Large amounts of data from different channels accumulate in companies of every size everywhere every day. Analytics tools are helping to harvest valuable information from this data. By means of artificial intelligence, huge amounts of data can be assessed, and patterns and contexts identified a lot faster than human labor could. But what is the potential of artificial intelligence in analytics? Peter Schmitt, CTO at ASC, explains the advantages for contact centers, financial service providers, and public-safety organizations based on different usage scenarios.
Artificial intelligence (AI) is one of the most far-reaching technological revolutions of our time. The possible applications of the technology are almost infinite and already now adaptive algorithms are used in many areas of our everyday life. For example, virtual assistants like Siri, Cortana or Alexa understand our questions and voice commands by deploying artificial intelligence. On the Internet, bots and AI help to update and monitor web pages or process messages in an entirely automated way, as e. g. in Microsoft MSN. Intelligent chatbots now even assist in communicating with customers in a chat. In industrial enterprises, service companies, and even in public authorities, AI systems are a common sight. They perform repetitive tasks, analyze corporate communications or do the accounting. Insurance companies, banks, and even the revenue service have started to deploy intelligent programs to check the submitted forms or make the decisions whether to effect payments.
As a consequence, it does not come as a big surprise that in an increasing number of sectors artificial intelligence is considered a key technology for the near future which is expected to bring far-reaching transformations in the coming years. According to a study on machine learning (ML) carried out by the research company IDG, about 30 percent of German companies intend to deal intensively with artificial intelligence and machine learning in the coming years. And already now, 57 percent of companies in Germany already use at least one ML technology to automatically generate knowledge from data.
The major fields of application are speech recognition and text analysis as well as big data analytics
. As digitization promotes the accumulation of increasing amounts of data from different channels, viewing, organizing, analyzing, evaluating and draw conclusions from this data by hand is positively out of the question.
Advantages of artificial intelligence in analytics
In addition to being able to handle large amounts of data, AI-supported systems integrated into analytics applications may also significantly contribute to ensuring data quality and selecting relevant information from an otherwise overwhelming flood of digital data. AI-supported analytics solutions provide much more information processing capacity than any human ever would. They create a reliable foundation for decision-making – almost in real time – that human intuition and action can build upon.
Today, the AI behind data analytics usually is an explorative data analysis. It reveals high-risk patterns and thus supports human actions; it helps to quickly navigate through data in a target-oriented way, to speed up work processes, and eventually to act and react faster. On top of that, AI enables new processes, strategies, or services unprecedented not so long ago. It is not seldom that new approaches of collaboration between humans and machine are developed in which the AI enhances human capabilities and thus allows for new and sometimes improved ways of working.
Some of the advantages of artificial intelligence in the field of analytics are:
Shortening work processes
Thanks to advanced language processing, statistical correlations can be investigated, data can be grouped and attributed to one or several categories. That way, companies can speed up repetitive work processes and use the now undeployed resources for other tasks.
Comparing content quickly
When comparing content, the possibilities of language processing go far beyond those of a common search engine. Since the speech converted into text is also comprehended, even substantially rephrased passages with matching content can be detected.
Fast and target-oriented analysis
By automating the analysis of documents using natural language processing, companies can access information that is especially relevant to them very quickly and in a target-oriented way.
Automatically structuring previously unstructured data
AI-driven automation analyzes structured and unstructured data resulting in an advanced understanding of the customers’ needs, of market conditions, and of business opportunities.
Improving customer relations
Artificial intelligence makes it much easier for companies to automatically collect customer opinions from different sources and to extract information. The results help to implement customer feedback more quickly and thus improve customer relations.
Use cases demonstrating the advantages of AI-driven analyses come from a wide range of industries and economic sectors. Usually, the companies hope to reveal weak points in their corporate communications to be able to take measures increasing efficiency and optimizing processes. Due to the accumulating amount of data and its ready availability by means of cloud solutions, for instance, using this data adds tremendous value and is an ideal way to better understand and as a result evaluate and control business processes in a target-oriented way. Analyzing communication is a must, especially for companies with a high volume of calls.
