How to automate your customer service? Controlling satisfaction and reducing costs, the impact of virtual agents.
The guide to getting started and getting immediate results
Introduction
More than 71% of customers prefer to solve their problems independently rather than having to go through a customer service department. Gartner's latest research shows that the self-care trend is gaining momentum among B2B and B2C customers.
Yet many companies have yet to deliver a satisfactory self-service experience with only 13% of requests being resolved without human intervention.
This discrepancy weighs on customer service teams who are overloaded with requests that could be automated. Internally, processing costs and team motivation are directly impacted, while satisfaction and NPS (Net Promoter Score) are not optimized on the customer side.
The use of a virtual agent equipped with conversational artificial intelligence (AI NLP) has the potential to permanently change the performance of customer services by allowing them to automate up to 70% of incoming requests.
This guide will help you understand the different steps to follow in the evaluation and implementation of a virtual agent integrated to your telephony.
Identify your needs
To act effectively on your customer relations, it is essential to take a step back and define your objectives. Improving customer satisfaction, the pick-up rate or the qualification of incoming calls are not at the same level of the customer experience and therefore do not require the same strategies.
To save time, you can represent your customer journey in a funnel to visualize at which stage your objectives are. This will allow you to identify the bottlenecks that could prevent you from reaching your results .
This method avoids the phenomenon of the pierced basket, a situation in which the efforts are reduced to zero because of a problem that is present higher up in the path. To simplify the process you can use our 3-step model
Our 3-step modeling simplifies the visualization of objectives and potential obstacles to their success. The funnel follows the chronology of an incoming call.
1. Quit Rate (QS)
2. Resolution rate (FCR)
3. Caller satisfaction (CSAT)
If you have already done this type of modeling you can skip to the number corresponding to your goal. However, we recommend that you compare your performance with the observed averages to ensure that you act where necessary.
We will detail each objective here to allow you to compare your situation with the best performances and the contribution of a virtual agent on this stage.
1. QS (Quality of Service)
This measure is one of the most closely monitored by the customer and sales departments. The alert threshold for this step is below 80%. Concretely, if you miss more than 20% of your calls because of a lack of human agents or because the calls are made during a time slot when you are not available, then priority should be given to improving this indicator because it impacts all the following steps. To refine this calculation, you can use the linear QS calculation formula which allows you to calculate it on the active time slots of your customer service.
Its formula is :
Linear SQ = Number of time slots > your SQ target / Number of total time slots in the service x 100
Regardless of your starting level, setting up a virtual agent will automatically get you to 100% answered calls.
With 24/7 availability, the virtual agent can either take calls that cannot be handled by busy human agents (overflow) or take all incoming calls to qualify them or resolve requests in real time (first instance).
2. Resolution rate (FCR)
The ability of a customer service department to reduce the number of steps required to answer incoming calls is a guarantee of efficiency. The FCR or First Call Resolution will allow you to identify the capacity of your call center to answer incoming requests from the first call.
Its calculation formula is :
FCR = (number of calls resolved at first contact/total number of calls handled) * 100.
The resolution rates are variable and depend directly on the complexity of the requests. We observe rates between 40% and 80% for calls managed by humans.
If your FCR is below 70% and the previous indicators are optimized, your priority is to act on this resolution rate.
The impact on your FCR of a virtual agent equipped with a conversational AI will be immediate. For simple requests, your agent will be able to directly solicit the internal databases to which it is connected (picking tool, CRM, e-commerce, etc.) to provide the requested information to the caller. For complex questions or procedures, sending an FAQ article can also approach a 100% FCR rate.
However, a virtual agent cannot compete with a human for complex requests. The qualification of the call and the redirection are essential to solicit the resources able to best meet the customer's expectations.
3. Caller satisfaction rate (CSAT)
Customer Satisfaction Analysis (CSAT) has become one of the key measures for most customer services. This measure allows to evaluate the satisfaction of a customer after a contact. It does not evaluate the probability of recommendation as does the Net Promoter Score (NPS) but it is an excellent barometer of the general health of your customer service.
The formula for calculating CSAT is as follows:
CSAT = number of positive responses/number of responses * 100
The commonly accepted range for assessing CSAT health is between 50% and 80%. Above 80% is considered to be a healthy level of customer satisfaction.
