Phocuswright Europe 2026: Eclectic panel offers wider perspectives on agentic commerce

Representatives from a fintech innovator, an AI-native corporate travel and expenses platform and an enterprise technology behemoth shared their world view on agentic commerce at this week’s Phocuswright Europe event in Barcelona.

The session kicked off with a short presentation from Pablo Laucirica from Microsoft Advertising, who introduced the audience to Microsoft’s recently launched “three eras of the web” strategy. The three eras are the human web, the LLM web and the agentic web, which serve to help me find it, help me choose it, and do it for me, respectively.

“The agentic web fundamentally changes the architecture of the web that we know today,” he explained. “The web today is built for humans, not agents. An agent doesn't get stimulated by an image or respond to a call to action. Agents are built to analyse tons of information, they are fast, they are precise, they are rational, and they will follow exactly the prompt that we are instructing them to follow.”

He crystallised one of the recurring themes across this year’s Phocuswright Europe, telling attendees, “if your business is not able to kind provide the information the agent wants to read, then probably you're not going to be visible, so you might end up not playing a role in this agentic economy.”

At the same time, he also revealed that Microsoft’s approach is that “the three webs are going to co-exist.”

He also raised a potential headwind for businesses in light of this co-existence. “Today, we think around 80% of websites block bots but, guess what, in many cases, agents are considered as bots. So we need adjust the infrastructure layer to enable agents to read the information and cite it as the user is actually searching for it.”

Laucirica then joined Stanislav Bondarenko from Revolut and Nikita Miller from Perk in a panel discussion moderated by Mitra Sorrells from Phocuswire, who targeted the discussion around the specific role of agentic AI for booking travel. She teed up the conversation with Phocuswright data which shows “AI is clearly becoming a research layer, but the booking has not quite taken a hold just yet.”

Miller agreed, while pointing out “a lot of the data that we're seeing is very much on the consumer side, which is very different from the pace at which businesses and companies are actually adopting.”

As chief product officer for the corporate travel and expenses platform, her current focus is to develop agentic AI products whose purpose is to build trust - trust with the corporate traveler and trust with the CFO.

“As with all product management, it's ultimately about meeting customers where they are but also laying the infrastructure for where we believe things are going to be,” she said.

“Most of our customers aren't there [booking with agentic AI] yet. So a lot of what we're doing is bringing AI into our product in a way that starts to build trust.”

For example, every time someone sends an approval, Perk shows whether it was in policy or not, but also shows what the alternative options were, where Perk might have saved the company some money. “That starts to build trust about the capabilities and how we're using the data in meaningful ways to drive value for the business,” she explained.

She added that Perk is helped by having 10 years’ worth of structured data for individuals and companies. Bondarenko’s team also has access to data - the spending patterns of Revolut’s 75 million account holders. He is responsible for all non-financial AI products at Revolut, which includes hotels, experiences and its loyalty scheme. He is building an AI agent which he described as “an octopus, with tentacles, reaching out to different parts of the app like a regular chatbot but which also supports users with proactive tips and notifications.”

The agentic AI is hooked into to “extremely rich” transaction data which can deliver deep personalization. “The data is extremely rich,” he said. “We know not only what the users say they want, but also what they actually vote for with their money, as well as knowing where they buy. This understanding is how we recommend the next best experience to the user.”

One aspect of the agentic user experience which the moderator raised with Laucirica was advertising. “When a Copilot agent returns one answer to a user's query instead of a page of links, what does an ad even look like in that environment?” Sorrells asked.

His reply suggested that it wouldn’t be much different from what we have on the web today, and reiterated that “the number one principle is trust.”

“In the new LLM chat reality, we have still two types of spaces, as in the traditional web or search engines,” he explained. “Organic results are not going to be altered with any advertising. What we’re moving towards is curated conversations based on precise information about the customer, their interest, their needs, etc. From a marketing standpoint, is a super powerful conversation to be in, so there’s going to be placements for advertising alongside the organic.”

“The two will coexist, but they will not merge,” he concluded. “Because the moment they merge, then you are basically breaking that first principle of trust.”