Ensuring successful AI integration: Best practices and key strategies

Adopting a strategic and thoughtful approach is essential to ensure the successful integration of AI. Grégory Gruber, Deputy Director of Outsourcing & Cloud Services at Proximus NXT, shares his insights on how businesses should prepare for this transition, the challenges to overcome, and the best practices to follow in order to maximize the benefits of AI while ensuring security, compliance, as well as ethical and responsible use.  

How should companies prepare to integrate AI into their processes, and what are the key steps IT leaders should follow to ensure successful adoption?  

Grégory Gruber (GG): "Preparation starts with a clear understanding of the objectives that AI should achieve. Companies often approach AI from a technological perspective, trying to solve problems through technical solutions. But in reality, AI should be thought about in terms of what it can bring to the company. For example, it can help automate manual or repetitive tasks, such as data consolidation or decision-making assistance. AI does not replace humans but rather facilitates their work by processing large volumes of data and providing predictive models. Humans then make the final decisions based on these recommendations. Actually, AI is not meant to replace humans but rather to enhance their ability to act more efficiently.  

AI can be applied to a wide variety of business processes. It is essential for a company to begin by identifying use cases to ensure AI is deployed effectively. In finance, for example, AI can automate document analysis or even detect fraud. In sectors like communications or marketing, AI can be used to personalize customer interactions or automate repetitive tasks. The key is to carefully consider how AI can contribute to three main objectives: improving customer experience, increasing employee satisfaction by freeing them from repetitive tasks, and generating cost savings. These savings can then be reinvested in new AI projects, creating a continuous development process."  

The integration of AI creates new roles and skills. What professional profiles are essential for this transformation, and how is Proximus NXT preparing its teams, particularly in terms of training and change management, to succeed in this transition?  

GG: "To integrate AI, it is essential to train all actors in the company, including legal, compliance, and business departments, on the technologies and associated processes. The goal is to demystify AI and explain what it can and cannot do. This enables each team leader to understand where AI can add real value. Once these foundations are established, it is important to stimulate creativity within the company through ideation sessions to identify processes that could be automated or improved by AI."  

Once these ideas are collected, use cases should be evaluated according to criteria such as customer experience, employee satisfaction, cost savings, and implementation complexity. The simplest and most impactful cases will be prioritized to generate quick wins and facilitate AI adoption. Another key aspect is the establishment of an AI Center of Excellence, where 'AI champions' play a vital role in spreading best practices and training other employees.  

The first stages of an AI project require specialized profiles, such as Data Scientists and Data Engineers, to ensure data quality and security. Then, roles like AI Architect and AI Developer will be necessary to integrate and customize existing models. As AI evolves, more expert profiles, such as AI Engineers, may be involved to develop custom algorithms. These roles will be coordinated under the leadership of an AI manager, often heading the AI Center of Excellence.  

However, AI integration also presents a change management challenge. It is important to guide employees through role transformations. If certain tasks are automated, it is crucial to think about the new responsibilities employees can take on to continue adding value to the company. This reflection should be integrated from the beginning to ensure smooth and successful AI adoption."  

What are the main challenges and best practices that companies should consider to strengthen their policies and principles around cybersecurity, data, and AI?  

GG: "While AI provides security solutions, it also introduces new risks, such as its malicious use in cyberattacks and phishing. To address this, it is crucial to employ AI ethically, with safeguards to prevent misuse, particularly the abusive use of chatbots or generative systems. As a result, securing AI requires not only robust measures but also careful consideration of its role in facilitating attacks, with adaptive security policies designed to protect sensitive data and promote ethical practices.  

Security measures must be adapted to the technologies deployed. It's not just about firewalls but about strict access controls for the data used by AI. For example, when using techniques like 'retrieval-augmented generation' (RAG), it is crucial to restrict access to sensitive information. Each role, such as that of a sales employee, should be limited to accessing relevant data. This security and compliance aspect should be integrated from the start of AI implementation, alongside the definition of use cases.  
This approach continues throughout the initial experiments, which help identify early successes, potential biases, and necessary adjustments, particularly regarding security. Once these adjustments are made, production can begin, while continuing to monitor and adjust security and compliance to ensure a smooth and responsible adoption.  

Protecting confidential data and intellectual property is paramount. Companies must classify their data and opt for secure solutions, such as private clouds or data sovereignty options. Public cloud services can expose sensitive data to compliance risks. To address these challenges, Proximus NXT offers secure solutions like Clarence and U-Flex, ensuring data confidentiality and sovereignty, either in on-prem infrastructure or in Luxembourg."