Use Case Contact Center
The number of customer calls in contact centers is growing steadily. Evaluating and assessing this data manually is a physical impossibility going far beyond any financial means or time resources. However, this data promises valuable information and can provide important insights into customers’ behavior. Analytics tools in combination with artificial intelligence help to reveal previously hidden information in customer communications – automatically, close to real-time, and at an affordable price. For contact centers, this opens up a wide spectrum of possibilities:
With analytics tools, effective approaches to improve service quality
in contact centers are just a click away. Furthermore, contact centers may gain new stepping-stones to leverage measures for continuous improvement.
• Employee qualification
Calls can be selected automatically to be used for training purposes. The evaluation of the collected information leads to improvements in employee training. In addition, time-consuming preparations for coachings become obsolete.
• Taboo word spotting
It is possible to create blocklists to identify and avoid banned words. On top of that, calls may be searched for specific words and evaluated how often and by which agent they were used.
• Reports on current service quality
Automated evaluation and categorization of all communication allows determining the quality level. The results may be displayed in graphical reports and dashboards.
• Quality monitoring and alerts
Keywords indicating quality issues can be defined and when spotted in a call the management can be informed. An immediate reaction and individualized training measures for agents can prevent repetitions in the future.
Analytics solutions designed specifically for contact centers have a major impact on key processes that can increase efficiency as well as on the flip side affect it adversely.
• Reduction of call volume
Analytics solutions support contact centers in identifying and resolving the issues leading to repeated calls revolving around the same question or problem. This results in an improved first call resolution (FCR) rate. The FCR rate indicates the percentage of cases in which a customer's query could be resolved upon the first call.
• Compliance with defined processes
By analyzing all calls, the need for process optimization as well as the degree of process compliance by the agents can be determined.
• Efficient search
Relevant calls can be retrieved and replayed within seconds. Manual search is a thing of the past. Contact centers can search for calls containing specific keywords and see how often they were used.
Transcription tools convert speech to text for documentation and archiving purposes. The transcribed calls can be made available to the customer as a text file. That way, all involved parties can get an overview of the conversation content if required.
Boosting customer experience
By analyzing interactions across all channels, contact centers can get a consistent, uninterrupted impression of the customers’ experience and reveal the leverage points for further improvement.
• Customer feedback
Analyzing data from customer surveys provides insightful feedback for more personalized customer service. On top of that, it offers the possibility of searching for calls that already contain customer feedback.
• Identifying trends
To identify trends and current hot topics at an early stage, calls can be categorized at the click of a button and thus responded to immediately.
• Avoiding customer churn
By means of an early identification of issues that affect customer loyalty, contact centers can prevent customer churn. Calls can be monitored and in the event of problems, alerts are issued immediately to be able to take countermeasures without delay.
• Emotion detection
Analysis tools detect emotionally charged conversations based on verbal and acoustic characteristics and can be configured to send respective notifications.
The insights gained from the analyzed data help to make reaching out to customers more effective and to increase sales and revenue.
• Recognizing competitive advantages
Analyzing customer calls may provide valuable information about the business behavior of competitors. And in the same vein, analyzing a competitor may help to accelerate the process of marketing one’s own products. Even information that the agent missed during a call are not lost for good but can still be used to the company’s advantage thanks to later analysis.
• Identifying sales opportunities
Analytics solutions shed a light on customer acceptance as well as on demand for specific products and identify effective arguments when reaching out to customers. Furthermore, contact centers learn about new trends and the latest hot topics discussed in a call. This contributes to successfully developing new sales strategies.
• Avoiding sales risks
Escalations can be revealed precisely by analyzing customer experience especially with regard to dissatisfied and upset customers.
• Revealing current trends and hot topics
Recurring topics can be treated in FAQs for customers to try to resolve minor issues themselves before calling the hotline.
