Pandora: How Much Freedom Should a Shopping Agent Have? (AI Summit LDN)

Pandora: How Much Freedom Should a Shopping Agent Have? (AI Summit LDN)

Author: Brave Software June 24, 2026 Duration: 22:02

Riccardo Arnaldi, Agentic AI Product Manager at Pandora, shares what it actually looks like to build and scale a brand-facing LLM shopping agent live with millions of customers. He explains why too much autonomy is a liability, how a deterministic routing layer keeps the AI from going off-script, and why traditional QA teams had to rebuild their processes from scratch. He also breaks down a real customer interaction that exposed the gap between automation and appropriate human sensitivity, changing how the team handles frustrated users.

Key Takeaways

  • Which metrics Pandora uses to measure success of Gemma, their customer-facing AI agent

  • What breaks when traditional QA processes meet agentic products, and how testing an agent differs from testing software

  • How a brand's voice gets engineered into an AI agent, including exact words it's allowed to use and the persona it borrows

  • What a frustrated customer's outburst revealed about the line between automation and human sensitivity

Guest Bio:

Riccardo Arnaldi is a product manager with 6+ years' experience building and iterating on digital products through data, experimentation, and AI. Currently, he owns end-to-end development and strategy of LLM-based shopping and customer support agents at Pandora. Previously, he drove X60M+ DKK in incremental revenue by scaling Pandora's website and CRM personalisation programme from the ground up across 7+ markets. An Italian based in Copenhagen, Riccardo is a sports lover and guitar strummer, and is always on the lookout for the next great travel destination (with hype-worthy local foods).

----------------------------------------------------------------------------------------

About this Show:

The Brave Technologist is here to shed light on the opportunities and challenges of emerging tech. To make it digestible, less scary, and more approachable for all!

Join us as we embark on a mission to demystify artificial intelligence, challenge the status quo, and empower everyday people to embrace the digital revolution. Whether you're a tech enthusiast, a curious mind, or an industry professional, this podcast invites you to join the conversation and explore the future of AI together.

The Brave Technologist Podcast is hosted by Luke Mulks, VP Business Operations at Brave Software—makers of the privacy-respecting Brave browser and Search engine, and now powering AI everywhere with the Brave Search API.

Music by: Ari Dvorin

Produced by: Sam Laliberte

 


Hosted by Brave Software, The Brave Technologist is a series for anyone curious about the real human impact of our digital evolution. It moves past the surface-level buzzwords to ask how technologies like AI, machine learning, and Web3 actually affect everyday decisions, privacy, and the fabric of society. The conversations here are grounded and thoughtful, prioritizing the user's perspective in a landscape often dominated by abstract visions of disruption. You'll hear from builders, thinkers, and technologists who are shaping these tools, discussing not just what is possible, but what is responsible and meaningful. This podcast sits at the intersection of technology, business, and marketing, examining how these forces converge to shape our collective future. It’s for listeners who want to understand the "why" and "for whom" behind the code, offering a nuanced take that is often missing from tech commentary. Tune in for a deeper, more inclusive dialogue about the systems that are redesigning our world.
Author: Language: English Episodes: 50

The Brave Technologist
Podcast Episodes
Why Meta Threw Out the Interviewing Playbook [not-audio_url] [/not-audio_url]

Duration: 24:03
Danit Nativ Navon, Senior Software Engineering Manager at Meta, unpacks how the company became the first major tech firm to build AI-enabled coding interviews. She breaks down the process of rethinking what engineering i…
How a Solana NFT Project Landed on Amazon Prime [not-audio_url] [/not-audio_url]

Duration: 40:55
Andrew Pelekis (CEO of HEEBOO) and Nicholas Cabana (Co-Founder of Claynosaurz and Chief Creative Officer at HEEBOO), break down how a Solana-based digital collectibles brand and a small team of Hollywood animators grew i…
Beyond the Yield: What Staking Is Really About [not-audio_url] [/not-audio_url]

Duration: 23:49
David Kinitsky, Chief Corporate Development Officer at Everstake, shares what distinguishes this cycle of enterprise crypto adoption from every previous one, and why the shift matters. He explains why staking is a founda…
Why Non-Dollar Stablecoins Might Strengthen the Dollar [not-audio_url] [/not-audio_url]

Duration: 33:59
Navin Vethanayagam, Chief Brain of IQ and Co-founder of KRWQ, unpacks his "Stablewoods" theory: why non-dollar stablecoins like KRWQ, the first Korean won stablecoin, may end up reinforcing dollar dominance rather than c…
What Binance.US Is Betting on Next in American Crypto [not-audio_url] [/not-audio_url]

Duration: 20:12
Steve Gregory, CEO of Binance.US, shares what it actually takes to bring a compliant crypto exchange product to market in the US. He explains how his background as a compliance lawyer shapes the way he leads product deci…
The Enterprise AI Mistake Most Companies Will Regret [not-audio_url] [/not-audio_url]

Duration: 25:27
Joel McKelvey, VP of Product Marketing at Glean, shares why the enterprises winning with AI right now aren't the ones chasing the latest model—they're the ones who got their context layer right first. He explains why AI…
Why Most Enterprise AI Projects Never Make It to Scale [not-audio_url] [/not-audio_url]

Duration: 21:54
Natalia Konstantinova, Lead Enterprise Architect in AI at Natwest, shares why most enterprise AI projects don't fail because of the technology, they fail because organizations skip the foundations. She also explains how…
The Global Fight for Free Speech Starts With a Hamster [not-audio_url] [/not-audio_url]

Duration: 41:02
Preston Byrne, Tech & Free Speech Lawyer at Byrne & Storm, P.C., shares how a small American imageboard became the first test case for whether foreign governments can legally force American companies to comply with their…
How to Know When Personalization Becomes Creepy (AI Summit LDN) [not-audio_url] [/not-audio_url]

Duration: 22:00
Dr. Mohsen Ghasempour, Chief AI Officer at Kingfisher, unpacks why the most dangerous thing a company can do with customer data is use all of it. He explains how a global, multi-brand retailer can draw the line between p…
What Counter-Terrorism Taught the NFL About Cyber Risk [not-audio_url] [/not-audio_url]

Duration: 18:36
Kam Karaji, Director of Cyber Security & Risk Management at the NFL, shares what 13 years in counter-terrorism and criminal intelligence taught him about protecting high-stakes environments, and discusses how that thinki…