Potential applications of AI and thinking machines - not just in consulting

What director Spike Jonze tells us about artificial intelligence in his 2013 film "Her" seems a little crazy: Theodore Twombly, the main character of the movie, falls in love with Samantha, an operating system on his computer, with whom he communicates via headset and video camera. Samantha is an artificial intelligence. She learns through social interaction with Theodore and her behavior becomes more and more human. In the end, Theodore falls in love with the artificial intelligence and even begins to develop an intimate relationship with Samantha.

Of course, we're talking about a movie that doesn't claim to correspond to reality in every respect. But how far away are we from such a scenario in 2019, six months after a man in Japan took a hologram as his wife? How intelligent is artificial intelligence today and how well can it communicate with us?

 

What is artificial intelligence anyway?

To answer these questions, we first need to clarify the issue: What is artificial intelligence anyway? This is not easy, as the term artificial intelligence, or AI for short, is not uniformly defined - partly because AI research has been interdisciplinary since its beginnings in the 1950s. However, the following definition has proven to be useful for practical applications: "Artificial intelligence is the ability of an IT system to exhibit human-like, intelligent behavior."

Many systems that we consider intelligent do not rely on artificial intelligence at all. Although dictation functions and voice control are helpful for us, they do not reflect an intelligent understanding of language. Instead, they work with preset keywords that trigger certain functions, such as "set alarm clock to 7 o'clock" or "light on".

In more detail, the term artificial intelligence describes computer applications that aim to demonstrate intelligent behavior. This requires the following four core capabilities:

• perceive,
• understand,
• act and
• learn.

They extend the basic principle of "input, processing, output" of all IT systems. Artificial intelligence should support people in achieving their goals - not make them superfluous. However, the aforementioned core capabilities are necessary in order to provide optimum support for humans. The really new thing about today's AI systems is learning and the associated understanding.

 

Learning machines and systems - for humans?

Modern systems can be trained in the processing component to achieve better and better results - usually better than with conventional methods. The latter are essentially based on rigid, clearly defined and pre-programmed sets of rules (if..., then...). Examples of this are image and speech recognition. AI systems, on the other hand, not only recognize letters in an image, they also know what the word "complaint" means in a scanned letter and can initiate a complaint process. And they are often much more powerful than humans - for example, when it comes to searching millions of websites and providing all images that show a dog. That would be impossible for humans.

Modern AI systems are currently being used to supplement clear system control. The special thing about them is that they learn during the test phase and during operation based on their errors and feedback.

 

Machines learn in a similar way to humans

In principle, machines learn in a similar way to humans. For example, a computer program can learn to recognize certain objects. To do this, it is first fed with data and trained. For example, it is told which object is a horse and which is not. The program then receives regular feedback from the programmer as to whether it made the correct distinction between "horse" and "not a horse". The algorithm uses this feedback to improve itself until it finally makes the distinction between "horse" and "not a horse".

Machine learning systems usually consist of three components:

1. from a model that makes predictions and identifications,
2. parameters, i.e. signals or factors that are used by the system to make decisions, and
3. from the learning system.

The learning system adjusts the parameters and therefore also the model by looking at the differences between the prediction and the actual result.

At the beginning of the model, a forecast is often made that applies to a specific situation. Initially, the results often deviate from the forecast. The system therefore has to learn. To do this, it continuously checks the data fed in and learns from it. To this end, mathematical algorithms are used to adjust the original assumptions and thus continually optimize the model.

 

AI in (service) companies

Service companies such as banks and insurance companies in particular are already investing a lot of time and money in artificial intelligence. For example, they are focusing on AI disciplines such as robotic process automation (RPA), knowledge management software, digital assistants and predictive analytics. They see the future benefits of artificial intelligence primarily in contact with customers. With the help of AI, products and customer contact should be more individualized. Today, however, customers are still mostly served by personal customer advisors; AI only has a support function, so to speak - also because it is usually alienating for people to talk to a computer voice that does not understand our natural language.

However, at the Developer Conference 2018, Google presented Google Duplex, a technology that makes it possible to have natural conversations and perform real tasks over the phone. The video in which Google Duplex is presented has achieved over 2.5 million views* on YouTube. Watching it, it quickly becomes clear that people cannot distinguish the computer voice from the voice of a real person, and the system understands real human speech and responds appropriately. For example, it arranges a hairdressing appointment for the person it is talking to or orders a table in a restaurant.

 

AI systems take over human tasks

It is therefore conceivable that the system will also be used by companies - not just banks and insurance companies - in the future, for example to take and process orders, complaints or damage reports over the phone. The system is constantly learning from caller feedback.

The British mathematician and computer scientist Alan Turing, one of the most influential theorists of early computer development, developed the Turing test in 1950. It is designed to determine whether a computer has a thinking ability comparable to that of a human and can therefore no longer be distinguished from a human. If a human questioner no longer recognizes whether they are communicating with a human or a machine, the test is considered to have been passed. Based on this criterion, Google Duplex has passed the Turing test, because: At least in the video, the people on the other end of the phone line do not realize that they are talking to an artificial intelligence.

However, the Turing test is controversial because it only tests the functionality of a system, but not whether the artificial intelligence also has consciousness and intentionality. Furthermore, in its basic form, it is only based on a conversation via keyboard and screen: It does not involve any auditory or visual contact between the participants. Google Duplex is only about acoustic contact. Nevertheless, Google Duplex passed the test at first glance because people on the phone can no longer distinguish between human and machine.

