Artificial Intelligence is a strategic priority, but it requires great attention to the risks involved. Here, AlixPartners explores a series of successful, carefully-constructed GenAI adoption strategies, with use cases from Fastweb, Italiaonline, Lutech, Nexi, and UniCredit, focusing on where to start, how to manage risks, and the lingering doubts about true costs and benefits.

Artificial intelligence is no longer a novelty. Technology is revolutionizing various industries, opening new opportunities and unprecedented levels of efficiency. AlixPartners confirms the trend in its Digital Disruption Survey: 75% of leaders say that AI will be extremely important for their industry.

That said, many companies still struggle with key questions: from where to start and how to manage risk to doubts about the cost and actual benefits of AI solutions and specific use cases.

GenAI is definitely one of our clients’ top technological priorities,” says Davide Antonazzo, Director at AlixPartners. “Our survey tells us that, for 89% of companies, AI can help increase operational efficiency. But the technology can go much further, supporting an increase in creativity that has not yet been experienced at the speed promised by GenAI."

It is worth remembering that generative AI models, which are used to produce new data, differ from discriminative AI models, which are otherwise intended to sort data according to differences. "Discriminative AI is already a commodity,” continues Antonazzo.

Download our PRACTICAL AI PLAYBOOK: 

In many enterprises, such as those with call centres, AI is being used to speed up flows, expanding the number of use cases that set the pace for others. Yet for many today, the most appropriate approach would be to start with customer needs and build on them to understand the most effective use of AI within their strategy.

These are sound approaches which are not exclusive. “On the one hand, we have the organization that already knows the topic and knows where it wants to go,” explains Marcello Bellitto, Partner & Managing Director at AlixPartners. “On the other hand, use cases can help understand how to approach the topic in a vertical and specialized way. GenAI will certainly change some industries and ways of working, but it is the result of an evolution not so dissimilar to the technological revolutions that have characterized the history of humanity.

How to shape the strategy

In a few years, AI will also allow you to consider the topic of business sustainability within the context of the now-established technological models, operating at hugely increased speeds.

An example is the cloud,” explains Antonazzo. “Considering the Italian landscape – which is somewhat peculiar compared to other countries – AI will allow companies, including SMEs, to cross-reference data and obtain insights that would have been more difficult to identify before. Vision, skills and a network of reliable partners are the factors that can make a difference.”

But there is one aspect not to be neglected: Understanding risk. “There is a lot of hype about the world of Gen AI,” continues Antonazzo, “but we have not defined the applicability formulae yet, despite the will of the European Union to regulate the sector. Therefore, risk management skills can convey success even amid ongoing uncertainty.”

For more on risks, read our Harvard Law Business school article “The Board Member’s Oversight of AI Risk – Moving from Middle to Modern English”:

In this context, some companies are focusing on the skills that can drive the adoption of AI for practical applications. This is the case of the Lutech Group, which has set for itself a pivotal goal in 2024: 800 new hires, the majority of which will be dedicated to the growth of the AI-based market.

To attract the best talent, we must talk about interesting, very futuristic projects,” says Giorgio Ancona, chief delivery officer Industry Capabilities of Lutech Group.

To do so, we decided to invest in both resources and territory, opening offices in Bari and Cosenza so as not to leave anyone behind. For the delivery under my responsibility, we will create activities strongly focused on AI, which are under the broader umbrella of data. No-one can yet say if in the future, in addition to organic growth efforts, there could also be further momentum for vertical acquisitions.”

To confirm the willingness to use GenAI and evaluate its opportunities, Lutech has launched Lutech Brain, a GenAI solution to enhance Corporate Knowledge Management by optimizing sales processes, thus speeding up customer response times.

For example,” Ancona explains, “GenAI allows us to better manage information that is useful in response to public tenders and also in preparation for testing phases.” This references the creation of sophisticated test cases based on customer insights and the needs of his own business. “Not to mention the skimming of curricula for HR, applying generative models to vertical uses, which today represent the greatest advantage of such tools.

The drive for training

Change management is the other piece of the group’s ‘augmented’ strategy. Here sits Lutech Next, the business advisory and consulting line that promotes a culture of innovation, adopting an approach that goes beyond purely the field of IT.

Implementing new processes using AI requires DNA evolution from within,” says Ancona. “We follow this paradigm first-hand and then transfer it outside. AI forces us to prepare workers to work in areas different from those of the past. Several years ago – as Lutech Group – we started experimenting with quantum computing. The ability to anticipate trends in order not to suffer from change is what drives us.

