Marcello Bellitto
Partner & Managing Director - Milan
This article by Antonino Caffo was originally published in Italian by Data Manager.
Operational dynamics and AI agents: only 20% of European companies are ready for the transition. Lutech, Impresoft, and TAS are focusing on data governance, integration, and vertical expertise. The challenge is no longer technological but industrial.
Artificial intelligence is already in corporate budgets, buying habits, and processes. Yet, this acceleration brings a contradiction. The more adoption increases, the more the illusion grows that AI is simple to manage. Companies experiment, launch pilot projects, and multiply agents. But transforming all of this into concrete, measurable, and sustainable value is another story.
This is why it's useful to understand the strategies of companies like Lutech, Impresoft, and TAS, which are living this transformation in real time, and the insights of AlixPartners’ Technology, Media & Communications team, which is monitoring AI's impact on the European software ecosystem. According to AlixPartners’ recent report, 2026 Enterprise software technology predictions, only 15-20% of companies in the region are prepared to face this transition. A sobering statistic, especially considering that since the start of the year, the market value of software companies globally has dropped by an average of 30%.
The problem is structural. AI is upending business models: what previously took months to develop can now be realized in days. This destabilizes old schemes based on per-user licensing, because AI agents increasingly replace human tasks, reducing the number of licenses needed.
The picture isn't entirely negative, at least for Europe. Unlike the United States, where software represents over 50% of the tech market, in Europe the share stops at 30%. The rest of the market consists of IT services provided by players specialized in consulting and system integration.
“Beyond this, four other factors differentiate the European market from the U.S. one” explains AlixPartners Partner & Managing Director Marcello Bellitto. “Regulation oriented toward greater control over innovation and greater protection of the workforce; greater complexity and fragmentation of technology stacks, built layer by layer over decades; the composition of the industrial ecosystem composed mainly of small and medium-sized companies compared to U.S. giants. Finally, there is a geopolitical issue—digital sovereignty—that is leading to greater awareness in data management.”
For AlixPartners, specialization makes the difference. Those operating in sectors like banking or healthcare, where data control is structurally higher, start with an advantage over generalist companies, which are more exposed in an increasingly AI-driven world.
The knowledge of processes
The paradigm shift tells of a transition from the ability to develop code to the ability to transform business processes. It's not a nuance but a transformation in how system integrators create value. “AI adoption involves radical change, which must be approached in a reasoned and conscious manner,” explains Partner Davide Antonazzo. “Our clients don't just ask us to develop a specific project, but to help them build an integrated strategy in a broader business context.” For Antonazzo: “The mid-market company that doesn't have the capacity to build an AI solution by itself relies on a partner. The transformation of operational processes that such an AI solution enables is a new business based on consulting, which runs alongside the existing software development business.”
This scenario includes Lutech, which generates around 50% of its margin from activities on clients' vertical operational dynamics. “Base AI technology will become a commodity,” says Giuseppe Di Franco, CEO of Lutech. “Competition shifts to the orchestration layer, to solutions that enable service interoperability and adaptability. Knowledge of core business processes is central. For us, it's already a more valuable asset than it was a few years ago.”
Di Franco cites a data point to frame the scope of change: according to recent Stanford research, AI adoption is happening faster than any other mass technology, including PCs and the internet. This adoption speed doesn't automatically translate into competitive advantage. “The ability to implement these solutions without additional lock-in, beyond everything IT has created in the past, is one of the guidelines to follow,” Di Franco continues. “It means moving from a sensationalist approach to a more realistic one, which requires planning, choice, and insertion into a broad development program of the company.”
For Lutech, the current trend isn't choosing which AI model to adopt. “There's a convergence between various models, now quite similar to each other,” Di Franco observes. “This reduces the gap between the U.S. and China. The real difference is no longer in technology but in execution. And in this, all business functions are part of the change, including management.”
