Stop Wasting Surveys on “Plain Vanilla” Calls: Using AI to Improve CSAT and Agent Evaluation

Source: Pixabay

By using AI to review routine contacts and concentrating surveys and evaluation on truly challenging issues, you can cut cost and fatigue while gaining a fairer, more actionable view of CSAT and agent performance.

Now that AI can quickly code the reason for contact and identify dissatisfaction and delight in both digital and voice conversations, there is a major opportunity to reduce both your tracking survey and evaluation expense by at least one third — if not half — while enhancing the effectiveness of your CSR and contact center evaluation process.

The core idea is simple: stop surveying the plain vanilla contacts that are “no-brainers.” Because almost all survey engines charge by the number of surveys dispatched and/or responses received, focusing your surveys on the 30% of more challenging contacts immediately cuts cost while dramatically improving what you learn. AI can scan the remaining 70% of contacts for signs of dissatisfaction and delight that merit closer examination. These actions will move the contact center out of firefighting and into a preventive mode of operation.

Simple in concept. Not easy in execution.

What Has to Change

The fundamental requirement is to change your survey and monitoring measurement strategy from a random sample of contacts to a focused set of challenging contacts — those most likely to lead to dissatisfaction and/or excessive resource consumption. This fundamentally changes your approach to evaluating staff and your service operation, and it changes how you analyze customer feedback. It will require your supervisors, analysts, and executives to think more intensively. And it will positively change how your frontline service staff views their own performance and their evaluation process.

Three basic processes must change.

1. Focus on Challenging Contacts

Between 30% and 70% of service calls are “plain vanilla” and can be easily handled by nearly all staff who have completed training. Sending a survey to these callers will seldom identify any unhappy customers, but it will create respondent fatigue, lowering response rates on future surveys about complex issues. Drop surveying these contacts. Your survey burden and cost will drop significantly.

Instead, send surveys to customers with moderately or highly complex issues—cases that present real challenges in reaching a resolution, such as those caused by product defects, customer misuse, misleading marketing, or external factors. Customer service representatives typically code the reason for contact; even better, AI using text or speech analysis can code every interaction, ensuring greater consistency.

2. Change Targets for CSRs, Supervisors, and the Contact Center

Change the target score for CSRs to account for higher contact difficulty. There are now very few free simple calls. Ideally, allow for some customer dissatisfaction in cases where the CSR is delivering genuinely hard news — telling an auto customer they are out of warranty and will have to pay for the engine repair, for example. A 75 or 80 may be a really great score. We’ve stopped using the grammar school grading system.

This strategy will produce much more robust data by issue type on how complex calls are being handled, supporting training and development of more effective responses. It will also identify those issues where a process or product fix may eliminate the problem entirely.

Supervisors may not be happy with this new mandate. They will have to provide CSRs with feedback not just based on a broad standard outline, but on how to handle particular types of contacts. AI can develop a first cut of such advice, but supervisors will still have to think and tailor the feedback to the types of calls handled and the individual skills demonstrated.

The HR process of ranking and evaluating CSRs will also have to change. Reps can no longer aim for 95% or perfect scores earned on a myriad of simple calls. Now, as Sally Hurley, CEO of VIPDesk, says below, all the calls will be challenging, and targets need to reflect that reality. You will also be able to set targets by type of issue: denied claims, beyond-warranty situations, and so on.

3. Use AI to Flag Challenging Contacts and Delight Experiences

AI can review the 40–70% of simple calls to confirm they were handled effectively and identify moments of intentional delight — instances where a representative turned a routine interaction into a genuinely positive experience. Surveys will not be sent to these customers, reducing burden and cost while AI captures the signal.

Is This Strategy Crazy? Two Veteran Executives Weigh In

Sally Hurley, CEO of VIPDesk — which services affluent customers of more than a dozen premium brands known for great service and value — says:

“I love this! I’ve had this opinion about quality for so long and while necessary, always felt demeaning to the agents especially if it isn’t their fault. We’ve been moving toward this approach to measurement has been necessary, but supervisors and HR will need to adapt.
 
I do think AI will eventually take this all off the table. It will be known how customers feel about the products and the service. We already see CSAT scores dropping for agents when AI is implemented and handling easy routine inquiries. Makes sense… more complex interactions obviously have more issues and unhappy customers. I love this for now.
 
In a year you may be writing… the death of NPS because AI will give feedback to agents and train them on gaps, clearly show where the problems exist (product development or production) and heck…. maybe a bot will send apology notes and flowers. We shall see. We’re already moving in the direction of evaluation by issue and development of flexible solution spaces to empower CSRs and support their success.”

Ibrahiem Atshan, VP of Contact Solutions at ACCPremiere, offers a complementary perspective:

“Survey fatigue is real, and we agree organizations need to rethink how often they ask customers to complete surveys. The approach suggested here is interesting, but in our experience it’s important to take it a step further.
 
We are only using surveys in more complex situations where it’s harder to fully understand the customer experience. We just see that as a more targeted use, rather than the main approach. Interaction analysis can cover the broader, real-time view, with surveys adding value in specific cases where extra context is helpful.
 
Rather than relying primarily on surveys, we analyze the interactions themselves. By using conversational intelligence and data mining across calls, chats, and other touchpoints, it’s possible to evaluate every interaction and generate a satisfaction signal at scale. That allows organizations to understand sentiment across the entire customer base, not just from the small percentage of customers who respond to surveys.
 
This also provides something surveys can’t: real-time insight. Instead of waiting for survey responses, interaction analysis surfaces trends, friction points, and emerging issues as they happen.
 
It’s also important to remember that customer satisfaction isn’t only about agent performance. Processes, policies, and the overall customer journey all influence the experience. Mining interaction data helps uncover those broader drivers and gives organizations a more complete view of what’s actually impacting CSAT. With conversational intelligence mining all interactions, we can spot trends and identify successful agent behaviors, language and opportunities that can be used in training/coaching to improve all agents’ performance and therefore overall CSAT.”

Five Benefits of This Approach

1. Greatly reduced survey costs. Half as many surveys need to be dispatched, but you will learn more details about an important call type from every response.

2. Reduced survey burden and higher response rates. Because surveys are not sent about trivial calls, customers are much more likely to take the time to respond when they do receive one.

3. Better use of supervisor time and effort. Supervisors will no longer waste time listening to simple calls where there is nothing to be learned. All monitoring can be focused on genuinely challenging contact types — product setup issues, cosmetic selection, out-of-warranty calls — producing an actionable improvement plan for each CSR.

4. More valid, helpful evaluations of CSRs. CSRs will be happier because they are no longer being evaluated to “get 100” on easy calls, but on how they handle complex contacts and sharpen their use of empowerment and explanatory skills. Tough calls will be recognized as tough. The fiction of the random sample is gone.

5. Accountability is fixed at the right level. Analysis by contact type allows responsibility to be assigned at the proper level — CSRs are accountable for their part of the call, while product management deals with the more strategic upstream issues.

Next: How to Execute this Strategy

The contact center industry has been measuring the wrong things for a decade. AI now makes it practical to fix that. My next article will lay out exactly how, from data collection and priority-setting to action planning, recognition, and where AI fits best in the new system.