Just like during the industrial revolution, the drive for re-skilling and training must come from with enterprises.

It is not enough simply to buy the latest hardware to advance your business,” Ancona continued. “The integration of skills also creates the inevitable convergence of the IT world and consulting companies. As AI permeates more and more industries as a technological topic, we will need to have ‘domain skills’. Getting to the root of business projects will mean knowing how to speak more languages, accompanying the customer on a path that is both technical and ethical, with important social impacts.

AI governance

So, who is responsible for overseeing how data is used and the logic of AI algorithms, both within the organization and in the relationship with suppliers?

To answer this question, we need to consider four themes,” explains Stefano Gatti, head of Data and Analytics at Nexi Group. “We need to focus on technology, on the organizational model, on human capital – that is on talent and the combination of strategy and culture. When it comes to ‘embracing the data’, one of the mistakes that companies have made in the past was to start with technology.

This was not always a winning approach, comments Gatti. “It is critical to have the right people and define the optimal organizational model before focusing on technology.” Over the last few years, Nexi has embarked on a path in which the distribution model is intertwined with the centralized one, with the idea of applying AI in core processes. “We defined a basic technological line, chose a unified cloud and supported the semantic understanding of data on a ‘lake warehouse’ to carry out AI and machine learning projects.” In fact, Nexi has created a real data community, eliminating silos by adopting a holistic perspective.

Governance is about retaining diversity and different ways of thinking and acting together. “In addition, we wanted to broaden the AI table by bringing in security and compliance professionals, trying to identify problems before projects started,” Gatti continues.

The data managed by Nexi has several owners. “Among these, there are merchants or the entities with whom payments are made – banks, cardholders and account holders. Being central in the Italian economic system, we have chosen not to follow data monetization but to use the information to enable fast, certain, and easy payments for all the stakeholders".

Data becomes an enabler of concrete innovation, especially in dialogue with AI. “It is important to consider that artificial intelligence in a company may not guarantee a return on investment as fast as you might expect,” explains Gatti.

To fully embrace the change that technology can bring to the business, it is essential to move beyond the short-term business case approach and adopt an Agile mentality. The goal is adaptive change that is AI-driven and no longer just data-driven. This implies the need to invest in people in the company who have strategic and technical skills and who act at the same time as gatekeepers and points of contact between departments without creating additional intermediate layers between the wider business and IT. Even this path has not been easy; training paths and redefining roles are steps that imply concrete commitment and openness to change, going beyond the simple innovation of the system.”

In our Practical AI for CEOs playbook, we outline the essential components for approaching and operationalising AI across an enterprise to drive value:

 

Enabling efficiency with AI

If GenAI is the buzzword of the moment, for Fastweb, the journey began as early as 2019 with a strategic approach based on individual use cases, with the aim of developing an organic program for reasoned use of AI.

We put data at the centre of the process, followed by technology and organization, and only later did we outline a shared roadmap for the implementation of possible use cases,” explains Giovanni Germani, manager of the Architecture & AI Center of Excellence at Fastweb.

From our point of view, AI must not be a topic for the few, but must be pervasive in the company. To understand the benefits in day-to-day work, people must be accompanied in the widespread adoption of technology.”

Here are the common challenges we see at AlixPartners: “5 hurdles to overcome for AI success”


Anticipating industry trends, Fastweb adopted a specific code of ethics for AI at the beginning of 2020. “We introduced principles, later taken up by the AI Act, many of which focused on the concept of an artificial instrument that must be ‘human’. After integrating AI into Fastweb, the next step was to extend this innovation to our products.”

Companies must adopt specific compliance and governance measures under the EU Directive. “On the one hand, we need to certify our solutions and, on the other, manage projects, analysing both the ethical use of technology and data protection. For Fastweb, our Data Governance team comes into play, working to qualify information and trace data from its origin through all of the transformations it undergoes.

To monitor the degree of AI pervasiveness for the Public Administration, the company carried out several analyses of public sector customers. Germani explains that there are two scenarios: “On one hand, you see those who have no problems introducing AI into the business processes, while others lack confidence in these tools, especially regarding data management.

The need to manage data in a secure and transparent way and provide companies and public administrations with powerful generative AI capabilities in Italian prompted Fastweb to develop a large language model (LLM) in 2023, trained natively in Italian, starting from scratch and sharing skills, a numerically relevant and certified dataset, as well as a dedicated infrastructure.

