Prada Taps Into Agentic AI For Procurement

The past five years were a culling season of sorts for the Prada Group. Over that period, the Milan-based luxury group engaged in more than 850 on-site factory inspections, including all-night stakeouts outside the facilities of subcontractors, the Financial Times reported earlier this year. As a result of this extensive audit, Prada terminated its relationship with 222 suppliers for allegations of worker abuse, which included excessive work hours and workers sleeping inside the factory.

While social and environmental audits have become table stakes, brands do not usually go to these lengths to inspect their complex supply chains. The report came out at a time when the domestic supply chain behind luxury fashion was under intense scrutiny by Milanese prosecutors for alleged worker abuse. While predominantly a labor story, it spoke volumes to the difficulty of staying on top of a sprawling supply chain, where seeking the origin of each material or component can lead to the discovery of another subcontractor in a huge network of manufacturers.

Prada’s portfolio, which includes Miu Miu and Versace, currently has over 1,000 suppliers for all sorts of requirements such as leather, fabric and packaging. It is in this context that Prada, a global leader in luxury fashion, is bringing in additional help—and its name is Iva, an agentic AI.

“Try to imagine,” said Fabio Francalancia, head of global procuring and cost management director, at Ivalua’s Procurement Innovation on Tour in New York earlier this week. “It’s 1,000 suppliers to be scrutinized, and it could take weeks of accurate review of documents, clouds, performance. That kind of activity is not only time-consuming, and you should need a lot of people to do it right and fast,” he said.

“It could be that you don’t find people that are available to do that kind of activity anymore. So it’s where AI, I feel, is entering into the equation to help,” Francalancia said in a short breakout session with other professionals, later calling Iva “the workforce you would never find.”

Iva stands for Intelligent Virtual Agent. It was developed by Ivalua, which has been doing source-to-pay (S2P) solutions for more than two decades now. The recent AI boom eventually led the company to explore generative AI, which resulted to Iva’s first version as an Intelligent Virtual Assistant that only does generative work, until later iterations led to what Iva has been since June: a singular AI agent that can learn a company’s rules and practices and do S2P on its behalf.

“The best analogy for people to think of Iva is thinking of Iva as this expert coworker,” Paul Zammit, who leads product marketing for Ivalua’s AI solutions, told Sourcing Journal after the session.

“Iva can help you onboard a supplier because it can create a supplier record and fill out the fields. Iva can launch a sourcing event and help you draft it and gather the requirements. It can help you execute a contract,” he said. “It can do all of these things because it uses the application like a person would.”

Prada is one of around 90 early customers using Iva, Zammit said. He declined to disclose names, but Ivalua serves a broad range of clients, from multinational corporations in luxury to major defense companies and federal governments.

An important distinction to remember with AI is that its results are probabilistic, as opposed to being deterministic that can make guarantees, Zammit said. It’s the humans who decide where to deploy the AI and what it can do without human approval. This also means the management involved is also deciding parts of the business that AI could not touch without permission.

“So I think that’s all part of an organization looking at their processes and acknowledging, ‘Okay, I want to deploy AI first of all where it’s going to deliver the most value, but part of it delivering the most value is de-risking it as much as possible, and that comes with realizing ‘Hey, this is probabilistic.’”

Prada’s business case before Ivalua

Francalancia, who joined Prada earlier this year, said its business case usually has two main drivers: compliance and value generation. “But there is a third driver that is the one that has been somehow declared in each coffee room, in each conversation with the users, which is speed,” he said. “People in the company expect that a procurement department is an accelerator in our company.”

So it’s usually a tradeoff, he said, where having one driver means having less of another. “We have always perceived, at least in my experience, control, compliance, and speed as a trade-off. I think this is no more the era of ‘OR’. It’s the era of ‘AND’.” This sense of having it all is made possible by the fact that the company stays in control of Iva.

Iva comes with Iva Studio, a dashboard which serves as the AI control tower of the company, where leaders can manage the skills Iva uses, the tools it can call, the external systems it connects to through Model Context Protocol, and the Large Language Models it runs on. “I feel I have the total control, and this is enabling us to try, adapt, and measure what will be the consumption of AI and AI cost for doing what,” said Francalancia.

Prada was brought in the event to talk about its experience using Iva. During the breakout session, Francalancia shared the stage with Alex Saric, Ivalua’s chief marketing officer, who echoed the point about control. “I think we’re all learning, but having that control will allow you to really maintain the avoid any kind of unpleasant surprises and kind of stay within budget while still driving the value and the excitement behind it,” he said.

It’s still early days and Prada and Iva are still learning, but so far Francalancia is seeing practical benefits. For example, in the case of training, he said it is hard for workers to dedicate hours of their workday outside of their core job so they can learn AI. The technology eases this burden. “We found out that AI can be the right trainer in the right moment in the right screen when you just need a bit of an help to do something obvious, but in the right way immediately for the first time, instead of trying to remember what I had been told one week ago in that room,” he said.

Some lessons, however, are more introspective, which is just as important. “The other challenge is be humble. each week there is something new. Each week you can hear from someone else something much more clever and focused,” he said.

“We really need to be a community around a project like this and to win from each other, learn what has been successful and why, and see if this can be adopted somewhere else—because if we try to learn by ourselves in silos, it will be too much.”

Source link