The AI Edge Podcast

Amazon rebranded its entire ad platform — there's no more DSP, the default is now a PMax replica, and they're calling it Amazon Ads Agent. Meta's Muse started crawling Amazon and got shut down within hours. OpenAI launched Dots as a small business tool overnight — apparently because Zuckerberg can't beat Sam Altman, not even on a Tuesday. And President Trump signed an executive order renaming AI to SI. Superintelligence. Because it's super.

This week, Shiv and Myles sit down with Mike Finnerty, US President of Mutinex — the agentic MMM platform that has spent the last four years doing what the measurement industry has been promising to do for 80 years, but never quite pulled off operationally.

Mike breaks down why every walled garden platform grading its own homework creates a math problem that doesn't add up — when you sum the self-reported performance of five or six platforms, the total is always higher than actual sales. It was the ad network problem of 2008 all over again. Mutinex's answer: inject a neutral third-party incrementality signal back into the bidding algorithms so platforms optimize for actual growth instead of their own metrics. And every single time they've done it, platform-reported performance has improved.

He also introduces one of the cleanest frameworks in the episode — read vs. write for LLMs. Until recently, everything AI did was read: research, information gathering, journey changes. What's happening now with Muse, Dots, and Amazon's branded conversations is write: actually taking actions on behalf of consumers. Someone's Muse agent negotiated a Facebook Marketplace deal without the other person knowing. The write era has started.

They also get into why the conventional wisdom that brand is a long-term investment is about to be flipped on its head — if your consumer doesn't say "I want a Stanley water bottle" before the agent fires up, the agent will just pick whatever's cheapest and available — why this holiday season will be the first where agentic transactions actually show up in commerce data, and why the entire marketing ecosystem should stop counting clicks and start asking one question: how much went in, how much came out?

Shiv and Myles will be at Ad Week — send a DM if you want to connect.

Ready to level up your AI skills? Join the U of Digital AI Accelerator.

Learn more about the AI Literacy Alliance at uof.digital/alliance

New episodes every two weeks. Get your AI Edge here.

Amazon just partnered with ChatGPT to let advertisers buy AI inventory through Amazon's DSP — and ChatGPT still won't let Amazon crawl its product catalog. Meta charged advertisers a software fee for the first time in company history. Big tech captured 56% of all ad spend this year, up from 53% last year, and AI is the tailwind driving every bit of that growth. And OpenAI's IPO apparently isn't happening this year — which tells you everything you need to know about where they actually are.

This week, Shiv and Myles welcome Paul Boruta, Founder and CEO of Slingwave — an independent AI-native marketing decision engine built for eCommerce and DTC brands. Paul started at Overture and Yahoo, built and sold a startup in the YouTube space, and has spent the last several years building what he describes as a closed loop system that puts all of your marketing data — paid, organic, competitive, and contextual — into one model that gets smarter with every campaign.

Paul breaks down why the Amazon/ChatGPT partnership makes perfect strategic sense once you map the funnel correctly: Amazon and Google are the cash register, ChatGPT is excellent at the mid-funnel research and discovery phase where neither of them is particularly strong. He also makes a notable admission: 18 months ago he would have told you synthetic data for incrementality testing was too flawed to be useful. That's changed — and changed fast. And he explains why doing it wrong — one person, one deck, no validation, no closed loop — is worse than not doing it at all.

They also get into why real-time competitive signals changed the measurement game entirely — because advertising isn't turn-based, you and your competitors are all in market simultaneously — why sponsored AI agents live or die on the neutrality question, and why the old world of deterministic multi-touch attribution is melting fast and everyone needs to be retooling for what comes next.

🎓 The U of Digital AI Accelerator starts Tuesday, September 22nd — a few spots still available.

U of Digital is proud to partner with the ANA to bring AI education to over 20,000 brands and 50,000 marketers.

New episodes every two weeks. Get your AI Edge here.

Salesforce just gave the "Force" suffix to another brand for the first time in its history — and the stock jumped 25%. ChatGPT ads are getting roasted on LinkedIn for underperforming. The Trade Desk launched Kokai Zuma, their most AI-forward interface yet. And Google quietly embedded its most powerful image generation model directly into Workspace. A busy couple of weeks.

This week, Shiv sits down solo with Marco Matos — CEO and co-founder of Adora, and a 20-year veteran of building ad tech products inside the walled gardens — for a conversation that moves fast, disagrees productively, and lands on some of the sharpest takes of the season.