We have purchased a system consisting of 31 NVIDIA DGX H100s that we will use to develop our national LLM and make it available to companies, start-ups, public administrations and other operators; an end-to-end system for the development of generative AI applications specific to the various verticals – from healthcare to education to mobility.

Like other telcos, Fastweb is no longer just a connectivity provider. After developing a network of edge mini data centers to offer cloud resources and services directly near businesses, the next step was to add AI to their portfolio. All this, with the support of the Center of Excellence that acts as an AI innovation engine for Fastweb, its partners and third parties.

The competence centre itself has become an example of transformation towards AI for client companies,” notes Germani. "This requires a strong availability of technical training, which still represents a gap to be filled in the Italian labor market. In 2025, I expect a real boom for GenAI at any business level.

Will we have enough professionals to satisfy the demand? “Probably not,” says Germani. “And that’s why forward-thinking companies have understood that they need to invest in training so that they don’t run out of resources tomorrow.”

Italiaonline, too, focused on a Competence Center when it wanted to involve the entire organisation in the AI adoption process.

The implementation of AI represented a diversified path that involved various activities and business sectors,” confirms Andrea Rondelli, head of Data Platforms at Italiaonline. “To maximize its use and develop innovative solutions, we have established a Competence Center that actively engages the entire organization. We chose to act proactively rather than react to market changes."

In early 2024, we started a process and set up an initial group to work on three key objectives. The first was to carry out a detailed analysis of the technologies and AI platforms used, crucial for the second objective, which is the focus on efficiency, especially regarding costs and the identification of the most relevant use cases. The third objective is to scout the most promising use cases, focusing less on the technology itself and more on potential practical applications.”

As Rondelli explains, the second cross-functional working group was created as a result of an internal recruitment initiative. “The diversity of people, from all business lines and company personnel, reflects the importance of cross-functional knowledge sharing for the success of the project. In the first four months, the group catalogued the AI technologies used, gathering a wealth of information. Alignment meetings every fortnight allow us to monitor the progress of the assigned tasks and establish new objectives, promoting a continuous cycle of improvement and knowledge sharing.”

Focus on data governance

Returning to the AI Act, its evolution shows how AI is a cross-cutting topic that affects very different fields.

I followed the development of GDPR at the time and looking at the AI Act, I think we have a comparable system. For example, both look to adopt a risk-based approach with requirements varying depending on the level of risk involved,” explains Cristina Paola Cabella, head of Group Data Protection Compliance at UniCredit.

Even before the AI Act was finalised, certain principles of GDPR applied to AI systems in the broadest sense, without there being a specific reference to AI. For UniCredit, as with any banking group, the starting point can only be sound data governance. The implementation of the AI Act is a unique opportunity to review the processes of development, purchase, and implementation of AI systems. At every stage, you need to have a transparent knowledge of data flows and their use. Correct data management is a priority because AI systems need data, and the AI Act pays attention to their accurate use.

In the ongoing debate on the responsible use of AI, the focus is on the purposes of technology. Simplifying somewhat, GenAI is a defined tool in its way of functioning, but its multiple purposes carry different risks. So how can we deal with this situation?

Controlling every possible use of a generative AI system requires a lot of resources, checks, and a complex process for examining the compliance of AI solutions – that’s why an AI strategy is crucial, Cabella says. “It is necessary to define what AI use cases you want to pursue and build a framework around these that ensures compliance with regulatory requirements.

Cabella also highlights the necessary transition from the idea that AI systems represent, above all, an opportunity for companies to become more efficient (reduction in costs, time, etc.) to the conscious and responsible choice of how to use AI to increase process effectiveness.

It is through effectiveness that the organization can evolve, focusing on the skills of people. We will continue to need people within the business because individuals will have to instruct AI systems and validate their results.

We need to overcome the novelty effect; we need to understand its potential and risks. “Let's not underestimate the potential of the technology, but let's make it usable in a responsible way. Combining innovation, risk, and responsibility requires a fair balance between ensuring robust data governance and defining the priority investment areas where AI can be leveraged,” comments Cabella.

In summary

Bellitto concludes: “From the various contributions it emerges that, in the short term, to reap the efficiency benefits and development lines of the business models enabled by AI in EBITDA, it is essential to start from strategy and objectives, selecting a manageable number of use cases where AI can make a difference, and progressively scaling from there."

It will also be important to develop internal skills and the partner network through cross-functional business models, integrating traditional business and IT functions and skills. It will also be decisive to organize AI and data governance with clear accountability, with risk management and compliance embedded in management processes.

This article was originally published in Italian on Data Manager: