Routine, statistics, analytics: what an AI assistant can help a technical support or hotline manager with

5.9.2024
Routine, statistics, analytics: what an AI assistant can help a technical support or hotline manager with

Content

97% of business owners believe that artificial intelligence will help their businesses (source→). But only 35% of companies have integrated AI tools into their processes! And this is a global figure: in Russia, the gap between “we consider it useful” and “we really use it” is even wider.

In this article, KT.team experts will explain how an AI assistant can help the head of a service company or service department reduce routine and get analytics for management decisions faster.

Another beautiful day

Let's imagine a typical Monday for a tech support manager.

He has a week already planned and will have to:

  • make a decision on 10 employees on probation: who to leave, who to fire;
  • make a decision on several current employees: who to raise their salaries, who to remove from customer calls and make a senior manager, and who to train;
  • make a decision on employees who are on his pencil due to poor feedback from clients and colleagues;
  • analyze frequently repeated requests together with trusted employees — how you can change scripts and standards on your side to reduce repetition of errors;
  • train the customer's employees to use the new system;
  • try to prevent the weekly window of chaos from growing to the size of a portal to hell.

Half or even more of these tasks are related to call analytics: how employees behave when communicating with customers, how they comply with regulations, and how current regulations are generally relevant. To get all this data, you either have to listen to calls in person or delegate this task to a trusted (and often the highest-paid) subordinate.

But maybe it's different?

Do you still listen to all calls in person? Then AI is coming to you

According to Deloitte The Technology, Media & Telecommunications (TMT) AI Dossier report→, an AI assistant is one of the five most promising developments for companies in the TMT sector. 54% → TMT organizations have achieved a return on investment of more than 20% thanks to investments in AI. This includes solutions that can structure and analyze call information.

Полезный материал по внедрению ИИ-ассистента

7 возможных ошибок в безопасности и как их избежать.

Scripts and regulations are followed “to the decimal point” on every call. And if they are not complied with, you know exactly who, how many times and with which client departed from the standards

The usual call center of a B2C company. Every day, he receives hundreds and thousands of calls from customers with questions and complaints. A protocol is provided for each scenario: how to answer, where to look for information, and how to refrain from emotional statements. And also: what and how to fill in the customer card, product (if there is a product complaint), employee...

But are you 100% sure that all rules are being followed? There are two ways to verify this: listen to the call and compare it to protocols, and choose “senior managers” who are almost exclusively responsible for supervising colleagues.

It is almost impossible to learn about all the jambs and shortcomings in this mode. A manager's time is not a tough one. To evaluate all calls, you need to hire as many experts as you employ line workers, which is not profitable. The most you can afford is a small random sample, which may not include the most problematic calls or employees.

In the meantime, every wrong answer or emotional explosion is a threat to the company's reputation and metrics. If a manager makes such mistakes on a regular basis, it's best to find out as early as possible.

An assistant who evaluates calls objectively and provides you with department statistics

Now imagine that you get data on every call from every manager. Moreover, this data was analyzed by an unbiased expert who easily compares calls to existing regulations, assesses the accuracy of compliance, and makes cuts by project or manager.

To make decisions, you just need to look at a simple table like this:

ИИ-ассистент оценивает звонки объективно и обеспечивает вас статистикой по отделу | KT.Team

And you'll see that half of the time, managers come up with solutions that don't match the script. Why is this happening: because the script is out of date or because managers don't know it well? You can find out this with reliable analytics data.

The manager himself will receive emails with a detailed analysis of his calls for compliance with regulations — and with recommendations on what needs to be improved!

What will be the outcome? Within a month, the quality of compliance with regulations will improve by 50% -95%, and the quality of the regulations themselves will also improve, because now you will receive information about their obsolescence faster.

More data for HR decisions

Underestimation, like overvalued employees, are twin problems where you lose money. But how do you get objective evidence that someone is not up to the required level, and someone is really already overqualified for their place?

Ask his colleagues, ask his senior manager, and yes, again! — listen to calls where this employee communicates with the client. You can't fully rely on the opinion of subordinates: human evaluation always depends on personal relationships. All that remains is personal involvement.

It's good if you have to make a decision one person a month. What if it's 10? 30?

