From reactive to proactive: 7 ways to use AI in sales and call centers

While Elon Musk and other experts have called for a pause in AI development, businesses are showing no signs of stopping. According to Statista, the AI software market will reach $126 billion by 2025, and 57% of companies already use AI to improve the services they offer.

Using this technology to handle calls is more accessible than ever. The idea was first raised years ago, but back then models needed to be trained on at least 20,000 calls to get useful results. Moreover, the technology required constant retraining. These days you can already get valuable insights after training on only a few hundred calls. This is great motivation for adopting AI to take your service to a whole new level.

In this article, we will consider what makes AI so attractive for the customer service industry, and in particular for companies that communicate with customers by phone. What opportunities are already available, what can we expect in the near future, and where is this technology headed?

1. The needle in a haystack: finding situations that require call QA attention

The duties of call QA staff are changing significantly. Call QA employees typically carry out mechanical tasks: listening to calls, searching for mistakes made during conversations, and providing feedback. Supervisors often do similar work.

A business in the US can expect to pay a supervisor about $17.53 per hour for this work. However, these specialists spend part of their workday listening in vain. It takes dozens of hours to analyze hundreds of conversations and find, say, the 20 that really need attention.

AI can do this task just as well as a human — or even better. After all, its attention doesn’t wander, and the numbers prove that it makes significantly fewer mistakes. There are already AI tools in operation 24/7, analyzing calls and looking for what are essentially needles in a haystack.

Many companies work with clients who speak different languages, requiring new foreign-language representatives to join the sales team every time a business enters a new market. All of these supervisors need their work closely monitored. Hiring individual supervisors for each language ends up being quite costly.

How will AI be useful here?

1. It transcribes the audio of conversations into text. It also automatically translates the text into English if the conversation is in another language. This makes it easy to monitor a sales team or call center operating in other markets. 
2. It records the key points of the dialog so supervisors or managers can get a 10-15 second summary of what was discussed.
3. It indicates whether a human review of the dialog is needed and notes why. For example, it will alert you if an employee needs training on working with customer objections or emphasizing the advantages of a product or service.

Equipped with the knowledge of which conversations require attention, a supervisor will be able to spend less time on repetitive, routine tasks. Or maybe their duties will be reduced so significantly that they can be combined with those of another employee. After all, listening to 20 conversations takes far less time than listening to 300.

2. Providing feedback to teams to improve the quality of service

Modern AI can detect mistakes made by sales reps or operators when communicating with customers. After training the AI, it “understands” employees’ weaknesses in conversations. For example, it can detect if a company representative:

• is not proactive enough or not focused on the client’s tasks;
• is not educated enough about the product and frequently asks colleagues questions;
• does not try to answer the customer’s questions in detail, but instead advises them to read the information online;
• is too technical when describing the product, which could frustrate or confuse the customer.

This enables supervisors to review errors that occurred during conversations and analyze them with the employee. The downside here is that this information is only available after the customer has hung up. The sales rep can’t influence the impression they have already made on the client.

The AI of the future will be able to give this advice during conversations. This will help reps adjust the course of the dialog so the client hangs up as satisfied as possible and the deal closes as quickly as possible. Maybe AI will show a pop-up message during a call that advises on:

• what to offer the client at the current moment;
• what to improve during the conversation, such as adjusting tone to be more friendly and trusting;
• offer a discount if the client is ready to buy and final doubts must be dispelled.

Call tracking will be useful for the first item above. In particular, it allows you to understand the customer’s path on the website before the call. The system knows which pages the user visited before the call and on previous visits to the site. This allows AI to better understand what exactly the customer is interested in — even if they don’t mention it during the conversation. The buyer will be pleasantly surprised if the sales rep offers them exactly what they had in mind.

3. Predicting the likelihood of a purchase and focusing on more promising deals

Of course, all the leads in your CRM are important. But the effort you need to expend can vary greatly from deal to deal. To make more efficient use of available resources — which are almost always limited — you should focus on those that are closest to being closed. There are always certain deals that can bring in profit immediately, while others require significantly more effort and may not necessarily lead to success.

