Internet of Things, Connected Devices, and Smart Devices

With already more devices connected to the Internet than humans, over 26 billion devices are poised to be interconnected by 2020. This empowers companies with a formidable amount of streaming data to tap into and mine for intelligence on product and consumer behavior.

Companies that grow a capability to turn raw data into intelligence, are companies that will dominate the space by delivering integrated user experiences for their constituents.

Predictive asset maintenance as a new opportunity for outbound revenue generation

Using predictive analytics techniques, the data from connected devices can be used to predict the health status of a machine. When the data suggests that a device is poorly configured or about to fail, contact centers can now perform outbound calls to alert the customer of a pending failure or necessary maintenance.

Example: consider a car manufacturer that has data on driving behavior, brake usage, road conditions, weather conditions, etc. By permanently analyzing the data of each individual car, algorithms can predict if there is a need for replacing brake pads or other parts. A contact center could then decide to proactively reach out to the customer to offer a maintenance session and initiate a sales flow. This provides a great user experience for the car owner. Automotive incumbents are planning to put these models in production as of 2020.

Pervasive listening and interactive devices

Devices such as Google Home, or Amazon’s Echo come with dozens of microphones that are constantly scanning and replying to a consumer’s request. Already widely proliferated showing no signs of slowing down, these pervasive listening devices will act as a connected home HUB, enabling a consumer to simply speak, after which the device will send commands to pick up laundry, lower the temperature, order groceries or answer support questions.

Artificial intelligence

Artificial Intelligence, or A.I., covers a wide range of techniques such as machine learning and deep learning to develop algorithms that are capable of learning and performing tasks, faster and more efficiently than humans are capable of. As such, Artificial Intelligence is globally recognized as one of the most prolific industries and poised to bring the most advances in operational efficiency and automation. In the contact center industry, applications of Artificial Intelligence are often called Robotic Process Automation or Intelligent Process Automation, and are mainly used to optimize current solutions.

Big data and deep learning to introduce next-generation contact center automation

Modern companies realize that data is the biggest asset they have. Because of the enormous size and exceptionally rich type of data, contact centers are in a unique position to benefit greatly from machine learning techniques.

Contact centers have always been optimizing performance by looking at performance indicators like average handling time, first call resolution, the average time in queue, abandonment rate, customer satisfaction, net promoter score, agent turnover rate, agent absenteeism, attrition rate, and much more.

Traditionally, this analysis was done in a reactive way by looking at historical data. Deep learning offers capabilities to move to predictive and even optimal analysis.

The potential role of a contact center with regards to CRM and big data brings the realization that data is the biggest asset there is. However, data that are not being analyzed only have potential value. A contact center that builds the capability of working with this data, not only unlocks the value of this data but also reaffirms its position of a contact center within an organization as a data pipeline for customer and product intelligence.

In order to efficiently and effectively exploit this data, artificial intelligence algorithms are required. Here, we see machine and deep learning as important techniques that focus on making computers learn from data without explicit human intervention at every step, and thus making the exploitation of big data affordable and actionable for a contact center.

Controlling operational cost with inbound volume forecasting

Planning the right amount and the right mix of agents is a crucial factor in balancing cost and average handling time. Inbound calls do not like to wait in a queue. Conversely, over-capacity eats away at the already thin margins of a contact center. New advances in the field of forecasting allow for more granular forecasting.

It is now possible to not only forecast the total call volume, but also the type of calls, required agent skills, expected handling time, etc.

Forecasting algorithms are becoming more and more accurate by using recent innovations like Recurrent Neural Networks, Long-Short Term Memory, and hierarchical forecasting.

These techniques can not only be used for forecasting actual values, but also for performing ‘what-if’ scenarios. For example, you can now estimate what the impact of an event like power outage could be on your inbound call center on a specific date and time. A power outage in the middle of the night will cause fewer inbound calls than during peak consumption hours. This can be used by contact centers to make sure they meet their service level agreements with an optimal number of resources.