A convenient visualization format and a variety of criteria for evaluating incoming information

Aggregated and visualized employee data

Now imagine that someone has already listened to all the employee's calls instead of you, transcribed them and evaluated them according to the parameters that are important to you.

You see a detailed table in front of you that you can sort by call type and date. And see, without even reading it, how an employee's assessment changes over time: whether their performance has improved or stayed at the same level, whether they make mistakes, whether they are polite to customers, and whether they respond quickly to complex requests.

15 seconds for any answer to the customer's relationship history

Imagine that a client you've been working with for a long time refers to a problem that happened a year or two ago as part of another call. He doesn't even remember when it happened. And your manager has to go through the entire history of recent years to understand what this problem was, how it was solved, and how it could affect the current situation.

Meanwhile, the client is nervous, time goes by, other tickets are being pushed back...

Get an answer in seconds

All you have to do is change the instrument and the situation will also change. Instead of searching for CRM or another system you're working with, your manager writes to the chatbot “give a link to a ticket based on the client's so-and-so situation”. In 15 seconds, he'll get a link, a summary of the problem, a solution, a link to a recording of the conversation, and a screencast... Everything he needs.

What is even more pleasant is that it is not necessary to reproduce the exact language that might have been used when describing the problem. An AI assistant understands both professional slang and human speech.

Scale without increasing staff

Let's take a closer look at what managers' time is spent on a standard 8-hour shift. Four hours to communicate directly with customers and solve their problems, and three hours to write brief summaries of meetings, minutes and memos. For another hour, he will look through the history of customer requests to understand whether CRM has “painful” patterns and what they tried to do about it earlier. In addition, it will be necessary to understand the scope of the problems by looking for historical information.

Routine tasks require recharging, so add coffee breaks and mindfulness meditation to this time if the problem is serious and you need to manage stress. A joke that has some truth to it.

Half of tech support workers report burnout at work, according to a study aboutMicrosoft 2023 Work Trend Index report→. BUT 66% say→that they don't have time to complete tasks within a working day.

With a staff of 24, a company can support, say, 12-15 clients. I would like more, but that would require adding one person to each of the three shifts.

Less routine means easier scaling

Any customer support call must end in a protocol. The question is how much time the manager will spend doing this and what will be the accuracy of this protocol.

Delegating this routine function to AI is a logical step. An AI-enabled assistant will transcribe the meeting, write a summary for the client's card, and won't forget about subtle (but sometimes very important) details. The manager will only have to slightly edit the wording and supplement the protocol with actions to solve the problem.

This approach will free up approximately 1-2 hours a day for each employee. It seems not much, but it takes 1-2 hours for more complex and important tasks. And an additional 12.5-25% of resources to scale the entire business.

How much does it cost to integrate an AI assistant into technical support or hotline processes

It depends on the goals you set for the future AI assistant.

If you want to “just decrypt calls”, you basically don't need an assistant — just buy a subscription to any of the millions of decryptors. This will cost you from $10 per month, plus a minor change in the work processes of each manager — each entry will have to be uploaded to the transcriber interface and the text removed from it.

But if you don't just want to receive additional text files, but also:

  • receive feedback on calls;
  • it is easy to analyze the quality of managers' work;
  • spend a minimum of time and effort on training new employees;
  • automatically add transcripts and protocols to customer cards;
  • receive any answers to any call in 15-30 seconds;
  • increase customer satisfaction,
  • and also take care of the confidentiality of the information that customers trust you when calling —

We are already talking about implementing an AI assistant, which, in addition to a decryptor, includes integrations, modules for anonymizing and encrypting information, a chatbot, and an analytics module. And here, of course, the order of numbers will be different. For example, at KT.team, we offer the introduction of such an AI assistant from 1 million rubles, for a period of 4 weeks.

You can read more about the product on our website→.

Go to other articles about AI implementation

Read
Другие статьи

Смотреть все

Popular PIM systems: functionality, features, cost

30/7/2021

Подробнее

Bitrix24 vs Salesforce: implementation features

22/2/2023

Подробнее

How to properly implement an ESB layer on the first try

11/5/2023

Подробнее

Смотреть все

We use cookies to provide the best site experience

Ok
We need advice on the implementation of AI