Here, again, AI can help optimize time and increase sales team profitability. It can predict the probability of closing a deal using detailed information on calls or chat messages. When making its predictions, AI considers many different signals, such as:

• whether the client was sufficiently interested during the conversation;
• whether the dialog contained words related to the lower stages of the sales funnel:  payment, signing documents, visiting the showroom, etc;
• whether the potential buyer expressed admiration for the product or responded positively to its description;
• the general mood of the conversation, as well as the mood of the employee and the client.

Based on the AI forecast, the sales rep will know which deals to focus on. This will be especially useful for small companies with many customer requests.

4. Writing follow-ups and tips for further action

No conversation with a customer should end without outlining next steps. AI will save you time by making sure you don’t need to write those next steps down after a conversation, when you might actually have another call you should be on. You also won’t have to think back to prior conversations, or listen to recordings, to remember what was discussed.

AI can suggest the best next steps by analyzing conversations. For example, it will remind you to send a detailed product description, issue an invoice, show a test sample of a product, etc. The sales rep will only have to check the text produced, edit it if necessary, and send it. AI will also be useful for new employees who don’t immediately know what to do after communicating with customers.

5. Smart, personalized call forwarding

Research shows that 73% of customers expect companies to understand their unique needs and expectations. One of these expectations is to be immediately connected to an operator who can solve a specific need or has perfect knowledge of a particular product. This isn’t so easy when your company has an online store with thousands of items and dozens of call center operators, and there’s no guarantee that the customer will be satisfied with the level of service they happen to receive.

How could AI handle this task? Let’s imagine an online store with different category products that require clarification by phone.

1. The customer calls after browsing the website. Based on the user’s path, AI concludes that the customer is interested in refrigerators. 
2. AI instantly looks through dialogs with the word “refrigerator” in them, as well as conversations from customers with the desired intent. Then it selects conversations where the operator got a high rating. Maybe it will even consider where there was a sale, if AI is integrated with CRM.
3. The tool selects the employees that participated in these conversations and checks whether they are currently available.
4. The call is directed to the employee who will most likely provide the best advice and close the deal. After all, they likely know best which product the customer actually wants. 
5. Every time a customer calls, AI will instantly create a virtual queue of operators who should receive the call. It will do this based on the user’s behavior on the website, their previous history, and the history of each of the company’s representatives when dealing with requests similar to the customer’s.

A similar principle can be applied if some employees are better at selling premium products and others excel at selling more budget-friendly options. It is important to understand that such scenarios will only work if your company has a long history of communication with customers. To be clear, we’re talking about training AI on thousands of conversations.

6. Automatic callback

AI will be useful for companies where customers make appointments, such as beauty salons, medical service providers, and service stations. In some cases, appointments can be made weeks or even months in advance. This means that clients sometimes simply forget about scheduled visits. When a customer doesn’t show up at the scheduled time, there’s an unplanned opening that could be filled by someone else. In many industries, representatives call customers to remind them to show up at the agreed time, to reduce missed appointments.

In the future, it will be possible to automate this process by linking AI and telephony. 

1. AI will retrieve a list of customers with the appropriate status from the CRM and call them to confirm the upcoming visit. 
2. The information received will automatically be transferred to the CRM, and canceled appointments will be removed. 
3. New openings can be used for clients with appointments several weeks out. This increases the business’s payload, as well as customer satisfaction.

Here’s another option for companies that ask for customer feedback. Virtual telephony can take data from the CRM and prompt customers with a recording such as: “Press 1 to 10 to rate how satisfied you are.” But summing up an entire experience in one number doesn’t give enough information. Ask customers to tell you in detail about their experience. AI will then convert their response to text and compile a summary of what they liked, what they didn’t like, and why.

