processes to turn call recordings into actionable data.

Using Speech Analytics to Redact Sensitive Consumer Information

Recording contact center calls is crucial for ongoing agent training. But retaining call recordings are also necessary if a company needs to prove what was said during a conversation if legally challenged. However, storing call recordings comes with its’ own set of compliance risks when sensitive consumer information is being shared.

Through speech analytics, sensitive information like credit card or social security numbers can be identified and scrubbed automatically before the recordings are ever stored. When redacted, sensitive information in the call’s audio file is removed.

This process allows companies to compliantly store calls without risking the consequences of a data breach or future lawsuit.

Using Speech Analytics for Sentiment Analysis

Speech analytics uses keywords and keyword lists to create a  separate sentiment analysis of both the agent and consumer’s conversation. This gives supervisors an indication of whether the overall sentiment of a call is positive or negative.

Sentiment analysis is helpful to assess individual calls, but also can be used to aggregate an agent’s sentiment across all the calls he or she has placed or received.

The 4 Steps of Speech Analytics & Machine Learning

Steps 1 and 2 of this process are refined and improved over time as more data is fed through the computing engines. This process improves the accuracy of transcriptions by allowing computers to learn regional dialects and specific variations in how agents pronounce certain words.

It also helps computers add important contextual data points necessary to customize its analysis to different business needs.