How to Use AI to Assess and Retain Your Best Agents

One of my favorite topics to build better customer experience (CX) and agent or employee experience (EX) is to figure out who your best agents are, find more frontline staff like them, and do everything you can to hold onto them. Out of the 80 CustomerThink articles I have written over the last 22 years, seven of them address finding and keeping your best agents including a pair of articles in late 2016 and early 2017, “Using Big Data Analytics to Find Your Best Agents and Supervisors.”

Considerations in agent retention

Now, thanks to the power of GenAI, LLMs, and new forms of predictive athletics, we are able to solve this challenge much more easily. Consider the following:

Cost of replacing top agents

When you lose one of your best performing agents or other frontline employees, including retail sales representatives, bank tellers, or field support engineers, you need to replace them with three headcounts. However, almost every company that I’ve seen replaces these best performers with only one headcount, thereby ensuring that they will never catch up with the demand from consumers.

Here’s a quick example. Lawrence has been a customer service representative for six years, able to juggle chat, phone calls, and social posts with equal aplomb, and not only is he fast at responding to customer inquiries and resolving their issues, but he is also really good at it. Customers routinely give him top box scores or NPS scores, his first contact resolution (FCR) is high, and everyone around him knows he’s the guy to go to if you have a question. Somehow, he can answer those questions and yet still maintain his full workload in such a distinguished manner. Now, if he is handling six contacts per hour with a 90% FCR, but a new hire is only able to handle three or four contacts per hour with a 50 or 60% FCR, customers will not get handled as quickly, nor as effectively. Only if the company hires two or three Lawrences will the performance level match what he was able to perform.

Agent turnover continues to be horrible

Recently, we’ve seen companies that have 135% annual turnover of their frontline customer service representatives, meaning that the average tenure for a new hire is only about seven months. Clearly, this is far less than Lawrence’s tenure as a CSR, so for every Lawrence who remains in the organization, many more will attrite faster than seven months. It’s almost impossible to maintain contact control, FCR, quality performance, and great CX with turnover that high.

Lack of visibility into third-party agent performance

Many BPO outsourcers and contract third-party companies for field service and installation do an outstanding job. However, all the evidence I’ve seen over the years is that captive agents or badged employees tend to perform better over time. Based on fears of co-employment, many companies have little to no visibility into the performance of their contracted, third-party agents or field service teams. This makes it even harder to determine the best agents, collect best practices, and fix the right staffing levels for anticipated demand.

Efficiency vs. effectiveness

In my previous articles, I shared frameworks and approaches to determine the best agents based on a selected number of KPI’s for efficiency compared to a limited number of KPI’s for effectiveness.

Among the better predictors of efficiency are:

• Percentage of long or short calls handled
• Shrinkage rates
• Dynamic individual handle time (DIHT) instead of using the same AHT for every agent.

Among the better predictors of effectiveness are:

• FCR = 100% minus “snowballs” (repeat contacts)
• Post-contact surveys
• QA scores

However, many of these metrics are based on samples instead of a complete analysis of every interaction. Most companies perform only four to eight QA audits per agent per month, and the response rate for post-contact surveys is often under 10%.

Now AI can help

Here’s where AI, including GenAI and large language models, is coming to the rescue.

In some of my previous articles, I talked about the role that Big Data can play and that’s still part of the solution. But now we can analyze 100% of customer interactions to determine an agent’s QA score and a complete contact effectiveness score (c-sat, NPS, or customer effort). In addition, AI can produce the dynamic individual handle time and scheduling challenge. DIHT includes the agent’s tenure, the complexity of the tasks assigned to the agent, and the needs of the customers being presented for support. This has always been a difficult metric to achieve, but now it is far easier thanks to the power of AI.

In addition, LLMs can provide precise coaching advice to help agents move from a weaker performance level into the best agent corner of the balanced scorecard. For example, if Renee works slowly with a higher percentage of long contracts, but routinely gets excellent scores for her effectiveness, she will want to know how to become faster without losing that effectiveness. On the other hand, Lawrence, our veteran whose departure prompted the need to hire three more agents, was, before he left, performing so well that what he needed to know new expectations so he could continue to be one of the top performers in the organization.

Striking a balance

In short, AI, together with big data and advanced analytics, can finally determine the exact position of every agent on a balanced scorecard of efficiency versus effectiveness. Then, you can roll up these scores and action plans to coaches or team managers, senior managers, centers, and if you have multiple locations — outsourced and/or captive — you can look at the overall performance very precisely company by company, center by center.

Finally, we are now able to answer with confidence the question:

“Who are our best agents and how can we hold onto them so we don’t suffer declines in customer experience and overall performance levels and quality?”