Before AI Week: The Five Questions European AI Strategy Must Answer in 2026
AI Week Milano takes place on May 19 and 20, 2026 at Fiera Rho. More than 700 speakers, 17 thematic stages, and 250 exhibiting companies will fill what has become the largest AI-focused event in Europe. The scale reflects something real: AI is no longer a specialist conversation. It has moved into the boardroom, the procurement office, and the risk committee.
Events of this size tend to generate more content than signal. The useful question to ask before attending, or before reading the coverage if you are not, is not what the industry will talk about. It is what it still cannot agree on. The five questions below are the ones that will be in the room at AI Week, whether they appear on the agenda or not.
Why AI Week is a useful mirror for European enterprise strategy
AI Week is not an academic conference. It is built around the intersection of technology capability and business implementation, with vertical tracks covering manufacturing, healthcare, finance, marketing, legal, and public administration. The speaker list skews heavily toward practitioners: CIOs, AI product leads, and operations executives rather than researchers.
That profile makes the event a reasonably accurate proxy for where European enterprise AI actually stands in 2026. Not where the research frontier is, and not where the regulation says it should be, but where the people responsible for making decisions inside organizations have arrived. The gaps in that picture are as informative as the progress.
Question 1: Is AI adoption at 88% a real number?
The Stanford AI Index 2026 puts organizational AI adoption at 88%. It is the headline figure that has circulated through every board-level AI presentation since the report was released in April. It is also, on closer inspection, a number that describes a very wide range of behaviors.
An organization that has deployed a general-purpose AI assistant for a subset of knowledge workers counts toward that 88%. So does one that has integrated predictive maintenance AI across a network of industrial facilities, with full monitoring, governance, and fallback procedures. These are not comparable states of readiness.
The more revealing figure is what happens after initial deployment. Most research on enterprise AI in 2025 and 2026 converges on the same pattern: adoption at the pilot or departmental level is widespread, but scaling beyond initial use cases is where most organizations stall. The barriers are organizational and structural before they are technical: unclear ownership of AI outcomes, insufficient data infrastructure, middle management resistance to process redesign, and the absence of feedback loops between AI system performance and the teams responsible for it.
The question for AI Week is whether the practitioner conversations reflect this gap honestly, or whether the event gravitates toward the success stories that make for better presentations. Both are useful, but they require different kinds of attention from the audience.
Question 2: What does compliance-ready AI actually look like after the Omnibus?
Four days before AI Week, on May 7, the EU and European Parliament reached a political agreement on the Digital Omnibus on AI. The deal shifts the compliance deadline for most high-risk AI systems from August 2026 to December 2027, with AI embedded in regulated products like medical devices and industrial machinery receiving an extension to August 2028.
The compliance conversation at AI Week was already going to be complicated. It just became more so.
On one side, organizations that had built readiness programs around August 2026 now face a genuine strategic question: pause, continue, or accelerate? The answer is not obvious, and it varies significantly by sector, by the specific AI systems in use, and by the commercial dynamics of a company’s customer relationships. A software vendor selling AI tools to regulated industries will face procurement-driven compliance pressure long before any regulatory deadline. An internal operations team using AI for back-office process improvement faces a very different risk profile.
On the other side, the synthetic content transparency deadline in Article 50(2) moved in the opposite direction. Organizations generating AI-produced content for external audiences now have until December 2026, three months earlier than the previous estimate. This is the near-term compliance priority that many organizations are still not treating as one.
The productive version of the compliance conversation at AI Week is not about deadlines. It is about what governance looks like in practice when regulators, standards bodies, and industry are all still working out the details. That is the conversation worth having. Read our full analysis of the Omnibus deal.
Question 3: How do you govern AI you did not build?
The Foundation Model Transparency Index, which tracks how openly major AI model providers disclose information about their training data, evaluation methods, and safety practices, fell from 58 points to 40 points between its 2024 and 2026 editions. The providers whose models underpin the largest share of enterprise AI deployments are, on average, less transparent about how those models work than they were two years ago.