But was the test really passed? That remains questionable, because: Google Duplex can only make calls if it has been thoroughly trained in the relevant area. The system cannot make general calls. In addition, communication has so far been limited to less than one minute. This is why Google Duplex has not actually passed the Turing test, because the system would have problems if it had to hold a longer conversation or a conversation about a different topic.

Nevertheless, accepting damage reports or complaints, for example, could be a future AI field of application - provided the system has been trained for this. This would also be possible in relation to other topics. This also opens up potential applications in and for change management.

 

Significance for change management

The importance of AI for change management should be considered from two perspectives: On the one hand, the introduction and use of artificial intelligence in the corporate context is associated with a change process. This must be shaped and accompanied. On the other hand, however, the consulting industry itself must also open up to the use of AI, as it will also play a key role in the future when advising (not only) on the topic of change.

The use of artificial intelligence inevitably requires a change process in companies, and according to a survey conducted by the research company Forrester in 2017, this is precisely where many decision-makers see a major hurdle - at least if the aim of using AI is for work previously carried out by humans to be taken over by machines, as humans will then only be needed to train the system at the beginning.

In addition to a change in business processes, this also involves a change in corporate culture. For this reason alone, the people affected in the respective companies must be involved in the process. What exactly this means must be decided on a case-by-case basis. There is no one-size-fits-all recipe for accompanying such a change. It is therefore all the more important to be aware that the introduction of artificial intelligence inevitably requires targeted change management.

The financial sector is already undergoing such a transformation - partly because FinTechs are challenging the market power of traditional institutions in many areas with their agile and customer-oriented way of working. This is why established institutions will also increasingly rely on AI in the future and try to distinguish themselves as technologically innovative service providers.

 

Consultants must deal with the use of AI

In addition to the financial services sector, the consulting industry is also affected by this change. It also needs to think about the use of artificial intelligence, as the interface to the customer plays a central role here too. So far, the possible areas of application for systems such as Google Duplex are still very limited, but this will change in the near future.

Let's stay with the example of the customer interface. Let's assume that a consulting firm is approached for a change process. The first step is to understand the customer's concerns. This requires phone calls and a good analysis of the situation. Until now, such telephone calls have been made either by a back office or by the consultants themselves. They then arrange appointments on site to take a closer look at the situation and get a concrete picture.

What would it be like if this analysis were carried out by artificial intelligence over the phone in the future? Imagine the following scenario: You as a customer call and ask for change management advice. Your inquiry is picked up by an artificial intelligence system that asks you the important and correct questions to clarify your request. The system then connects you directly with an expert. This saves you time and analysis costs - even if the expert or consultant subsequently carries out another detailed analysis.

 

AI elements are incorporated into the advisory process

Current AI systems are not yet able to carry out complex (telephone) analyses and consultations. However, there are already systems that support such processes. Although these are not yet AI per se, they are a preliminary stage. Dr. Kraus und Partner (K&P) also uses such a system in the area of change management.

Thanks to our many years of consulting experience, our organization has extensive expert knowledge in this area. All K&P consultants and clients should benefit from this. This is why K&P stores individual and collective knowledge and experience on the subject of change in a software package. However, this concentrated know-how is only of interest to customers to the extent that it matches their needs. They would quickly be overwhelmed if K&P, metaphorically speaking, simply dumped all its knowledge on their desk and left them alone with it - possibly with the comment: "You 'just' have to tell us what you need."

Instead, the change management system is used to show customers the various options (for action) based on their situation. Customers then decide for themselves which options to choose. In the next step, they are again shown the various alternatives for the selected option, from which they again select the options that are relevant to them. In this way, the system guides customers step by step through the key questions they should ask themselves in their current situation. This enables customers to deal with their concerns immediately without having to understand the whole world of change management. And the K&P consultants? They can enter into discussions with customers better informed.

In its current version, the system still works without artificial intelligence. Instead, it is continuously updated and adapted with the knowledge of K&P consultants. However, its mode of operation is similar to that of AI systems. This means that system branches of the software program that have proven to be successful are retained, while branches that were less effective are adapted and improved.

 

AI helps to design customer-specific solutions

K&P's experience of using this system has been extremely positive to date. Regardless of the size of a project, it helps to quickly gain an overview of the customer's plans or concerns and to develop customer-specific solutions. The response from customers has been correspondingly positive, as they are generally only interested in the consulting knowledge they need at the time. And if it turns out in the course of the consulting process that another option should be considered and thought through? Then the software program simply has to be run through again based on the current level of knowledge and understanding.

Based in part on the experience gained with the use of the current software, K&P is convinced that the use of AI - including in the consulting process - will continue to advance in the future, and both internal and external change consultants must adapt to this. On the one hand, because consultants themselves must change and be prepared to use new technologies, and on the other, because they must be able to support the change processes resulting from the use of AI by clients - regardless of whether these take place at the level of a company's strategy, structure or culture.

 

Preparing for the future now

The current state of development of AI does not yet permit any conclusive application scenarios. However, it does reveal numerous possibilities for accelerating business processes in the future and making them more efficient. However, it is important to familiarize yourself now with the possibilities and requirements of using AI-supported systems in order to identify potential areas of application at an early stage and prepare for their operational use. This strategic preparation can determine which players will be among the winners in the increasingly dynamic business environment of the future and which will disappear from the market due to technological disruption.

Whether we humans will get on so well with artificial intelligence in the future that we fall in love with it like in the movie "Her" remains to be seen and, above all, is up to us. But one thing is certain: it will accompany us even more intensively in the future than it already does.