AI Agents. Buy & Sell Side. What's Next for Humans? Andrew Mole & Andrew Gilbert (pubX) - Media Tech Talk #39
Media Tech TalkShow Notes
For years, automation in advertising meant software following rules: bidding, buying, optimising. What is emerging now is something different. AI agents that reason, negotiate and strike deals on your behalf, on both sides of the trade.
This episode brings together the two people building that for publishers and advertisers at the same time. Andrew Mole is co-founder and CEO of pubX, his second ad tech company, built on five years and more than a trillion auctions of pricing data. Andrew Gilbert has just stepped in to lead pubX across Australia and New Zealand, after senior roles at Yahoo, Afterpay and IAS and a spell running his own AI advisory.
Across part one, Daevid Richards asks what advertising actually becomes when the buying and the selling are handled by machines talking to each other. Both Andrews trace how they got here, why an industry this inefficient was always ripe for disruption, and what a sales agent and a buying agent really do once you take the bid request out of the middle. Andrew Mole makes a call on whether agentic supplements programmatic or eventually replaces it, and Andrew Gilbert explains where the genuine ANZ appetite sits, why every publisher conversation starts from suspicion, and how AI assisted, autopilot and truly agentic are three very different things.
If you work in programmatic, publisher monetisation, agency trading or brand safety, this is the clearest look yet at where the supply chain is heading, and what happens to the people who have always managed those relationships.
Part two goes deeper into what really happens when two agents negotiate a deal, the honest question of what could go wrong when machines trade at this speed, and Andrew Mole's call on what it all means once this becomes the norm.
CHAPTERS
0:00 - The shift no one has explored on this show yet 1:17 - Welcome, and how many Andrews are at pubX 1:47 - Cotswolds shorts and the Central Coast commute 2:44 - Andrew Mole: Platform360, PubGalaxy and founding pubX 5:02 - Andrew Gilbert: Yahoo, Afterpay and going deep on AI 7:14 - Why agentic and AdCP moved this fast 8:51 - The swing back towards the publisher 10:47 - What is pubX, and the problem it is solving 14:08 - Signals agents, governance agents and the new stack 15:21 - Does this supplement programmatic, or end it? 17:12 - Who actually captures the media dollar 18:10 - What a sales agent does on a publisher's behalf 20:54 - Why pubX's pricing heritage matters here 22:38 - The ANZ reaction: appetite, and suspicion 24:53 - AI assisted, autopilot, or truly agentic 26:04 - brands.json and a new level of advertiser control 26:47 - How the buying agent works for an agency 28:15 - Collapsing the pre DSP and the activation layer 29:09 - Wrapping up, and what is coming in part two
CREDITS
💬 Guest: Andrew Mole, Co-Founder and CEO - pubX 🔗 https://www.linkedin.com/in/andrew-j-mole/
💬 Guest: Andrew Gilbert, Country Manager ANZ - pubX 🔗 https://www.linkedin.com/in/andrewgilbert01/
🔗 pubX: https://pubx.ai/ 🔗 https://www.linkedin.com/company/pubxai/
🎧 Host: Daevid Richards 🔗 https://linkedin.com/in/daevidrichards/
🏷️ #MediaTechTalk #pubX #AgenticAdvertising #AdCP #AdTech #Programmatic #AIAgents #PublisherMonetisation #DigitalAdvertising #Podcast
More from Media Tech Talk
There's a moment on the biggest screen in the home, just before you choose what to watch, that every advertiser now wants a piece of. In part two of his conversation with Daevid Richards, Alex Blundell Jones, Commercial Director, Australia at LG Ad Solutions, looks at the challenge he believes is reshaping the future of television, and what the TV can really offer advertisers in the AI era. It starts on the home screen. Alex explains how LG builds ads into a premium device without annoying the people who paid for it, why some categories have been quicker to get on board than others, and how new creative tools are turning a website link or an out of spec asset into something ready for the big screen. Countdown timers, QR codes and 3D creative all get a mention, along with a few campaigns that literally broke out of the frame. Then to performance. Television was built to sell brands, so what does an outcome look like for a TV manufacturer? Alex unpacks how LG Ads proves incrementally, how a home screen ad can be connected to what a viewer actually does days later, and what it takes to follow a conversion from the TV to the web. Daevid also tackles fragmentation: with video budgets spread across BVOD, streamers, FAST and the home screen, where is unduplicated reach hardest to solve, and how can device data help budgets go further? Finally, AI and sport. Alex tells the story of an advertiser who arrived late to the World Cup with every broadcast package already gone, and the engagement result that followed. He closes with his crystal ball: mergers and acquisitions, an in-home experience that may be hard to fathom, and where operating systems could turn up next.
Streaming has overtaken cable and broadcast in the United States, and the television itself has become something new: an operating system, a home screen, a bundle of free channels and an ad platform in its own right. So can a television manufacturer really compete with the broadcasters for video budgets? Daevid Richards puts that question to Alex Blundell Jones, Commercial Director, Australia at LG Ad Solutions, the person sent from London to open the Australian market from a standing start.
Knowing an ad was served is one thing. Knowing it was looked at is another. But was it the right ad, and did it actually land with the person on the other side of the screen? In Part 2, Rob Hall, founder and CEO of Parallel, tells Daevid why he came back from a year off to start all over again, and reveals the thread that runs from the attention business he sold to the one he is building now.
Media Tech Talk #041 - Rob Hall (Parallel) Part 1
Part one introduced the agents. Part two asks the harder questions. When software is negotiating with software at machine speed, who is actually in control, and what happens when something goes wrong?