On Thursday, October 15th, at the Nicolaus Hotel in Bari, we held our seminar as part of the 1st FARE National Conference: “What if an AI Agent were the new typewriter for public tenders?”.
The room was packed beyond expectations — so much so that chairs had to be added after the seminar had already started — a clear sign that this topic genuinely interests the whole supply chain today: contracting authorities, economic operators, and legal experts alike.
Below you’ll find the full video of the seminar and a recap of the main points raised by our four speakers.
The contracting authority’s perspective
The session opened with Gianluca Viganò, President of ALE, Head of Clinical Engineering and Interim Head of Procurement at ASST Spedali Civili di Brescia, who posed a simple question: when we talk about AI applied to procurement, where do we actually want to get to?
His answer centered on three concrete goals: process optimization, support for procurement officers, and guided data digitalization. Some applications — such as automated market research and product catalog standardization — are already being validated in Brescia; others, from predictive needs analysis to support for evaluation committees and monitoring during contract execution, are the next step.
A key takeaway from his talk: these tools are assistants, not replacements. “A high-speed technical co-pilot,” as he put it, reminding everyone that final responsibility and decision-making always rest with the procurement officer — and that using these tools well requires serious investment in training, starting with so-called prompt engineering.
Regulation actually arrived on time (for once)
Amanda De Cosmo, Partner at Studio Brugnoletti&Associati, offered a perspective rarely heard in these conversations: unlike almost every other technological innovation, the law did not arrive late for artificial intelligence. The EU AI Act, Italian Law 132/2025, Article 30 of the Procurement Code: a regulatory framework that doesn’t hold back adoption but actively accompanies it, setting a few clear rules.
At the heart of it all, De Cosmo explained, is the so-called human reservation principle: responsibility for the decision always rests with a person, and it must be demonstrated in practice — simply “signing off” on what the AI produced is not enough; the reasoning must genuinely be made one’s own. She cited a recent ruling from the TAR Marche regional court on exactly this issue, and offered a practical recommendation for every contracting authority: adopt an internal regulation covering procurement, traceability, and training — a legal safeguard, but also a tool for greater operational effectiveness.
The SME perspective: closing the gap with multinationals
Federico De Rovere, CEO of Corios and president of the Lombardy association of medical device suppliers, brought the perspective of a company competing daily against tender offices with dozens of people, while having one or two of his own. For an SME, AI agents aren’t a technological luxury: they’re a way to level the playing field against better-resourced competitors — in tender research, feasibility analysis, and compiling information that repeats, tender after tender.
A point rarely considered: the more SMEs manage to participate, the more contracting authorities have access to a genuine market, not just the usual handful of large competitors — and SMEs, De Rovere noted, are often the real source of innovation that larger companies later acquire. Like the other speakers, though, he stressed the same point: human supervision remains necessary, because AI “misses pieces” and is, in his words, “extremely accommodating.”
From chat to agent: how it actually works
Closing the session, Luca Visconti, CTO of TenderBrain and CEO of OpenGate, explained the technical distinction underpinning everything: “chat answers, the agent works.” An agent plans, uses tools, checks the result, and repeats the process with a patience no human operator would have — analyzing a thousand documents one by one, without skipping any.
Visconti also shared the story of how TenderBrain came to be, born from merging the domain expertise of BEC’S (preparing tender responses since 2011) with the technology expertise of OpenGate (software development since 2004, Google Cloud and Scaleway partner). He emphasized two technical pillars that are non-negotiable when working with AI in sensitive fields like healthcare: zero data retention (no customer data is ever stored by the models) and European data residency. He also flagged a well-known pitfall for anyone working with these tools daily: AI makes mistakes “with confidence” — which is why source traceability matters, and why, in the most sensitive cases, you need agents that actually verify the work of other agents.
Thank you
A sincere thank you to Gianluca Viganò, Amanda De Cosmo, and Federico De Rovere for the conversation, and to FARE for hosting such a well-attended conference. If you work in public procurement and want to see TenderBrain in action on your own tenders, we’re here.