The risk of “DIY” AI
But there's a dynamic to watch: AI gives the illusion that employees can do everything themselves: create agents, automate flows, build solutions independently. This is deceptive, especially in the enterprise sphere. “Artificial intelligence raises the level of complexity; it doesn't simplify it,” warns Lutech's CEO. “It introduces another element that must be managed. The consultant and the system integrator must draw the client's attention to security and data sharing issues, fundamental for an aware use of the technology.”
This is compounded by the challenge of human capital. For Di Franco, skills haven't evolved at the same pace as technology, creating a misalignment between opportunity and concrete know-how. He notes that in Italy there are 15 million people between ages 6 and 66 who lack basic digital skills. “AI could worsen this divide or, at best, become an accelerator to catch up with more advanced economies—both in terms of technology and skills. AI as a driver of growth for society.”
The architecture of the future is agent-based and requires independence. “The shift to agent-based architecture is now a technical necessity,” Di Franco continues. Customizing solutions “on top” of existing AI platforms allows for faster adoption and sustainability. But this also requires reflection on dependence on large international vendors. Not by chance, Lutech is investing in this direction with a project called BrAIn, designed to free AI architectures from the proprietary platforms of large global vendors, focusing on integration and concrete business support. “Companies destined to emerge from this phase won't be those that invest the most in technology, but those capable of transforming it into processes, competencies, and measurable value. A technology-driven vision isn't enough: a business case-driven approach is needed. Before being technological, this is a cultural and managerial transformation,” says Di Franco.
The Italian mid-market
There's a slice of Italian productive ecosystem that risks falling behind. Alessandro Geraldi, Group CEO of Impresoft, knows it well. Impresoft Group works primarily with the mid-market, about 70% of its revenues come from this segment, and from this position observes a recurring contradiction: high interest in AI but almost absent governance.
The problem is that in medium-sized companies, artificial intelligence is entering through the back door. Employees use consumer tools, chatbots, assistants, text generators without the company having a system to control what is shared, with whom, and on which platforms. Sensitive data, internal processes, strategic information: “Everything can end up in systems that don't guarantee protection,” Geraldi explains. “If you don't manage this, your employees will use AI anyway, but you'll have no control over what they're doing.”
Added to this is subtle but pervasive cultural pressure. Consumer AI has accustomed everyone—managers, employees, entrepreneurs—to the idea that you just need to ask to get results. That technology is simple, immediate, infallible. In a business context, this perception is dangerous. “It leads to delegating decisions without verifying, accepting outputs without checking, building workflows on probabilistic foundations in contexts that require certainty,” Geraldi continues, and he's clear on one point: “Companies live in a world where every statement must be demonstrable, every process must be replicable, every result must be verifiable.” AI doesn't change this need; it complicates it. And those who don't understand this may risk having to redo everything from scratch.
Accompanying change
Impresoft's response to this scenario is a strategy that guides companies from initial assessment—to understand where they're starting from, where implementation can be most useful, and what sector-specific risks should be considered—through implementation and ongoing governance. A distinctive element is the cross-cutting control system: who has access to AI tools, how much is spent, what works and what doesn't, where there are anomalies, how usage patterns change over time. Not a decorative dashboard, but an operational tool for those making decisions in complex contexts. In this vision, governance isn't a bureaucratic constraint: it's the condition that makes AI sustainable in the long term, scalable, and above all useful rather than costly and fragile.
What makes this approach credible, according to Geraldi, is something Impresoft has built over decades of work: proprietary software and vertical knowledge of the sectors it operates in, from manufacturing to retail, from fashion to luxury to healthcare. It's precisely this process expertise that transforms into competitive advantage when AI comes into play. “You need to know how a credit management flow works, or a regulatory approval flow, or a clinical diagnosis chain, to be able to verify that the output is correct, coherent, and reliable. Those who know only the technology but not the process can't detect a hallucination, can't understand when something doesn't add up, lack tools to distinguish a plausible result from a correct one. We know the data system, we can semanticize the process. This allows us to benefit from what AI produces and also to detect any errors and ensure the implementation withstands the test of facts.”
