Oracle AI World Tour Milan: From Agentic AI to Digital Sovereignty, Pragmatism as a Strategy
Oracle AI World Tour Milan: From Agentic AI to Digital Sovereignty, Pragmatism as a Strategy
There is a precise moment when a corporate event ceases to be a communication exercise and becomes a snapshot of the state of the art. The Oracle AI World Tour, during its Milan stop on April 1, 2026, was that moment. Not so much for the quantity of announcements—which were abundant—but for the coherence of the message: enterprise AI is not an add-on module but an architectural principle that permeates the entire technology stack.
The underlying theme was set by Cormac Waters, SVP Applications of Oracle EMEA, with a metaphor that has the merit of clarity: the transition from propellers to jet engines did not change the mission of aviation; it changed what was possible. In Oracle's vision, AI is doing the same for businesses: it does not replace skills; it elevates them. Less time reacting, more time deciding. Less process management, more outcome governance.
This narrative that Oracle constructs revolves around three pillars: trust, choice, and scale. It finds its concrete translation in a series of announcements made the previous week in London and contextualized in Milan for the Italian market, as confirmed by the interview with Andrea Sinopoli, Vice President Cloud Technology Country Leader for Italy.
From Agentic Applications to Outcome Systems
The most cited data point of the day was the adoption of AI in Fusion Applications: one year ago, less than 10% of approximately 15,000 Fusion customers actively used AI features; today, that percentage is close to 70%. A leap that Waters deemed surprising, but which deserves a more granular reading. A good part of these activations pertains to basic functions: customer service assistants, HR automations, reconciliation in finance. The more interesting question is not how many are using AI, but how many are actively redesigning processes.
In London, Oracle announced 22 new agentic applications for the Fusion platform. These are not just simple agents or chained workflows, but applications built to achieve specific outcomes, capable of reasoning, predicting, and acting continuously. Kerry Gallagher, head of Fusion Applications, summarized the paradigm shift with an operational concept: the transition from a “record system” to an “outcome system.” The agent does not merely close a process; it has a goal and continues to operate until it reaches it.
To measure the effectiveness of this approach, Oracle introduced dashboards with KPIs dedicated to the performance of agents: execution times, resource optimization, ROI. Andrea Sinopoli framed the issue with pragmatism: “We are industrializing the measurement of agent performance because without concrete metrics, there is no enterprise adoption.” This point is not trivial: in a market where many companies are still in the exploratory phase, the ability to demonstrate measurable return makes the difference between a pilot project and a large-scale deployment.
The Database at the Center: Data, Security, and Determinism
The second focus of the day was the data platform. Oracle presented the AI Data Platform, an architecture that combines the Oracle database (increasingly multimodel, multilanguage, and multimodal) with tools to integrate heterogeneous data sources, including non-Oracle ones, into a single governed access point.
The most significant announcement concerns agentic capabilities integrated directly into the database: a framework for running agents close to the data, an agent factory for building them, and an autonomous vector database for indexing unstructured data. Kumar Maddipatla, head of database technologies, emphasized a precise architectural point: when the agent lives next to the data, the database's security policies are automatically applied, whether the query is generated by an agent, a natural language interface, or a human operator.
There’s also the issue of determinism. In a world where generative AI is probabilistic by nature, Oracle introduced Trusted Answer Search, a component that allows natural language queries on the database with deterministic results. This is a point that Sinopoli insisted upon in the press session: “The ability to add a deterministic component within the database model is a unique issue that we can now bring to market.” For regulated sectors, where traceability and auditability are not optional, this combination of AI and determinism represents a functional requirement before being a competitive advantage.
Sisal, Generali, UniCredit: Three Paths, One Direction
The most concrete session of the day was the roundtable with three Italian clients already using Oracle AI in production. Three different sectors, three varying levels of maturity, but a fundamental convergence: enterprise AI works when it creates measurable competitive advantage.
