Creating Value from AI in 5 Steps

Organizations have high expectations regarding AI and its potential. These expectations come from both executives, who recognize AI as a key lever for improving performance at all levels, and end-users, who have discovered AI’s capabilities, notably through tools like ChatGPT. “AI contributes to enhancing value in many ways. Within our group, we use it to improve our services, while helping our clients, through Codit, accelerate their transformation using AI,” explains Grégory Gruber. “The expertise we have developed internally, across various AI use cases, allows us to better guide our clients in adoption and support them in managing risks and regulatory challenges.”

I. Define a Strategy

Leveraging AI cannot be improvised. The challenge lies in fully understanding its multiple dimensions. “It’s not just a technological issue,” emphasizes Eva Gram. “It requires a multidisciplinary approach and the creation of an internal AI community, bringing together technical experts, business representatives, and employees from legal, risk, and data management teams. This collective expertise defines how AI can serve the company’s development objectives.”

The first step is setting a clear vision: determining what the organization wants to achieve with AI. “The goal is to create value for users and support change. By monitoring technological developments, this community can explore, evaluate, and validate new use cases,” adds Eva Gram.

II. Assess, Set the Framework, Choose the Right Solution

Beyond strategy, implementing any new AI use case requires evaluating expected outcomes, defining deployment scope, identifying required data, and assessing confidentiality levels. “When using AI for the first time, it’s recommended to work with limited datasets. The success of an AI project largely depends on the quality and availability of data. Gradually, organizations should evolve their governance as they gain maturity,” explains Eva Gram.

For instance, if AI is used to help employees access HR data, it is crucial to ensure that users cannot access unrelated information. For contract data retrieval, confidential information must remain protected. “The architecture depends on these aspects. Based on constraints and opportunities, we can determine whether to deploy via cloud solutions, hybrid approaches (U-Flex), or a disconnected sovereign cloud, like Clarence, which opens the door to AI sovereignty,” adds Grégory Gruber.

III. Compliance, Security, Control

Choosing the right solution also means considering regulatory compliance, confidentiality, and security. “Organizations must implement AI governance policies that satisfy regulators. In some cases, use cases must be reported to authorities, security policies adapted, and procedures established to identify, assess, and mitigate risks,” notes Eva Gram.

IV. Ensure Implementation

Once the framework and AI platform are selected, the next step is implementation. “Several elements need consideration. Teams must be trained both to use and supervise AI,” explains Grégory Gruber. AI models should first be trained on datasets to ensure relevant results. Through reinforcement learning from human feedback (RLHF), results can remain as accurate and useful as possible.

V. Deployment

AI deployment occurs only after proper training. At this stage, employees can use the solution, or it can be exposed to clients.

The framework should allow for scaling AI across departments and users. “The speed of transformation depends on team proficiency. Change management is crucial for success, alongside ongoing supervision of deployed AI solutions,” adds Grégory Gruber.

While deploying AI solutions may seem complex, Codit and Proximus NXT emphasize that it can be manageable. Starting with simple use cases that deliver tangible results allows organizations to gradually embrace AI. “It’s important to work with internal teams interested in the technology. By creating a network of AI champions, adoption is easier. These employees can share successes and encourage others to embrace AI,” explains Eva Gram.

Indeed, placing AI at the service of employees is how true transformation can take place.