Where to Restart With AI in September 2026

Every September, Italian boards return to the same ritual: review what's been spent in the first eight months, discuss what's left of the budget, set priorities for the last four. For AI, this year, that ritual lands at a particular moment. The 2026 regulatory calendar has just stopped moving, after months of delays and corrections, and for the first time in a while companies have a window stable enough to plan properly, instead of chasing the latest amendment.

Worth being precise about what's actually settled, because more than one shortcut is circulating in conversations this week. Since August 2nd, the AI Act's Article 50 transparency obligations are binding, penalties are enforceable, and national supervision sits with ACN and AgID under Law 132/2025. The high-risk obligations for Annex III systems, on the other hand, have been pushed by the Digital Omnibus to December 2, 2027, and those for Annex I product-embedded systems to August 2028. The new prohibitions on non-consensual intimate imagery and AI-generated child sexual abuse material take effect December 2, 2026. In Italy, completing the picture, there is the decree given final approval on August 4th that adds artificial intelligence to the predicate offenses under Legislative Decree 231/2001, still awaiting publication in the Official Gazette.

Sixteen months is not an excuse to wait

The temptation, faced with deadlines pushed out to 2027, is to read them as time gained and not to be used yet. That's the wrong read. Sixteen months is exactly the amount of time it takes to build calmly what, done under pressure right before a deadline, always ends up worse and always costs more: a register of AI systems in use, a risk assessment process for every new use case, governance that doesn't depend on one person's memory but on a written, repeatable method.

The companies that arrive at December 2027 ready won't be the ones who waited until November 2027 to start. They'll be the ones who used the remaining months of 2026 to measure where they actually stand, before deciding where to invest in 2027.

Measure before you allocate, not after

This is where it connects to what we've written over the past few weeks on cost and adoption: companies that blow past their AI budget almost always do it because they scaled a project without first checking whether the organization could sustain it, and the 71% who started without reaching structured governance shares the same root cause, the absence of an honest starting picture.

Our AI Rating exists to answer exactly this, before any decision on budget or roadmap. It measures maturity across four dimensions, readiness, how prepared the organization is on data, skills and strategy, delivery, whether technical processes hold up in production and not just in a prototype, risk, how well governance covers compliance and ethics, confidence, how much the board and users actually trust what's been built. The result is a score from zero to five and a class from A to D, with critical gates that can't be compensated for: excellence in three areas out of four isn't enough to cover a serious gap in the fourth.

Three questions before closing the September budget

Before allocating the AI budget for the last quarter, three questions separate decisions made on evidence from decisions made out of habit. Which AI systems are already in use across the company, mapped with an owner and a risk level, not just the ones formally approved. Which use case, among those underway or proposed, has a success criterion written down before it started, rather than estimated after the fact once it's too late to stop. And if a regulator or a board asked today for an account of the company's AI governance, would the answer come from a verifiable document or from a shared feeling in a meeting room.

Whoever answers all three with confidence has likely already done the work described in this article, and is ready to talk about scale rather than foundations. Whoever hesitates on even one has just found the place to restart in September, before looking at the rest of the budget.

The advantage of measuring before everyone else

A structured rating also produces a ninety-day roadmap, with priorities ordered by impact and risk, not a generic list of recommendations. It's what lets a board arrive at December with a concrete plan for 2027, instead of the feeling of having chased the previous year one month at a time.

Organizations that choose to measure before investing in September arrive at the new year with an advantage that isn't visible right away but matters a great deal over time: they know where the money is going, and why.