Use Case Financial Services
Regulations such as MiFID II
have a significant impact on corporate and customer communications in the financial industry. Among other things, they stipulate the uninterrupted recording and archiving of all calls dealing with financial and investment advice. This is where the use of analytics tools in combination with artificial intelligence offers great potential for financial service providers.
• Meeting compliance requirements
Reliably detecting potential fraud and compliance violations is an important prerequisite to be able to take immediate countermeasures to protect a company. Analyzing communication data supports financial service providers in meeting compliance requirements.
• Checking for compliance statements
It is possible to verify whether calls contain a mandatory compliance statement. Have the compliance requirements been fulfilled and has the customer been provided with all relevant information during the call? Do the calls include risky topics or violate the compliance specifications? Calls which lack the compliance statement are thus documented in conformity with legal demands.
• Automated categorization
The entire communication is automatically categorized according to its relevance for compliance. This is followed by automated archiving with retention periods that can be defined individually depending on the category.
• Documenting consultant calls
The transcription of calls enables turning audio content from consultant calls to text at the click of a button. For one thing, this is an apt means to preserve evidence to be submitted at a later moment in case of litigations; for another, it leads to reduced post-processing time since handwritten documentation is no longer required.
• Improved risk management
All calls are automatically analyzed for atypical behavior of customers or agents. If a potential risk is identified, financial service providers can react immediately and avert any threats of litigation.
• Fraud detection
Suspicious interactions regarding possible breaches of compliance regulations or insider trading are automatically tagged and reported to the management or the compliance team at an early stage. That way, companies are able to recognize risks and non-compliant processes in time and initiate appropriate measures. Fines and sanctions can thus easily be avoided.
Use Case Public Safety
In recent years, the number of calls reaching public safety control centers has risen notably. Emergency calls as well as threat calls cause an increased workload for employees at control centers. It is crucial that emergency communication remains efficient to be able to continue to guarantee assistance without unnecessary delay. At the same time, the entire communication process must be traceable in case of subsequent investigations
. Highly efficient analytics solutions can help to significantly improve these processes.
• Process improvement in emergency situations
Analytics solutions support event analysis and process rationalization. A structured categorization of calls based on incidents and the evaluation of the entire process leads to process improvement in emergency situations.
• Facilitating work for employees
Analytics solutions minimize the time required to document tasks by automatically transcribing recorded calls, creating protocols, and completing forms thus speeding up the entire process chain. This significantly reduces the overall processing time.
• Warning notifications in case of threat calls
As a result of categorizing calls based on previously defined keywords and phrases, the solution issues real-time warning notifications in case of crises, threat calls or specific topics.
• Quick prioritization
Recorded calls are categorized according to urgency and prioritized based on predefined keywords and phrases. This allows identifying, evaluating, and reconstructing crises and critical topics quickly. In case of emergencies, real-time alerts are issued.
AI assistance systems/voice-controlled systems of the future – an outlook by Prof. Dr.-Ing. Mohammed Krini
Voice control is part and parcel of digital life these days: Whether in onboard navigation system of cars or as an intelligent personal assistant on the smartphone – voice control is a handy functionality. Digital voice-controlled assistants at home and at work will soon use AI to make daily decisions easier and improve their quality. Intelligent AI systems are predicted to adapt even better to the needs of users and the relationship between humans and machine will become even more intense and efficient thanks to AI in the future. AI assistance systems will soon be able to detect even slight changes in facial expressions, to react to different pitches in the voice and to understand the content of calls by means of microphones, sensors, and cameras.
Companies that record their communications – be it for compliance, training, or business purposes – have access to an extensive wealth of data. This data represents an enormous added value for any company as it provides the raw material for valuable insights. However, the sheer volume and complexity of the data often makes it impossible to test and question any conceivable and unconceivable combination of factors. Today’s challenge is to make use of this valuable data quickly and intelligently. This is where artificial intelligence and analytics come in.