Natural language understanding and generation

Natural language understanding (NLU) is a subfield of artificial intelligence that focuses on processing natural language, both written and oral. Enormous lapses in the understanding of natural language have been made in recent years, enabling machines to semantically understand human written and spoken words, sentences and paragraphs. State of the art natural language understanding models can detect a user’s intention and extract information with large accuracy.

Agent decision support systems and reply suggestions

In a first phase of using Natural Language Understanding, a robotic agent is listening to the same call as a human agent. The robotic agent transcribes the call in real-time and offers suggestions to the human agent on how to proceed. Especially for agents with a low tenure, this can be useful as they will be able to learn from the digital agent on the fly.

By using machine learning models to offer real-time reply suggestions, contact centers can move away from anecdotal-driven troubleshooting (‘the last 3 times I saw this problem that was the cause’) to data-driven (‘there is a 73% probability this is the cause”). This increases first call resolution and decreases average handling time.
The same principle can be used for outbound sales calls. By analyzing the available transcription history, a machine learning model can learn which answers to objections work the best.

Instead of relying on scripts, or agent gut feeling, an optimal dialogue can be reached every single time, drastically increasing the close and retention percentage.

Real time sentiment analysis and emotional profiling

Humans are guided by emotions. Contact centers all over the world are currently experimenting with real-time sentiment analysis. Understanding emotions can offer valuable insights into how best to proceed to interact with a customer.

For example, if the sentiment algorithm picks up that the tone is becoming more agitated or even hostile, the operator can receive an alert on their dashboard and initiate a different script.

Derived insights regarding the emotional state of customers can be stored in their customer profile. In turn, this information can be used for more personalized interactions.

For outbound sales calls, knowing a person’s emotional triggers can be a powerful asset in a sales process. Voice analysis is already being used in the insurance industry as an indicator of false claims.

Voice synthesis and agent mimicry

Voice synthesis technology is the field of creating realistically sounding voices, including accents and emotions, based on real voice examples. Although already used by companies in the gaming and media industry, as well as by cybercriminals to perform CEO fraud, we expect this technology to mature and become commercially available by 2020 within contact centers.

At that point, software will be available to train and subsequently mimic the voice character of an agent. Combined with real-time reply suggestions, agents will then be able to handle multiple calls at the same time.

The role of the agent changes from having the conversation to merely approving the suggested, and synthesized responses.

Only when no suitable script is available will the agent take over and talk in free form. Since the dialogue was in the actual agent’s voice, the switch from digital agent to a human agent will be seamless and almost unnoticeable to the caller.

Continuous voice authentication

Security questions for validating a customer’s identity take between 30-60 seconds per caller and are considered annoying by the customers. Recent advances in voice recognition algorithms have made them reliable enough to authenticate users by simply speaking a full sentence.
Voice authentication algorithms work by calculating voiceprints. These are characteristics of a person’s voice that are unique to the speaker. Recent advances have made voice authentication algorithms robust even for text independent cases. In the next 1-3 years, a customer will not have to repeat an example sentence anymore, not even for setting up the voice print for the first time. Because of that, voice authentication is non-invasive and provides a secure authentication method, enabling to deliver a safe and personalized experience right from the start.

New channels

Commercial enterprises are already recognizing that integrating social media into the contact center as a core customer interaction channel is the way forward, estimating that within five years, the number of relevant social media interactions will be equal to the number of phone interactions, with 70% to 80% being service-oriented and requiring attention.

Using social network listeners, contact centers can tap into new ways of delivering great customer experiences by joining the social discussion and responding to public tweets, Facebook posts, and other social network interactions.

As technologies like Virtual Reality and Augmented Reality mature, contact centers will find their way to get closer to the customer and adapt these.

Technology maturity

To assess the technology maturity and readiness, we have analyzed the above-mentioned technologies and scored them in terms of maturity and value.

A clear emphasis can be seen on how natural language-based solutions are gaining traction fast, while already having reached a maturity level that allows companies to start using this technology for automating the handling of up to 80% of recurrent questions at a contact center.

We can clearly witness how machine learning and natural language technologies will continue to play a pivotal role in the evolution of the traditional call center into a data driven contact center.