7. Getting data to increase advertising ROI

Training advertising algorithms based on call quality data

Google Ads, Meta, and Bing’s algorithms have long been able to learn from call data. If they learn that there were calls from a certain campaign, the system will raise the associated bids and start showing those ads more often, since calls are a sign that advertising is attracting customers.

Algorithms can be trained much better with data collected by AI. After all, not every call is from a potential customer. The next step is to train based on targeted calls, or calls of a certain length with customers. But even this parameter doesn’t mean that the advertisement led to a customer who is likely to buy.

Some companies may tag quality leads and submit them to Google Ads, but unfortunately this is manual work. Employees have to enter specific tags after each conversation, such as “consultation,” “purchase,” or “spam,” which is a time-consuming process. Few businesses use tagging.

In the future, artificial intelligence will help here as well, by determining metrics such as probability of purchase based on features of the call. Similarly, conversions can be transferred to Google Analytics and Google Ads. This will allow AI to provide more accurate signals for the algorithm and automated bidding strategies so that they learn more effectively. This will let you optimize your campaigns to get more profit from them while spending less on promotion by shutting down ads that do not generate quality calls.

In Google Analytics, you can also build a report on the funnel that each lead goes through. A simplified example would be: transition from ad to website → call → call with a high probability of sale → sale. If you create a report like this, you can easily analyze which campaigns generate the most converting calls.

Gaining insights for advertising campaigns

AI can also provide other valuable insights for marketing. For example, it can analyze callers’ first words — how potential buyers formulate their requests and what non-obvious words they use to describe a service or product. There’s always a difference between how a user Googles something, what the keyword in the advertisement is, and what a potential customer calls something in real life.

By doing this, AI can collect non-obvious keywords that marketers can later use in contextual advertising. Without AI, this would require additional research and dozens of hours spent listening to call recordings.

Another advantage of this method is that it provides a more successful entry into foreign markets. After all, a brand might not always understand what foreigners call a certain product, even when using a translator. AI will suggest more common synonyms and slang phrases that can be added to ads and even a website. 

We have listed only seven possibilities related to calls. However, AI opens up many more, such as smarter chatbots. These assistants will no longer be limited to pre-prepared phrases but will be able to fully engage in a dialog.

AI-powered chatbots can also be used for initial call processing. For example, they can immediately answer simple questions that customers ask often. In addition, chatbots can be used instead of IVRs.

Here we should mention Google’s Dialogflow CX, which allows you to create smart voice menus. They can understand customer requests and intentions, and provide useful answers. If a potential customer’s request requires a call to be forwarded to a specific department, the tool uses routing rules. If a customer doesn’t know which department can solve their problem, AI can use clues from how their question was formulated to figure it out. This can even happen automatically after a preliminary conversation with the client.

Conclusion: Can we say goodbye to the old ways of working?

Should we still be afraid of the quick development of artificial intelligence, as Musk and other experts suggest? I don’t think so. True, for many people, this is the end of their career as they know it. For example, according to a Goldman Sachs report, AI could replace the equivalent of 300 million full-time jobs. But it also states that it can create new jobs.

Some studies predict that a true dream will be realized — the four-day work week. Google Cloud experts believe that AI will increase productivity so much that by 2025 this dream could be a reality. It seems like we can hope for more free time and less burnout in the future.

In addition, there’s hope that many professions will be transformed, not wiped out. Someone who used to do a certain process manually will now train or monitor AI. Even if call center staff is reduced by 90%, 10% of the most experienced operators will need to remain. After all, an electronic consultant can easily answer common questions, but there will always be non-standard situations that require human intervention.

Some tasks cannot yet be entrusted to hardware, such as closing complex deals with large clients, which are always preceded by long negotiations. In these cases, buyers want to communicate with a specific person and buy from them. It’s the same when a client is planning a large purchase they may only make a few times in their life. Here, a “live” seller is what sends a message of reliability. Human charisma and the gift of persuasion are always a plus if the client has any doubts.