Italian companies invest in the power of artificial agents

Following the advances made with robotic process automation technologies and traditional and generative artificial intelligence tools, the next evolutionary leap for Italian companies is agentic AI, meaning autonomous systems capable of proactively planning, acting, and adapting, collaborating with each other and learning from data. Eva Terni, Managing Director of Insights & Data at Capgemini in Italy, is clearly convinced of this, and it stems from a specific perception. "Italian companies," she explained to Il Sole 24 Ore, "are evaluating whether to develop solutions internally or adopt market models, based on the level of customization required, data sensitivity, and the desired competitive advantage." Despite the concerns raised by some analysts (specifically, Gartner) regarding the likelihood of failure of agentic AI projects, the path is effectively mapped out, even if we are only at the beginning.

A recent report by Capgemini Research Institute (“AI in action: How Gen AI and agentic AI redefine business operations”) confirms that artificial intelligence is no longer a laboratory experiment, but an already operational lever for reducing costs, increasing efficiency, and transforming key functions such as supply chain and procurement, finance, human resources, and customer service, with expected benefits in terms of reduced operating expenses of around 30%.

According to the research, 60% of organizations in Italy have already developed a roadmap for scaling AI adoption, supported by ROI metrics and solid use cases. However, the use of agents, despite strong interest, is currently limited: only 1% of Italy's largest companies have already implemented these tools in their processes, while 43% are planning to adopt them within 2–3 years. The rest of the world, however, is ahead of the curve, with projects related to these tools expected to grow 48% by the end of 2025: currently, 21% of companies with revenues of over $1 billion are already using agentic AI solutions and multi-agent systems (almost double the figure from last year, when the figure was 10%), 16% plan to adopt them within the year, and a further 31% plan to use them in the next two to three years.

"These data," Terni continues, "confirm a strong growth trend in Italy as well and indicate that the market is at a turning point. Companies that have so far experimented with a variety of independent use cases are preparing to industrialize solutions with a concrete impact. To do so, however, requires a shift in mindset, which means rethinking processes, enabling new ways of working, and adopting new technologies, as well as knowing how to manage data as a reliable and strategic asset."

The transformation, as the report continues, involves virtually all sectors, albeit with different priorities. Evidence gathered in Italy suggests that the industries most likely to adopt Gen AI and agentic AI, albeit with implementations still in the exploratory phase, are manufacturing, high-tech (where 45% of companies have already adopted agents), pharma & healthcare, energy & utilities, and retail. Insurance and banking, as the Capgemini manager explains, are focusing on intelligent automation to reduce operating costs (cost savings in finance & accounting functions can reach 40%), simplify complex core processes, and improve efficiency. Automotive and manufacturing in general are focusing on rapid innovation to accelerate new product development and optimize production, while in the consumer products sector, large-scale automation is aimed at improving the customer experience (the use of agents is estimated to increase customer satisfaction by up to 44%) and strengthen consumer loyalty. The path, as mentioned, is clear, and in a scenario reflecting a certain caution among Italian companies, at least two other findings from the study are revealing. The first shows that 52% of Italian companies will have increased their Gen AI budget by 2025 (compared to 62% globally). The second confirms the central role of hyperscalers as suppliers of the adopted AI models, with a percentage reaching 87%, while only 1% of companies adopt hybrid approaches combining proprietary and open-source models.