The Prompt & The Pitch

Sean and Cameron sit down with Harry Tong from PubMatic, to explore how agentic AI is changing programmatic advertising—from campaign execution and audience discovery to the guardrails needed before advertisers hand over control.

Cameron shares his experience setting up a campaign through PubMatic’s MCP connection, while Harry explains why automating the clicks doesn’t remove the need for expertise. Understanding how the system works becomes essential to defining what agents can—and cannot—do.

They discuss:

• Why commercial roles are becoming more technical—and how AI can help people make that transition
• PubMatic’s internal AI culture, hackathons and everyday workflows
• Harry’s framework for deciding when to delegate, collaborate or consult AI
• The risk of producing impressive work without retaining the underlying knowledge
• How PubMatic’s agentic capabilities connect buying, audiences, inventory and curation
• Financial guardrails, approval workflows and business rules for autonomous campaigns
• Why independent agencies have been able to experiment quickly
• A future where buyers choose their own interface and switch between platforms more easily
• Why proprietary intelligence and strategic judgement become more valuable as execution gets easier
• Practical ways to start experimenting with MCP, Claude and Notion

The value of programmatic expertise isn’t disappearing. It’s shifting from operating the platform to designing the system, setting its boundaries and deciding what it should achieve.

What happens when clicking an advert starts a conversation instead of opening a landing page? And when your AI assistant buys something for you, whose interests is it really serving?

Sean and Cameron return with an AI news roundup exploring sponsored agents, personal assistants and what happens when advertising becomes part of the conversation.

They unpack the sponsored-agent experience discussed for ChatGPT, why Meta’s distribution could give it an advantage in personal AI, and the trust problem facing any company that wants to both recommend products and sell advertising.

They also discuss:

• AI safety concerns, IPO speculation and the economics behind frontier models

• Why brand recognition still matters when agents influence purchasing decisions

• How Shopify and HubSpot integrations could help bring AI advertising to smaller businesses

• Cameron’s experience testing ChatGPT ads—and whether early results will survive greater competition

• WhatsApp-based assistants and the opportunity to build a Western super-app

• Autonomous agents that complete work and come back for review

• Why cheaper, specialised decision models could unlock new programmatic advertising applications

• Using AI to filter thousands of pages, analyse site lists and identify relevant signals

• Low-cost video generation, interactive livestreams and personalised entertainment

The big question: if the same company supplies your assistant and sells access to your attention, how do you know who that assistant is really working for?

Sean and Cameron sit down with Dr Cecilia Dones - lecturer, researcher and AI transformation adviser - to explore the human skills that become more valuable as machines take over execution.

Cecilia argues that tomorrow’s professionals must learn to operate as creative directors, orchestrators and curators: setting the vision, coordinating different technologies and applying the taste and judgement machines cannot provide.

They discuss:

• The three skills students and early-career professionals need in the AI era

• Why portfolios should demonstrate judgement rather than button-pressing

• How CMOs can cut through the pressure to “AI everything”

• Why leaders cannot understand AI theoretically—they must experience it

• Which customer experiences should be automated and which should remain human

• How AI is lowering the barrier to experimentation for smaller businesses

• The security risks created by agent-to-agent communication

• How agencies, brands and technology companies offer different career paths

• Why consumer resistance could make genuine human interaction a premium

• The difference between accumulating contacts and building real connections

• How speaking rudely to machines could change how we communicate with people

• The risks of emotional dependence and parasocial relationships with AI

AI may change how work is delivered, but it does not change why the work matters.

Cecilia’s advice is to begin with the people you want to help - then decide which technologies can help you do it.

Sean and Cameron sit down with Brandon Keenan, co-founder and CEO of ViVV Labs, to explore building the AI agency and why he believes the traditional holding-company model cannot survive the agentic era.

They discuss:

• Why attribution and marketing mix modelling still fail to answer the CMO’s most important questions

• How ViVV Labs aims to provide “WPP for $75 a month”

• Why AI could enable ten-person companies worth tens of billions

• Whether the era of the marketing specialist is ending

• Why tomorrow’s marketers will need to become AI-enabled generalists

• The challenge of giving junior talent the experience needed to develop judgement

• Why people forgive human mistakes but quickly lose trust in algorithms

• How companies can turn their expertise, playbooks and historical knowledge into an AI moat

• Why the future of marketing technology could be multimodel, headless and interface-free

• The critical-thinking skills people risk losing by outsourcing too much thought to LLMs

Workflows and code can increasingly be replicated. Brandon’s argument is that accumulated intelligence knowing what works, what fails and why is what becomes genuinely defensible.

Sean and Cameron discuss Anthropic’s move to watermark AI-generated text and whether audiences will reject content once they know a model helped create it. The conclusion: context matters. AI assistance is acceptable for sharpening an argument, but it feels very different when used for something requiring genuine human care.

They also explore:

• Why clear thinking will become more valuable as AI produces more content

• Whether businesses are moving away from the newest, most expensive models

• Why model-routing platforms such as OpenRouter could become essential infrastructure

• The rapid growth of Chinese models among American businesses

• Meta’s attempt to bring vibe-coded games into social feeds

• Why distribution—not development—is becoming the barrier to building successful apps

• Whether developers must evolve into product thinkers

• How Lovable could become the Shopify of software development

• Why Omnicom outsourcing parts of its AI platform raises questions about its technology moat

• What Google could gain from buying an airline’s data out of bankruptcy

Vibe coding may still sound like a shortcut to some people. But as AI takes over more technical execution, taste, judgement, distribution and understanding what people actually want become the valuable skills.

