Frank Ferro: ‘Mastering GenAI within your Data Journey’

How do you make GenAI a successful part of your overall data journey? For over 10 years, Frank Ferro has been generating business value with data and AI. After many years as a senior executive at PostNL and a temporary role as Program Director GenAI at ANWB, he will soon take up the position of Director of ICT at Amsterdam UMC. During a recent CIO Masterclass at Anderson MacGyver, he shared his vision and experiences. Here’s a summary.

In various sectors, the focus has long been on physical products and services, but attention is increasingly shifting toward data and algorithms. These are no longer just used to enhance customer experience and optimize business processes but are also becoming marketable products in their own right.

To effectively work with traditional and generative AI, three things are recommended: a safe and controlled environment, proper governance, and maximum independence from major tech players. The foundation for this is a solid data management structure and IT architecture, which organizations have often already been working on for some time.

Moreover, it is crucial to know exactly what you are doing. This usually starts with improving the necessary knowledge level within the organization. This requires genuine attention, as even a basic level of data literacy is often lacking.

This approach allows you not only to work on the required quality of data—both centrally and from within the business—but also to explain why you are using AI, where, and how. For example, when an algorithm makes decisions in a call center environment, it should be transparent what the AI is doing and for what purpose. The same applies to fraud detection applications.

In the early stages, it’s best to choose a straightforward, engaging “moonshot case” that generates visible business value for everyone. Without active involvement and support from the business, initiatives are unlikely to succeed.

Controlled Environment

A key recommendation is to use GenAI in a safe and controlled environment. Just like WhatsApp, where the data itself may be secure, metadata—such as who you contact and when—can reveal more than you might wish.

For instance, the free version of ChatGPT is not entirely watertight in this regard. However, the Enterprise version provides a secure, controlled AI environment. This allows even non-IT professionals to experiment and work with it, using their own personal and professional configurations. Examples include rewriting executive announcements or other official communications in plain language, which can save executive and communications teams significant time.

Another example is deploying a large language model (LLM) to answer customer inquiries via email at scale or to generate code for programmers—a practice already commonplace in many companies.

For most applications, humans will still remain in control for now. Microsoft Copilot, for instance, is explicitly not an autopilot. It acts as an assistant that can significantly boost productivity, but it is not error-free.

Proper Governance

Step two: ensure proper governance. Start by establishing a Data & AI Governance Board, which can assist at the executive level in making decisions about the potentially sensitive or impactful use of data and algorithms.

Equally fundamental is control over reliable data—not only to ensure process efficiency and achieve goals but also to comply with laws, regulations, and privacy and security requirements. Implement robust central and decentralized data management and quality control.

Interestingly, the “data owner” is often not the “data lord.” In other words, the person responsible for data quality is not always the one using the data to create valuable and intelligent solutions for the business.

Therefore, it is wise to embed data management within the business itself. Teams can then take responsibility for cleaning their own datasets for specific initiatives or use cases, even when data quality across the organization is not yet optimal. Waiting to start with GenAI until all company data is clean and ready may result in never getting started at all.

Meanwhile, you should work on the master data, ensuring, for instance, that all customer data is collected and accessible within a single system. For new initiatives or technical possibilities involving customers, there should always be a connection to the master data management system.

Virtual Federative Approach

A central vision and policy are necessary for an organization to reach the desired maturity level in data and AI. However, to make progress within specific business units or divisions, you can adopt a virtual federative approach. This involves assigning staff from the central tech or data department to specific business units or projects.

Over time, data governance and data management can increasingly be handled and discussed at the decentralized level—for example, during planned meetings or sessions. Initiatives with potential risks related to security, privacy, or ethics can then be escalated to the central Data & AI Governance Board.

You should also consider conducting an ethical assessment to determine whether new digital possibilities and plans align with the organization’s values and principles. Often, this will quickly clarify whether something is feasible or not. Such assessments can also help identify risk mitigation measures.

Maximizing Independence

Finally, strive for maximum independence from big tech. It’s fine to use services from Microsoft, Google, and other major players, but avoid becoming dependent on them. Instead, select services and providers that add the most value in your context, and don’t put all your eggs in one basket.

Ensure you have a control layer capable of integrating with multiple LLMs. This enables you to use the most effective LLM for each use case.

In summary, GenAI offers unprecedented opportunities to create business value but requires a well-thought-out vision and approach. Ready to take the next step? Anderson MacGyver combines years of experience with challenges in strategy, organization, data, and technology to help drive new digital developments. Want to know more? Contact Anton Bubberman to share your ambitions and challenges. We’re happy to help!