Mario Martinelli, CTO South Europe and Africa of Flutter Entertainment (the group that controls Sisal, Snai, and PokerStars in Italy), portrayed a radical transformation: from one million active online customers per year to 2.4 million per month, with an infrastructure redesigned on microservices and Oracle cloud to handle enormous transactional peaks. Every sports betting event generates updates on around 3,000 types of bets in real time, multiplied by hundreds of concurrent events. Sisal has developed over 150 AI use cases, 25% already in production, with applications ranging from risk management to responsible gaming. Martinelli also hinted at a project to migrate the ERP to Oracle Fusion specifically to leverage agentic AI integrated into back-office processes.
Alessandro Protasoni, Head of Group Organization and Workforce Planning at Generali, presented a different perspective: that of organizational impact. In a group present in 50 countries with 88,000 employees, the question is not only which use cases to activate but how to rethink the organization around the cohabitation of humans and machines. Protasoni advanced a radical thesis: in future RACI (Responsible, Accountable, Consulted and Informed) schemes, the R for “Responsible” will tend to belong to an agent and no longer a person. A position that raises concrete questions about governance, especially in regulated sectors like insurance.
Gabriele Chiesa, CTO of UniCredit, described a structured approach along three lines: individual productivity tools with AI, redesigning banking processes, and chatbots for direct interaction with customers. The most significant element, however, pertains to infrastructure: UniCredit is migrating around a thousand applications to Oracle Exadata Cloud Service, which runs on Google Cloud. An emblematic case reflecting Oracle's multicloud logic, which sells its cutting-edge technology also within the infrastructure of a direct competitor.
Digital Sovereignty: From Geographic Perimeter to Operational Model
The theme of sovereignty permeated the entire day but took on clearer contours in the press session. Waters traced a precise evolution: from European sovereign cloud, operated by EU personnel and under European legal entities, to national sovereignty, up to the Alloy model, where a local partner takes the Oracle cloud in white label and provides it as its own service.
In Italy, this model is already operational with the partnership between TIM and the Strategic National Pole: the infrastructure is Oracle, but management and customer interaction are under a fully Italian entity. Sinopoli clarified the mechanism in a video interview with Edge9: “There is the possibility of not precluding the use of world-leading technologies while simultaneously providing the guarantees that address the issue of sovereignty.”
Waters added a counterintuitive but relevant element: to be sovereign, one must be global. A Middle Eastern customer, he recounted, asked to physically move its data to Europe while maintaining sovereignty due to geopolitical instability in its region. Infrastructure flexibility, in this scenario, becomes a survival requirement before it is a compliance issue.
The Knot of SMEs and the Infrastructure Question
From the Q&A with the Italian press, a picture emerged of the most active sectors in AI in Italy: public administration and healthcare (driven by PNRR funds), gaming and betting due to high transactional volume, supply chain, and logistics. However, the risk of polarization is real: these sectors are advancing rapidly while the fabric of SMEs risks falling behind.
Sinopoli acknowledged the problem, focusing on two levers: the simplification of the integrated stack (the message “One Oracle” that increasingly recurs in corporate communication) and the partner ecosystem as a channel to reach smaller realities. The logic is that large companies invest first, obtain results, and the same consultants and system integrators then transfer skills and experiences to the mid-market.
On the infrastructure front, Oracle continues to invest in building data centers on a global scale. When asked whether geopolitical tensions and rising energy costs are slowing these plans, Sinopoli responded cautiously but unequivocally: “The willingness to be ready when the wave of AI becomes even more disruptive is our priority. If one is not ready, it will be too late.” A declaration of intent that, in the absence of specific details about Italy, remains, for now, a strategic promise rather than an operational plan.
The Oracle AI World Tour Milan presented an image of a company that has found a coherent narrative around enterprise AI: vertical integration of the stack, agents as a new application paradigm, sovereignty as a flexible operational model. The challenge, as always, lies in the distance between vision and execution. The cases of Sisal, Generali, and UniCredit demonstrate that for large enterprises, that distance is shrinking. For the rest of the Italian manufacturing fabric, the game is still on.