The opportunities that artificial intelligence and analytics offer are multiplying rapidly thanks to new concepts and algorithms. Current achievements in the field of machine learning and analytical methods indicate enormous progress. The development and application of AI systems are made easier and accelerated more and more – to the benefit of experts and end users alike.
Definitions and distinctions
Terms like data analytics, big data, machine learning or deep learning are omnipresent in the field of AI. But what exactly distinguishes them from one another? And what do natural language processing (NLP) or automatic speech recognition (ASR) mean? In the following, the most important terms related to AI are explained in detail.
The term big data describes large amounts of unstructured and semi-structured data that companies produce every day. Loading this data into a relational database for analysis consumes a lot of time. Data sources are either too large, too fast or insufficiently annotated, so that classic database technologies and analysis methods fall short of the task. To comprehensively analyze such data, methods from the field of artificial intelligence, among others, are frequently used. Big data is often associated with cloud computing, as the real-time analysis of large amounts of data requires a lot of storage capacity and computing power.
Data analytics use mathematical methods with the objective of extracting precise information from large amounts of data and thus to better understand the process that produced the data. Data analytics combines deep knowledge of the respective field of application (domain knowledge) with an analysis specific to each data set. The evaluation of the results is done manually by human experts. The findings of the data analysis support the creation of an AI.
Artificial intelligence is the umbrella term for applications in which machines perform intelligence services similar to those of humans. The basic idea is to use machines to approximate important functions of the human brain. This includes learning, judging, and finding solutions to problems. Artificial intelligence has gained importance in recent years which is partly due to the easy availability of big data. For example, AI is able to recognize patterns in data more efficiently than humans could thus allowing companies to stay on top of the flood of data as well as to gain a target-oriented insight.
Machine learning is a subsector of artificial intelligence and refers to the artificial generation of knowledge from experience. For this purpose, classical algorithms are trained using a variety of data from different sources. These sophisticated algorithms can recognize patterns in unstructured data such as images, text, or spoken language and make decisions independently based on these patterns. This form of learning has opened up uncountable new perspectives – in natural language processing, for instance.
Natural Language Processing
Natural language processing revolves around processing text and natural human language. NLP uses different methods and learnings from linguistics and combines them with modern computer science and artificial intelligence. The aim is to create the most extensive communication possible between humans and computers via speech. NLP is used in automatic text translation, mood analysis, and automatic speech recognition (ASR).
Automatic Speech Recognition
ASR is a method for automated interpretation of human speech. A computer-based system analyzes, classifies, and stores the entered speech information using automatic speech recognition. Automatic speech recognition may include the recognition of speech, of keywords and sentences along with their meaning as well as the identification of a speaker for security-relevant functions such as access authorization. The stored speech information is correlated with spoken words, with their meaning, and with the characteristic features of a speaker.
The deep learning algorithm enables machines to learn beyond the available data. The machine recognizes structures, is able to evaluate them and use them to build knowledge in repeated forward and backward processes. The more data a machine receives, the greater its ability to learn and the more "intelligent" it can become as a result. Problems are assessed and analyzed in several consecutive layers to obtain optimum results. Each layer goes a bit deeper (hence "deep learning") into the problem.
Cognitive computing uses technologies of artificial intelligence to simulate human thought processes. The goal is to develop own solutions and strategies based on experience. The systems interact with their environment in real time and are able to process large amounts of data. With the help of AI and cognitive computing, the ultimate goal is to create a machine that interprets images and language, simulates human thought processes, and then responds conclusively.
Author: Peter Schmitt
Peter Schmitt has been working for ASC Technologies since 2008 and heads the departments PreSales, Project Management, DevOps, and Manuals and Documentation in his position as Chief Technology Officer. His expertise extends across solution design for complex, large-scale projects, software development, relational databases, open standards as well as applications for telephone systems, digital video, and access control. Since July 2020 he has been responsible for Product Management at ASC, too.