Agentic AI Does Not Wait for Multinationals: the Change for Italian SMEs
When discussing agentic artificial intelligence, the examples provided are almost always the same: tech giants, investment banks, and multinationals with IT departments comprising hundreds of people.
This is misleading: it also concerns companies with 30, 50, or 150 employees that represent the backbone of the Italian production fabric.
And yet: Italian companies have a more widespread cloud infrastructure than the European average1https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/.
In fact, 75.6% of companies purchase paid cloud services, the second-highest figure in the Union after Finland, compared to an EU average of 52.7% (Eurostat, 2025)2https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20260203-1.
AI adoption, however, is still lagging</strong>: 16.4% compared to the European 20%3https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20251211-2. The gap suggests that the bottleneck is neither technological nor budgetary: the basic infrastructure, in many cases, is already available. The issue is therefore organizational.
76% of Italian SMEs have not invested nor plan to invest in AI4https://www.ai4business.it/intelligenza-artificiale/pmi-piu-spesa-digitale-ma-poca-visione-sullai-il-nodo-resta-competitivo/, and only 7% have initiated structured training on the subject (Digital Innovation Observatory in SMEs, Politecnico di Milano). Adoption has more than tripled in two years (from 5% in 2023 to 16.4% in 2025 (Istat)) but a clear gap remains based on company size: 53.1% in large companies compared to just 15.7% in Italian SMEs.
Among companies that evaluated AI without adopting it, nearly six out of ten (58.6%) indicate the lack of internal skills as the primary obstacle. This is good news: the window of opportunity for those who move methodically remains wide, because the majority of direct competitors have not yet taken action.
What is an AI Agent?
It is worth clarifying a point, as the confusion between “chatbots” and “agents” fuels much of the skepticism among entrepreneurs.
A chatbot answers a question.
An AI agent perceives a context, plans a sequence of actions, and interacts autonomously with company systems (ERP, email, CRM…) to complete a task, requesting human intervention only in case of an exception.
The difference is not cosmetic: a chatbot helps write an email; an agent writes it, sends it, updates the CRM, and only reports anomalous cases.
Where it Really Works for a Small-to-Medium-Sized Company
Among the most cited applications in consultancy practice for an SME, five areas recur more than others:
- document management and back office (data extraction from invoices, delivery notes, and orders, automatic reconciliation with the ERP system);
- customer service (autonomous management of standard requests, with escalation only for critical cases);
- sales (lead qualification and personalized follow-ups);
- suppliers and supply chain (monitoring of contractual deadlines, comparison of offers, early reporting of bottlenecks);
- operational reporting (real-time aggregated data instead of manually prepared weekly summaries).
Regarding estimates of economic return, industry figures are circulating that do not rely on a public methodology or a representative sample of Italian SMEs, and should be taken with caution. The practical principle remains valid: measure the results of your pilot against a clear baseline before scaling.
Why Method Matters More than Technology
A multinational company can afford for a pilot to fail and try again with different budgets or teams.
A SME, almost always, can only bet once: if the first attempt disappoints, the subsequent project never starts.
For this reason, a different method must be adopted for SMEs, namely treating adoption as an evolutionary project.
The methodology exists, and it is called Project Management.
It is not a small-scale transposition of what large companies do.
In fact, in SMEs, the following process should be followed:
- choose a high-impact but low-complexity process;
- define two or three measurable KPIs first (to understand time saved, error rate, and unit cost);
- plan the project, create and engage the team, and entrust it to an appropriate project manager with clear and defined responsibilities. The project manager can be someone who knows the process, as those who live the activity are best placed to judge if the tool produces real value or just plausible outputs;
- test it on a subset of real cases for a few weeks and scale only after validation.
A question before the method: the culture of change
Even before the process and the budget, there is a more uncomfortable question: how much is the urgency to evolve truly perceived within the company?
If a process that worked yesterday also works today, the drive for change struggles to find energy. This is a legitimate reasoning: the cost of a missed evolution often remains invisible until it is made visible by a faster competitor, a customer asking for something that can no longer be offered, or a margin that erodes without an apparent cause.
Then there is a second question, even before the one about AI: is the organization capable, as a whole, of tackling process innovation?
Introducing an agent into a workflow is no different, in organizational substance, from any other change: it is a project, with an identified manager, periodic progress reviews, real support for those leading it, clearly assigned responsibilities, and the tools to exercise them.
Many Italian SMEs have never needed this discipline until now: growth has relied on rapid decisions by the entrepreneur or a few key collaborators, without formalizing project management roles. A model that holds up until innovation requires being followed over time, measured, and corrected along the way.
The fundamental question, then, is not only whether one has the skills to use AI, but whether the culture of evolution is already a consolidated practice or needs to be built from scratch alongside the first project launched.
The Real Bottleneck: People, Not Software
The 2026 Istat Annual Report confirms that qualified human capital correlates with adoption: companies with more graduates show an AI adoption rate 8.8 percentage points higher, and those with more technical specialists 5.2 points higher. SMEs struggle to bridge the gap not for lack of will but due to concrete obstacles: time taken away from production, training offers tailored to large organizations, few contacts perceived as credible, and training lacking verifiable certificates.
The investment priority for an entrepreneur is therefore not the software but the skills of the person who will lead the first pilot, also leveraging incentive schemes such as Interprofessional Funds or the New Skills Fund.
Where to Start
In summary:
- map internal processes and choose a high-impact, low-complexity candidate, avoiding critical systems;
- define two or three measurable KPIs in advance;
- entrust the project to those who know the process best;
- test it on a real subset for a few weeks; measure the results before extending;
- reinvest part of the generated value into training before adding a second automated process.
The gap that separates Italian SMEs from more structured companies today is not, ultimately, a technological gap. It is a gap in method and organizational preparation: the type of gap that targeted support can bridge before it becomes a requirement imposed from the outside, by a client, a tender, or a regulation.
In this regard, experience counts as much as method. Roncucci&Partners has been supporting Italian SMEs in digitalization and innovation paths for years, applying the same rigor to AI transformation that has guided hundreds of growth and internationalization projects: process mapping, definition of measurable KPIs, and targeted training for those leading the pilot. Not standard solutions, but support built on the organizational reality of each individual company. To learn more: https://www.roncucciandpartners.com/consulenza/digitalizzazione/
Related Articles:
