I regularly hear the following three objections when I talk to companies about the potential of generative artificial intelligence (GenAI). First: we fear that this may violate data protection guidelines. Second: AI's hunger for data is exceeding our IT capacities. Third: we would love to, but we don't have the staff. Justified concerns that can be mitigated with secure corporate AI tools from the public cloud.
According to a survey by the German Chamber of Industry and Commerce (DIHK), the technology and finance sectors in particular are using artificial intelligence.1 But it is also worth taking a look at industry. The first companies are moving towards Industry 5.0 with GenAI. This is because they take collaboration between humans and machines to a new level, as both can communicate with each other thanks to generative language models. Automotive supplier Schaeffler, for example, is equipping its first machine tools with a co-pilot from Siemens and Microsoft. If the company needs a new code for machine control, the co-pilot can take over. Humans only have to dictate the desired new functions. Audi's own generative AI supports the development of new rims. Bosch, on the other hand, generates images of faulty products with GenAI. The industrial group then feeds the images into a software program as training data so that it can detect defects in new production lines.
When I look around on social platforms such as LinkedIn, I get the impression that there is hardly anyone who has not yet been infected by the enthusiasm for generative artificial intelligence. I think this willingness to experiment is great, especially when companies are already using GenAI as a source of new (business) ideas or using generative AI as a planning tool or sparring partner. However, as the use cases are now becoming increasingly complex, the demands on the company's IT infrastructure are increasing. Simply having a large amount of computing power is not sufficient. If you want to become more effective and efficient with GenAI, you need a high level of IT security and full compliance with existing data protection regulations. The European Union's AI Act also puts GenAI on a leash for greater legal certainty: operators of basic models such as ChatGPT should disclose which data they use to train their models.
I just mentioned the keyword “basic model” – I could also have said “foundation models” or, more precisely, “large language models”. These are synonyms for the actual core of generative AI. The models are trained with vast amounts of raw data so that they can learn to analyze and understand this data. The quantity and quality of the data are crucial for GenAI to be able to generate new content – text, images, music, and even software – based on learned patterns. The chatbot, the most prominent member of the GenAI family, is based on the Generative Pretrained Transformer (GPT) language model from Open AI. You probably communicate with it regularly – whether professionally or privately. But do you know which data you can and cannot entrust to it? And: do you have enough good data in your company to establish your own GenAI applications?
This brings us to perhaps the most important challenge for generative AI: the issue of data protection. Some 85 percent of respondents in a Bitkom survey2 state that data protection requirements are slowing down their own AI efforts, while 76 percent speak of legal uncertainties. This explains why 68 percent of German companies see the potential of generative AI and consider it to be the most important technology of the future, but only two percent actually use GenAI. If the thought of AI and data protection or compliance leaves you feeling uneasy, and you want to avoid potential risks with regard to AI governance, then you should rely on cloud services – and make sure that you choose a platform that combines these aspects. This is the only way you can generate content that complies with data protection and copyright law when you create graphics, texts, or program code with artificial intelligence. Find a partner who can support you as you move into the cloud so that you can tap into its potential with confidence.
Analysts at Gartner3 expect strong growth in the public cloud market in 2024 and forecast a 20 percent increase to a volume of 679 billion US dollars. This increase is also due to AI. After all, there are many hopes associated with artificial intelligence. According to McKinsey, GenAI solutions alone have the potential to increase global economic productivity by 2.6 to 4.4 trillion US dollars. But to do this, companies need to pave the way for AI with the cloud. It also offers companies that have neither specialist AI expertise nor sufficient specialist staff the opportunity to work more efficiently and effectively with smart algorithms. The hyperscaler platforms such as Azure ML (Microsoft), SageMaker (AWS), and Vertex AI (Google) provide frameworks, libraries, and collaboration tools that can be used to develop generative AI models and applications.
Anyone opting for a public cloud service should make sure that the provider secures the AI applications with clear guidelines on the responsible handling of sensitive data. The cloud infrastructure must not only be powerful, but also compliant with data protection regulations and regulatory requirements. If you are working with a technology service provider on your cloud and AI strategy, ask them whether they can guarantee with a GenAI proof of concept (POC) that you can use to exploit the AI potential in a compliant manner without having to invest in training AI systems.
There are already dedicated Software as a Service (SaaS) offerings for numerous GenAI use cases. And market leaders such as Microsoft, SAP, and Salesforce are increasingly integrating generative artificial intelligence into their SaaS solutions. The plus point: they can be implemented quickly and cost-effectively – without lengthy implementation times. The disadvantage: turnkey solutions prevent or severely limit individual adjustments. If you are planning more complex GenAI projects in the cloud, Platform-as-a-Service (PaaS) models are therefore recommended. They provide tools for quick results, but are flexible enough to be used to develop your own GenAI models.
Don't let your concerns or an unsuitable IT setup slow down your AI transformation. AI is no "nice-to-have" and it is not simply yet another technology project. No, it is a strategic imperative for companies if they want to remain competitive. My tip: exploit the potential of the technology. But make sure you control the data protection and ethical implications. A competent provider can help you with this. And: exchange ideas with others. Just like AI, humans can benefit from learning too. With this in mind, I look forward to exchanging ideas with you.
1 Digitalization Survey 2023, DIHK, 2024, dihk.de
2 German industry picks up the pace with artificial intelligence, Bitkom, 2023, bitkom.org
3 Gartner Forecasts Worldwide Public Cloud End-User Spending to Reach $679 Billion in 2024, Gartner 2024, gartner.com
