Success with Artificial Intelligence

Artificial Intelligence (AI) is no longer just a trend topic but has long been usable. But many companies find it challenging to start their own AI journey. This article shows how AI can successfully support companies. 

Many companies are skeptical about new technologies such as Artificial Intelligence, Machine Learning, and Big Data. They believe that these developments are merely trends that have no substance. But AI can increase the efficiency of companies and the quality of their offerings. Based on data analyses, conclusions can be drawn, which then can provide an essential foundation for entrepreneurial decisions. That is why AI must be placed on the agenda within the digitization strategy. Also, the importance of using AI-based solutions is now recognized by more and more CEOs and managers, as current studies show. AI functions can also be found in increased frequency in industry-specific solutions. And yet the conscious and targeted use of Artificial Intelligence is not widespread, because companies often do not know where they can benefit from AI. Furthermore, the success of AI deployment depends on companies implementing and integrating new technologies in a targeted manner.

In the context of Artificial Intelligence, a fundamental distinction is made between weak and strong AI. While strong AI is defined by the fact that it can plan independently, use logical thinking, and make its own decisions, the so-called weak AI can "only" solve tightly defined tasks. And since we are currently dealing exclusively with the latter category, AI-based solutions require clearly defined goals. Therefore, first of all, it is necessary to identify which essential tasks this technology should solve. Afterward, it can be examined at which points AI can support with its current state of development.

A further technical development has made it possible for companies to use AI in the first place, which is Big Data. Large amounts of data and the ability to analyze them very quickly are essential requirements for AI. To use Artificial Intelligence, companies must collect comprehensive data on their core business processes. That usually requires (technical) investments, since various systems which are used to collect data have to be connected. Alternatively, data can be collected according to requirements, based on a specific application or business models. Depending on each particular use case, it can be checked what amount of data is required for that case.

Not to mention: Artificial Intelligence causes change - and change can cause anxiety. It is, therefore, particularly relevant to include employees and prepare them for change processes. Digital transformation requires openness, the determination to adapt, and design. Professional change management is vital and pays off in any case.

1. Voice Commerce

Once companies have created the structural and technical prerequisites for the use of AI, the implementation phase begins. There are already numerous opportunities to benefit from AI in commerce: Digital butlers such as Amazon Echo Dot, Google Home, or Apple's Home Pod are now established in many households. Users can use these tools to shop conveniently by voice commands. Meanwhile, the devices are even capable of answering very complex queries such as "Find a color-matching T-shirt with a slim fit that fits with trousers A of brand C." The T-shirt can then be conveniently ordered - also by voice. Without AI algorithms, this would not work. Those work in a multi-stage process: First, the system must convert audio signals into phonemes. These are the smallest meaning-distinguishing linguistic units from which words and sentences are created. The logic then determines the exact user wants and converts the request into text-based commands to the backend of the system. That is usually done by calling web services. The result is then converted into speech using TTS (text-to-speech) algorithms and transmitted to the user. Thus, voice commerce represents another extremely attractive sales opportunity.

2. Detection of anomalies in the power grid

The use of AI is also highly attractive in the energy industry: To monitor electricity grids, many grid operators have so far used relatively inflexible systems that operate according to fixed threshold criteria. The systems monitor whether the target values under consideration are within a value range. If the company wants to examine an aspect - such as the load flow - it must therefore first define threshold values. With AI and Machine Learning, the power grid can be monitored much easier. Based on time-series measurements, this is also much more reliable. It is no longer necessary to define threshold values, as the analysis shows whether anomalies occur in the power grid. In this way, network operators can detect critical situations at an early stage and take appropriate measures to prevent the overload of lines and transformers.

3. Automatically generating metadata for image and video material

In the media sector, AI is part of reality for quite some time. Moving images play an increasingly important role. Due to the increasing amount of data, it is all the more important for journalists and producers to find video sequences as quickly as possible. The easiest way to do this is through so-called metadata, i.e., content information that is assigned to videos and images. The better the tagging within this description, the easier the search. If the amount of data is manageable, such data can be added manually. But when it comes to managing millions of images - for example, for TV stations - this quickly becomes a Sisyphos-project. With the use of AI, video files can be easily analyzed, and metadata can be automatically added. AI can recognize known persons, company logos, cities, or buildings and store them in the meta descriptions. Thus users can find video or image files very quickly.

These examples make it clear that there is not one ideal AI project which suits every company. Also, to what extent AI can be used depends on the digital maturity level of each company. Different projects or different measures may be required. In all cases, a step-by-step introduction is recommended: companies should start with smaller projects first and then initiate larger ones later. It is advisable to be supported by an experienced service provider. 

A four-step process has proven itself in practice:

1. Identify the use case with the most significant added value

While analyzing a current situation, the first step is to examine what foundations need to be laid for the use of AI in the company. How does the company currently use Big Data? What kind of know-how is already available regarding AI? Are there particular challenges to be overcome? Companies and service providers develop these questions in a workshop together. The findings can be used to compile a use case portfolio. In those, the potential applications for AI within the company are prioritized according to benefit and cost. The most promising use cases can then be further analyzed so that the so-called "readiness check" can be used to check which prerequisites need to be created for the implementation of the cases. In doing so, aspects such as data management, systems, and processes in the company are considered.

2. From the idea to the prototype

In the second step, it is essential to develop a concept for the procedure to realize the first prototype as quickly as possible. For this purpose, goals, and requirements have to be prepared in more detail within a design phase. A standardized development environment helps to eliminate disruptive factors and accelerate development. Thanks to an iterative procedure while taking feedback of the users into account, the AI prototype is created, which already works with real data. It quickly becomes clear whether the desired effects will occur, and if the set goals can be achieved.

3. From prototype to productive AI service

A Minimum Viable Product (MVP) is then created out of the prototype. The database and functionalities are iteratively extended. Continuous function, load, and integration tests ensure a smooth integration of the MVP into the system environment. Within a scaling plan, it can be recorded how an MVP becomes a scaling AI service. To do so, companies must analyze the future productive environment, influencing factors, and interfaces of the future AI service and take them into account within the implementation. The result: an AI service that can be integrated into existing applications, processes, and systems via interfaces according to requirements.

4. Stable operation and improvements

As soon as AI services are in use, it is essential to keep a continuous eye on them. Especially in business-critical activities, it is vital to monitor them 24/7. Constant re-training should ensure the quality of the AI algorithm. At the same time, a service management system built according to standardized ITIL processes takes over all routine activities necessary for day-to-day operations. In this context, companies should choose partners who can deliver comprehensive Managed Artificial Intelligence Services.

In summary, AI is no longer a dream of the future. The technical possibilities to implement AI projects are reality. AI is not hype, but has substance - it has a reason that AI is a component of numerous industry-specific solutions. That is why Artificial Intelligence will shape the business environment in the future. If CEOs and managers today do not deal with the technological possibilities offered by AI, they will suffer in the long run. Especially with professional support, potential integration challenges can be mastered well. The time to start the AI journey now has undoubtedly come.