Sean and Cameron sit down with Zaeem Bhanji, co-founder and CEO of Yolodex and former AI research and ads product lead at Meta, to explore what this means for media buyers, creative teams and agencies.

They discuss:

- Why the skills of traditional Meta media buyers were built for a world that no longer exists

- How psychographic personas can replace demographic targeting

- Why marketers are automating creative production when they should be automating customer research

- How Yolodex turns customer data and advertising performance into creative opportunities

- Why AI-generated content can damage trust and brand authenticity

- Whether Meta will eventually create, target and optimise entire campaigns

- Why 90% of advertising may be automated while the exceptional 10% becomes more valuable

- Whether creative strategists will replace media buyers as advertising’s most sought-after talent

- What AI means for agencies, employment, energy costs and the wider economy

The ability to generate thousands of adverts is no longer an advantage. The real advantage is understanding which customers matter, what they want and which creative will reach them.

AI-generated video is moving from novelty to genuine creative tool. But is it ready to transform filmmaking and advertising?

Sean and Cameron put Higgsfield’s latest capabilities to the test, creating cinematic scenes, personalised videos and complete product adverts from simple prompts.

They explore:

• Why Pixar co-founder Ed Catmull’s involvement gives AI filmmaking new credibility

• Whether storytelling becomes more valuable as actors become replicable

• Sean’s prediction that Paramount could acquire Higgsfield

• How convincingly AI can recreate Cameron inside different scenes

• Why smaller brands can now produce advertising previously reserved for major budgets

• How agencies could use AI to develop, storyboard and validate concepts

• Why imperfections still prevent many AI adverts from being production-ready

• Whether smart brands will respond to infinite content by deliberately creating less

  
  

When everyone can produce more creative for less money, volume stops being the advantage. Taste, storytelling and knowing what deserves to be made become the real differentiators.

Nielsen has agreed to acquire DoubleVerify in a $2.15 billion all-cash deal - but does combining two measurement businesses create something genuinely more valuable?

Richard Cavill joins Sean and Cameron to give their immediate reactions. Rich sees a natural fit between Nielsen’s strength in TV measurement and DoubleVerify’s verification capabilities, while Cameron questions whether the commercial upside can satisfy the demands of private-equity ownership.

They also discuss:

• Disney allowing TikTok creators to use its characters and the risk of losing control of its IP

• How generative video could produce an entirely new type of filmmaker

• Why HBO Max, Netflix and Disney are becoming obsessed with short-form video

• PubMatic’s decision to retire OpenWrap for the web while retaining its mobile SDK

• Whether declining web traffic will accelerate advertising’s shift towards apps

• A future in which AI replaces websites and apps as the primary consumer interface

• What smart glasses, AR and VR could mean for media buying

• Why audio, video and out-of-home may survive the decline of traditional display

  
  

The interfaces may change completely, but attention will still need to be earned and monetised.

Sean and Cameron explore Amazon’s unique advantage: it owns the audience’s attention, purchase intent, transaction data and much of the inventory - while its wider commerce and logistics infrastructure closes the loop.

They also discuss:

• Why gaming attracts just 2.3% of US digital ad spend despite having three billion players

• Whether Xbox advertising could help bring the price of games back down

• How brands can enter gaming without disrupting the experience

• Microsoft’s overlooked advertising assets across Xbox, LinkedIn and Xandr

• Why Adidas’s record sales matter more than its short-term share-price fall

• What Nike’s performance-marketing obsession teaches us about neglecting brand

• Why Amazon’s technology and user experience remain its biggest weaknesses

  
  

Google and Meta sell attention. Amazon can see when someone is ready to buy—and prove what happened afterwards.

01:50 — Can Xbox finally close gaming’s advertising gap?
07:58 — What good advertising inside games actually looks like
11:39 — Microsoft’s overlooked advertising empire
12:46 — Adidas, record sales and the case for brand investment
18:58 — Amazon’s $76 billion advertising machine
23:57 — Does Amazon have the best proposition in advertising?

AI in programmatic advertising didn’t begin with ChatGPT. Machine learning has powered bidding engines for years, but LLMs are now transforming how the people around those systems work.

Sean Halpin and Cameron Armstrong sit down with Puneet Gupta, co-founder and Chief Product Officer at Kayzen, to explore the two sides of advertising’s AI evolution: the specialist models making real-time bidding decisions and the agentic tools reshaping workflows, product development and campaign management.

They discuss:

• How Kayzen uses internal MCPs to make complex data accessible without SQL
• What the programmatic trader’s workflow could look like in a few years
• Why LLMs won’t replace specialist bidding engines anytime soon
• How AI is accelerating product development and model experimentation
• Why AI proficiency must be demonstrated by leadership—not delegated to an “AI lead”
• How Kayzen’s traders built their own campaign analysis agent
• Why transparency, proprietary data and performance matter more than simply adding an AI interface
• The litmus test every AI founder should apply to their business

Puneet’s test is simple: if the next frontier model becomes three times better and half the price, does that strengthen your business or make it obsolete?