Finally, there's a structural issue that for Geraldi affects the entire Italian productive ecosystem. Medium-sized enterprises, which represent the backbone of the country's economy, will hardly be able to attract the talent needed to manage an AI-based transformation autonomously.
“There's an element of attractiveness, scale, and ability to offer growth paths that more specialized profiles seek elsewhere. Complexity grows, skills specialize, and those without a partner with an in-depth understanding of operations risk finding themselves alone, facing decisions they can't evaluate. In this scenario, the IT partner's role changes radically: it's no longer that of the technical executor implementing decisions made elsewhere, but that of the strategic partner who arrives prepared with solutions already tested in similar contexts.
In practice, this approach is the difference between making noise about AI and innovation that generates real value.”
Accelerator of value creation
One sector, more than others, has already reckoned with the need to measure every technological investment in terms of concrete and verifiable return: finance. Salvatore Borgese, CEO of TAS, knows it from the inside. The TAS Group has been operating for decades at the heart of banking information systems: core banking, payments, card systems, capital markets. From this privileged position, it observes the transformation brought by AI with different eyes to those from less regulated sectors, who are less accustomed to data governance. The starting point, for Borgese, is a principle that sounds simple but has profound implications: “AI should not replace the architecture in use but accompany it, integrate with it, and enhance it. It's not about eliminating everything existing to start over. An expensive, risky, and often pointless operation. You must accompany artificial intelligence with assets already present in the company, including people's competencies.”
Managing the change in mission-critical processes and procedures like those managing current accounts or payment systems isn't like updating a user interface. Malfunctions or inaccuracies have immediate, measurable consequences regulated by law. This complexity, however, is also a form of advantage. “The financial sector was already accustomed, long before AI's arrival, to evaluating every technological project in terms of business impact,” Borgese explains. “Banks no longer buy technology but solutions with expected results. This means that when AI entered the operational scope, it found organizations prepared to think in terms of generated value.”
It's on this basis that TAS is evolving its business model in three directions. The first is building scalable, data-oriented architectures capable of transparently measuring solution impact to build pricing models aligned with the client's objectives. The second is value sharing: in some cases, TAS is exploring models where part of the benefit generated—in terms of higher revenues or lower operating costs—is redistributed between supplier and client, transforming a transactional relationship into a partnership. The third is scalability through agents. Here, Borgese shifts perspective: “The often widespread mistake is viewing AI agents exclusively as efficiency tools. As a technology that does the same things with fewer people. This reading is partial and in the long term risks being misleading. Agents don't simply replace human resources; they enhance and multiply them.”
The concrete example he provides: a financial institution with a thousand promoters that adopts AI agents could acquire capacity and coverage with a 5 to 10 multiplier. “Each person is amplified, not replaced. The result isn't a labor cost saving but a growth in the commercial-professional capacity to reach customers that would have been previously impossible to support economically.” This distinction between efficiency and value multiplication is central to what Borgese identifies as the great opportunity for the financial sector.
“Banks are focused on efficiency and have oriented much AI use toward the reduction of operating costs, automation of repetitive processes, and acceleration of workflows. Less focus is being paid to the second dimension: using AI to generate new business opportunities, access markets that were previously too expensive to reach, and build business models that without artificial intelligence wouldn't have been sustainable.” That's where, according to Borgese, the greatest value lies. It’s also the most difficult work to do, “because it requires a cultural change alongside the architectural change.”
But where will all the efficiency generated by AI go? “If it translates exclusively into cuts, organizational reduction, cost compression, the net effect on the economy could be very negative. If instead it's reinvested in securing new business, the ability to access new markets, and attracting more customers, then it can become a real growth engine.” Much will depend on the choices companies across sectors—not just tech—make in the coming years.
