AI Needs Organization
How Artificial Intelligence Becomes a Success With the Right Measures
Interview with Niels Pothmann, Head of AI, Arvato Systems and Martin Weitzel, Innovation Lead, Arvato Systems
The correct organization of AI projects is considered a basic requirement for the successful use of AI in companies. The experts from Arvato Systems, Niels Pothmann, Head of AI, and Martin Weitzel, Innovation Lead, explain which AI competencies should be built up and which conditions should be created within the company.
Implementing AI projects and using AI touches many areas of a company. Which AI competencies and roles are necessary?
Niels Pothmann: Basically, a distinction can be made between technical and professional specializations. On the technical side, a data engineer or data architect is needed to collect and consolidate the data. That creates the basis for the Machine Learning Engineer or Data Scientist, who combines technical competence and specialist orientation. He/she transfers the task definition with machine learning and AI into automated procedures. On a technical level, an AI ambassador demonstrates the advantages of an AI solution, and a product owner drives the topic forward. Together with experts from the specialist department, the functional orientation of the solution is specified.
What prerequisites must companies create to use AI successfully?
Martin Weitzel: The central prerequisite for successful AI projects is cross-departmental collaboration. Employees must be able, willing, and allowed to work together. Even today, companies often have to create technical, procedural, and cultural conditions before the employees required for AI adoption can find each other. To implement all of this, a clear and tangible vision is needed, on which everyone works together. And yes, that costs time and money.
Implementing AI projects and using AI touches many areas of a company. Which AI competencies and roles are necessary?
Niels Pothmann: Basically, a distinction can be made between technical and professional specializations. On the technical side, a data engineer or data architect is needed to collect and consolidate the data. That creates the basis for the Machine Learning Engineer or Data Scientist, who combines technical competence and specialist orientation. He transfers the task definition with machine learning and AI into automated procedures. On a technical level, an AI ambassador demonstrates the advantages of an AI solution, and a product owner drives the topic forward. Together with experts from the specialist department, the practical orientation of the solution is specified.
How can optimal interaction be ensured in organizational terms?
Martin Weitzel: That depends on the status quo. If you are starting out with AI and are still looking for the right ideas, you can start with a structured innovation campaign, for example. Because if the contents are unclear, it helps to make the process all the clearer. Those who already have ideas are more likely to set up a program that regulates the strategy, management, and coaching of AI implementations. How this is done depends strongly on the innovation DNA of a company. For some, the answer may lie in a decentralized AI transformation program; for others, it may lie in a central Innovation Lab. At Arvato Systems, we have an interdisciplinary Artificial Intelligence Competence Cluster, i.e., a cross-location network of around 120 experts, which is anchored decentrally in various business units. However, there are a few points that must always be observed:
1. Clear Responsibilities:
Especially in an integrative approach, it must be regulated who is responsible for what.
2. Define Ambitions:
It would help if you had business plans with hypotheses, which can be validated over time.
3. Agility, Flexibility, And DevOps:
Short decision-making, collaboration, and the joy of experimentation allow us to develop productive AI services quickly.
4. General Conditions:
Clear guidelines, for example, regarding budget and data protection, are mandatory.
5. Rules of the Game:
A top management godfather sets rules that may deviate from the usual rules: Maybe good stories are more important than turnover or cost savings.
6. Translator:
Roles are needed that can convey technical aspects in a technically understandable manner and vice versa.
How do the participants stay up to date in the dynamic AI environment?
Niels Pothmann: AI is not new, but many solutions can only be implemented practically today. They lack data and computing capacity. The challenge is not only to implement AI but mainly to keep up with the development. That's why companies need to train their AI team. On the one hand, it is worth looking at Hyperscaler, which is gradually expanding its cloud portfolio by integrating operating and application environments for AI.
On the other hand, it is advisable to make use of universities and research institutes' findings. Many pass on open source projects to their community. This way, new AI and machine learning methods can be applied quickly. Arvato Systems and the Bertelsmann Group rely on continuous learning: employees should become familiar with useful technologies. And nothing beats learning by doing. Using theoretical knowledge is indispensable.
Not every company can build the necessary skills. What distinguishes the right service provider?
Niels Pothmann: There is no general answer here. Some companies still have to find the right use case, for example. We can support them with our AI Journey method: We identify useful use cases, pilot them, implement a production-ready prototype, and transfer the AI solution to regular operations. We have also described this process model in detail in the Arvato Systems e-book, "How companies benefit from artificial intelligence. Prerequisites, deployment scenarios, and implementation paths for AI projects".
Martin Weitzel: So, individual consulting and implementation expertise are essential. To optimize core processes, many are already implementing pilot projects and integrating productive AI services. Does that often raise the question: standard or individual solution? A media company that wants to sign young artists or a financial service provider that wants to invest in lucrative start-ups does not need a standard solution. To be better than the market average, such companies need a unique tool. It does not have to be an individual solution. After all, developing it is time and cost-intensive. It has proven to be a good idea to individually adapt standard solutions to achieve performance advantages and stand out from the competition.
Niels Pothmann: A right service provider is halfway to the individual requirements of the company. It adapts the data landscape and business processes and adds further processes, platforms, data, and technologies. Also, he provides AI microservices for the specialist departments. During the practical implementation, he networks the data sources between the specialist departments and contributes his know-how for machine learning and data science. If you choose a provider who is convincing on these points, you are not doing anything wrong. And since AI solutions are continually evolving, long-term cooperation is advisable.
This article is part of the Luenendonk Magazin, which can be found for download in its full length here (German only).
Your Contacts for Artificial Intelligence