Agentic AI: why companies should build AI, not just use it

According to Emanuele Frontoni, professor of Computer Science at the University of Macerata, companies should not limit themselves to using artificial intelligence: they should become creators of AI, through agentic AI. He said so in an interview with AI4Business.¹

What Frontoni proposes

Frontoni is co-director of the VRAI Lab (Vision, Robotics, Artificial Intelligence), a research laboratory linked to the university. In his view, AI applied to business is not limited to LLMs (Large Language Models): it is a broad and varied set of technologies, of which agentic AI is a part undergoing rapid evolution.¹ In the Italian landscape of digital transformation, he argues, the game is not won through passive adoption of general-purpose tools. What matters is companies’ ability to produce original, proprietary technology.¹ Intelligent agents, he says, make it possible to move from being simple users to being creators in their own right. For this reason, in his view, companies need to be pushed to develop solutions in-house.¹

The CappellAI case in the Marche hat district

Frontoni cites the CappellAI project, born in the Marche hat district, which he says alone produces about 75% of Italian hats. Working with local companies and with the support of his laboratory, he explains, they are gathering what he describes as the largest dataset in the world dedicated to hats.¹ The idea, he says, is not to adapt a generic AI created elsewhere, but to build a specialized one together with those who design and make hats for generations.¹ For Frontoni, the novelty of agentic AI is that a domain expert, for example in logistics, manufacturing or quality control, can instruct a machine in natural language. In this way, they transfer their craft knowledge into algorithms.¹ In his view, many traditional manufacturing companies think that adopting AI means buying software or a general-purpose model trained by others. For Made in Italy, he points instead to the path of highly vertical agents and AI, that is, models tailored to individual production and management processes.¹

Doing better, not just faster

Frontoni maintains that companies should use AI to do better, not just faster. The most solid advantage, in his opinion, is improving activities and operations, for example through more effective forecasts and increasingly specialized foundation models.¹ In product ideation or strategic planning, he argues, what is needed is more quality and effectiveness, not more speed. Text rewriting and chatbots, he says, perhaps cover only 5% of high-value manufacturing processes.¹ As an example, he cites sales forecasting in mechanical engineering or fashion: predicting the season’s orders accurately, he explains, does not mean doing the same thing more quickly.¹

A more local and European AI

The professor also calls for the development of AI in companies to follow a European dimension, as an alternative to American and Chinese technologies. Italian companies, he says, should aim for an increasingly local AI, with smaller and more specialized models compared to those of the large operators that dominate the market.¹ Frontoni also holds the position of vice president of the Human AI Foundation, recently established on the initiative of the Fondazione Cassa di Risparmio di Jesi together with the VRAI Lab association. Since 2021 his name has been included in the Stanford/Elsevier ranking of the World’s Top 2% Scientists, which places him among the 2% most cited researchers in the world on AI applied to industry.¹

What changes

The thesis shifts attention from buying ready-made tools to building tailor-made systems, with the knowledge of those who work in the sector at the center of the project. In the case of hats, this takes the form of data collection shared between companies and the laboratory.¹

This article was written by artificial intelligence from the sources cited, without human review. Automatic newsroom by Redazia.

Sources

Footnotes

1. Frontoni (UniMc, Vrai Lab): “Spingere le aziende a creare intelligenza artificiale nell’AI agentica”, AI4Business, 2026-11-10 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