On the other hand, if this rate falls below 50%, it is a warning signal for the company because its customers may turn away from the service or convey a negative image of the company to their relatives.
The majority of companies measure CSAT between 50% and 80% and do not evaluate its improvement as a priority, but these figures are sometimes misleading.
Indeed, if on average more than 75% of companies measure their CSAT, they generally do so on a very small sample of their customers.
Several obstacles can be encountered, such as the exchange channel or the response rate. In addition, CSAT is often an aggregate average within the company and does not allow for efficient decision-making because the granularity of the analysis is not sufficient.
Companies at the forefront of customer experience are able to evaluate CSAT for each call and thus track the performance of an agent or department to make the best decisions.
This CSAT collection can be done automatically after a call managed by a human or virtual agent.
A scenario is triggered to ask on a scale of 1 to 5 what is the level of satisfaction of the caller, thus allowing an optimal follow-up of the health of its customers. Companies like Devialet or GoodWe have already implemented this system which can be easily extended to all numbers and languages used within the company.
The performance of a customer service department is directly correlated to its ability to integrate innovations that allow it to effectively manage incoming requests.
Faced with the scarcity of profiles and the increase in processing costs, the improvements in conversational AI used in virtual agents makes it possible to maintain an optimal level of service for companies that adopt these new methods. The integration of a virtual agent is still an unknown land for many companies, so we will address the 3 key points of choosing and installing a virtual agent in the second part of this guide.
Choose and configure your virtual agent
Now that you know the key stage on which your virtual agent must act, it is crucial to identify the right solution for your needs. The goal is not to rank good or bad solutions but rather to allow you to identify which one will deliver the best results for your customer service. Three parameters will come into play, the complexity of the requests handled by the virtual agent, the integration of the data with your existing software and the process changes within your organization.
To simplify your decision making we have segmented the situations into 4 scenarios that correspond to more than 95% of the companies. By selecting the scenario that corresponds to you the most you will know which elements are essential for your specific case in the choice and configuration of a conversational AI.
Scenario 1
Simple request management without integration
This situation corresponds to you if you meet these two characteristics
1. Your virtual agent will only collect the information given by the caller (Name, Phone, reason for the call etc...).
2. This information is not linked to your internal software and you process the tickets directly in your virtual agent interface.
This use case often corresponds to the management of a high volume of incoming requests for sales or support departments in SMEs. The installation time is extremely fast for this type of agent because it has no integrations. One day is enough for the configuration, the setting on line and the obtaining of results.
Scenario 2
Simple request management without integration
You fit this category if you have at least one of the following two goals:
1. The virtual agent must handle complex requests (order modifications, reservations)
2. The agent data must be integrated into a software (CRM, Ticketing, Payment) whose APIs are standardized and known (Hubspot, Zendesk, Aircall)
This is the most frequent use case. It allows to obtain the maximum performance while ensuring a synchronization of the data and the actions within all the internal software. The implementation time for this type of agent is 3 to 4 weeks.
Scenario 3
Simple request management without integration
If your software was created internally or is highly customized (SAP, Oracle), integration can be complex and require significant change management within the organization. The ROI of the virtual agent will not be impacted, but the implementation timeframe will be tailored to your needs.
.svg)
Scenario 4
Simple request management without integration
You fit this category if you have at least one of the following two goals:
1. Processing of sensitive data
2. Installation on an "On premise" environment
3. Caller population mostly elderly (+75 years old)
4. Telephony only managed on mobiles without VoIP
In most of the cases listed above, it is recommended to thoroughly study the anticipated ROI and to evaluate alternative solutions to avoid committing to a project that would have little chance of success.
Conclusion
The majority of companies are out of step with consumer expectations in terms of customer service innovation. The self-care trend is a strong demand from customers, with more than 71% wanting to solve their problems without human intervention. However, only 13% of companies manage to achieve this objective, even for simple requests. The implementation of a virtual agent allows to reduce this gap by automating more than 60% of incoming requests.
The first step of this guide should have helped you identify which stage of your customer journey will benefit most from conversational AI. The second part will have allowed you to assess the complexity of the applications and integrations required to adopt this technology. Not all solutions are created equal, and each company should be able to choose the solution that best suits their needs. We hope that this guide will help you to quickly improve your daily life and that of your customers.
