Identify and implement applications for artificial intelligence (AI) – Data-driven business architecture

How to identify and properly implement applications for AI in enterprises

Artificial intelligence (AI) has many practical applications for businesses. Think of AI as a larger framework that provides different capabilities. Companies implementing AI should review their vision and develop an AI strategy that aligns with their larger goals and focus. To implement, data-driven use cases are the basic requirement. Also, a flexible and modular architecture should be chosen from the beginning, which can be expanded and improved as needed in the future.

If you have already read the first part of the series “The Executive Guide to Artificial Intelligence”, you know why AI is essential for effective data processing and which seven components comprise a “Cognitive Enterprise”, the AI-infused company.

Artificial Intelligence has many practical applications for businesses and is being used today to enhance, improve and change processes or products.

A successful AI system is no coincidence

It is impossible to derive value from something if it is not understood, unless it is a happy accident. In the world of data processing, and therefore AI, there are no happy accidents; An AI solution is defined with meticulous precision, taking into account specific goals, available data, and the algorithms used.

AI basically has two core competencies. One is information gathering such as speech and image recognition, search and clustering. This allows machines to convert information from unstructured data to structured data. Second, AI can be used to contextualize information to each other to “understand what is happening.” This includes capabilities such as Natural Language Understanding (NLU), optimizations, predictions, and (context) understanding.

These different functions work sequentially and are usually also merged with other technologies (e.g. Robotic Process Automation, Cloud Computing, Internet of Things etc.) to form a single entity. Only by merging different approaches and technologies can the potential of data be fully exploited.

Data-driven use cases are the basic prerequisite

To introduce Artificial Intelligence use cases, companies should start with their strategic goals and identify urgent challenges or process/product/service visions that need to be addressed. For this purpose, there are various techniques such as design thinking workshops or user story mapping that can support this process.

The next step is to analyze the identified areas for economic viability and technical feasibility. Only if sufficient data is available, and necessary data from possibly different systems can be combined, can an AI-based solution be made possible. If this is not the case, the data management must be adapted to future requirements beforehand.

For a better overview, I recommend entering the results of the analysis in an AI heat map or otherwise documenting the information. With this overview, the various options can be evaluated. If a company decides to pursue AI projects, it can develop a business case for each one.

AI implementation

The first steps in an AI solution can be complex, as infrastructures have to be created and data has to be prepared, curated and merged accordingly. It is helpful to divide large use cases, such as the end-to-end automation of a core process, into smaller use cases and milestones.

It can be helpful to start with a Minimum Viable Product (MVP) and scale up afterwards. Creating the first AI component, no matter how small, is an important milestone for any company.

It is recommended to focus on an infrastructure and architecture that is flexible and scalable right from the start. As already paraphrased, AI solution are composed of a set of AI functionalities as well as other technologies that are continuously evolving. In order for the enterprise to continue to scale the AI solution(s) in the future and extend it to a “cognitive enterprise” to other areas of the enterprise, flexibility in implementation should be considered from the beginning.

Ideally, in an AI system that is composed of multiple components, it should be possible to replace any individual component to improve the function of the overall system.

For long-term success, companies will develop an “ecosystem” to support their AI and automation projects. This should reflect architecture teams that guide AI-related options through development and implementation, to operations and further scaling.

Additional data management roles such as a “chief data officer” may also be useful. AI systems are built on data, so skewed or inappropriate data can affect outcomes or even lead to misinterpretation.

AI enthusiasm pays off

The general enthusiasm for AI can be valuable for any company. However, this requires openness to innovative technologies and agile project methodology.

Artificial intelligence is a marathon, not a short-distance sprint. So far, I have seen the greatest successes and added value in intelligent computing when companies have established AI piece by piece into their vision and core processes. When the expertise of data and architecture experts is used and their own ecosystems are created in parallel.

Britta Daffner ist seit über einem Jahrzehnt in der Technologie- und Daten-Industrie zu Hause. Ihr Credo: Innovation und Digitalisierung von Unternehmen vorantreiben – durch Technologie und moderne Führung. Dafür befähigt sie als Practice Leader Data & Technology Transformation in IBM Firmen dabei, das volle Potential aus Daten zu nutzen und unterstützt als Coach Macher*innen, die in der Konzern- und Wirtschaftswelt etwas verändern wollen. 2021 erschien zudem ihr Buch „Die Disruptions-DNA“ (www.disruptionsdna.de), das dazu inspiriert, die Digitale Transformation aktiv mitzugestalten.

Comments are closed.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More