This creates a structural governance problem that most enterprise AI frameworks are not built to handle. Standard risk management assumes you can audit what you are responsible for. When the AI system at the center of a high-risk process is a model you accessed through an API, with no visibility into training data, known failure modes, or benchmark conditions, the audit trail has a gap at its core.
The EU AI Act addresses this through transparency obligations on GPAI providers and through the documentation requirements it places on high-risk deployers, but the practical gap between what the regulation requires and what providers currently disclose is significant. The question at AI Week, and the one that will define enterprise AI procurement decisions over the next two years, is how organizations manage that gap without either abandoning capable models or accepting unexamined risk.
Practical answers are beginning to emerge: contractual transparency requirements in vendor agreements, third-party model auditing, portfolio approaches that balance capability against explainability, and internal monitoring frameworks that compensate for what providers do not disclose. These are not elegant solutions. They are the workable ones.
Question 4: Where are the people?
The talent picture for enterprise AI in 2026 is more complex than the headline hiring numbers suggest. Entry-level developer roles have contracted in many technology companies as AI-assisted coding tools have changed what individual engineers can produce. At the same time, demand for roles that did not exist three years ago is outpacing supply by a wide margin.
AI governance specialists, people who understand both the technical behavior of AI systems and the regulatory, ethical, and organizational frameworks for managing them, are scarce relative to demand. The same applies to AI product managers with deep domain expertise in regulated industries, to data engineers who can build the infrastructure that enterprise AI requires at scale, and to the middle management layer that needs to redesign workflows around AI capabilities without either ignoring the tools or delegating critical decisions to them unchecked.
The skills gap is not primarily about knowing how to use AI tools. It is about knowing how to integrate them into processes in ways that preserve accountability, catch errors, and produce outcomes the organization actually intended. That kind of judgment is developed over time, through practice and failure, and cannot be hired in or trained over a weekend.
Question 5: What is Italy’s actual position?
Italy entered 2026 as the first EU member state with a national AI law (Law 132/2025, coordinated through ACN and AgID), a meaningful allocation of PNRR funding toward digital and AI infrastructure, and a manufacturing base that represents both a significant opportunity and a significant exposure to AI-driven productivity shifts.
The honest answer to where Italy stands is that the picture is uneven. Research capacity in AI is concentrated in a small number of university groups. The startup ecosystem is growing but remains capital-constrained relative to France, Germany, and the UK. Large Italian enterprises in manufacturing, finance, and retail have in several cases made genuine progress on AI adoption, while the broader SME population, which accounts for most of Italian GDP, is still at very early stages.
The PNRR allocations tied to digital transformation have moved more slowly than planned in many cases, and the absorption of funds has been uneven across regions and sectors. Italy has a regulatory framework, investment commitments, and genuine industrial depth. The question is whether those assets translate into competitive AI capability at scale, or whether they remain disaggregated across sectors and organizations that are moving at different speeds.
What to watch at AI Week, whether you attend or not
Three signals are worth tracking across the event’s output over the two days.
The first is which vertical use cases have crossed from pilot to production at scale. Not proof-of-concept results, but evidence of AI systems running in operational environments, at volume, with governance in place. The list of sectors where this has happened will tell you where the adoption curve has actually bent.
The second is how AI vendors are repositioning around governance and compliance. Exhibitors that have built compliance features, audit trails, and transparency tooling into their core offerings are signaling where enterprise procurement pressure is coming from. That repositioning is an indicator of what buyers are actually requiring, which is often more informative than what buyers say they require.
The third is what the labor market signals look like from the companies represented. Which roles are being hired for at scale, which are being eliminated, and what the internal reskilling programs actually cover will give a clearer picture of the organizational AI transformation underway than any panel discussion on the topic.
AI Week is a useful moment to take stock, not because it resolves any of these questions, but because it concentrates a large number of people who are working on them in one place for two days. The conversations that happen outside the formal agenda are often the more informative ones.
OneSynergy works with European organizations on AI strategy, governance design, and technology adoption. If you are attending AI Week and want to discuss any of these questions, reach out.