Marco breaks down why the Salesforce/Anthropic partnership is less about Claude and more about a deliberate business model pivot — from seat-based pricing to token-based consumption — and why Benioff pulling that off at Salesforce's scale is one of the hardest things any tech CEO has done in years. He also makes the case that the "ChatGPT ads are underperforming" narrative is based on statistically meaningless data and a measurement infrastructure that doesn't exist yet — and that last-click attribution being anyone's source of truth in 2026 should be a fireable offense.

They also get into why the agentic buying conversation is missing the point — the real unlock isn't using a platform's agentic workflow, it's having your own agent talk to theirs — why the ad tech industry keeps forgetting about creative while optimizing everything around it, and why every marketer building a Frankenstack of point solutions is setting themselves up to lose to whoever builds the end-to-end system first.

📣 A few announcements :

U of Digital is partnering with the ANA to bring AI education to over 20,000 brands and 50,000 marketers. More details coming soon at uof.digital and on social.

The AI Accelerator starts September 22nd — almost sold out. Get your ticket now!

And U of Digital will be at Ad Week — throwing a House Party with Unplugged Collective for marketers in media, ad tech, and MarTech on October 7th. Send a DM if you want an invite or register here: https://luma.com/thehouseparty-awny-2026

New episodes every two weeks. Get your AI Edge here.

Pepsi is shopping for an AI transformation partner and the whole industry is watching — consultancies or agencies, who wins? Alexa for Shopping now gives you one or two recommendations instead of a page of results, and brands are quietly panicking about how to become the one that gets surfaced. Google bought Spirit Airlines' entire data archive for $10 million. And SpaceX AI paid $60 billion for Cursor, the agentic coding tool, in an acquisition that slipped under the radar until it didn't.

This week, Shiv and Myles welcome Stephanie Layser — Global Sub-Industry Leader for Publishers and Advertising at AWS — for a conversation that connects the infrastructure story to the advertising story in ways most people aren't putting together yet. Stephanie has spent her career at the intersection of data, identity, and ad tech — at News Corp, AppNexus, and now AWS — and she brings a grounded, customer-facing perspective on what AI transformation actually looks like when the work is real and the stakes are high.

Stephanie breaks down what Pepsi's five-year data and cloud investment actually means in plain English — and why the companies that did that work early are the ones who can actually activate on AI right now. She explains why programmatic supply has historically been dumb pipes, and how the shift to agentic ad operations — including the work AWS is doing with Warner Brothers Discovery — is building the intelligence layer that changes what those pipes can actually do. She also shares what her own workday looks like now versus before AI: who she left on read, what needs her attention, and how writing a strategy document that would have taken weeks now takes a fraction of the time.

They also get into why a customer ran the same technology build evaluation two years apart and came back with a timeline ten percent of the original, why Alexa for Shopping going from infinite shelf to two recommendations is a bigger GEO story than most people realize, and why the AI advertising infrastructure being built right now feels bigger — not just similar — to the birth of programmatic.

Ready to level up your AI skills? Join the U of Digital AI Accelerator. Enrollment now open for September | Sponsored by Blis, a DSP by T-Mobile Ad Solutions

https://uof.digital/ai/accelerator/

New episodes every two weeks. Get your AI Edge here.

Time Magazine started selling sponsored FAQ sections to AI crawlers — and Perplexity shut it down, threatened to lower their quality score, and called it a trust issue. Publishers are wondering whether to block AI crawlers entirely. OpenAI quietly filed for its IPO while churning through senior executives. Anthropic is proactively watermarking Claude outputs ahead of the EU AI Act. And Mark Zuckerberg published a sprawling manifesto about open source AI that Shiv summarized with AI because nobody has time for that.

This week, Shiv and Myles sit down with Brian Wieser — CEO of Madison and Wall, a former equity research analyst, agency veteran, and the person Frienemies called "the industry's most widely respected marketing analyst." Brian brings a level of analytical rigor to the AI and advertising conversation that cuts through the hype in ways most guests simply can't — and this episode is full of moments where he says the quiet part loud.

Brian breaks down why AI will neither grow nor shrink the advertising pie — and why that's actually the most honest and interesting answer you can give. He explains the three real mechanisms through which AI could shift ad spending: changes in competitive intensity, creative destruction, and broader economic tailwinds. He also takes direct aim at Sam Altman's claim that AI can eliminate 95% of what agencies do — and dismantles it with data. Agency revenues are not decelerating. The budget is the budget. And one of his clients at Madison and Wall refused to buy a banner ad because the sponsored post lives in the LLM forever. Which, as Brian puts it, is fascinating.

They also get into why media planning line items are quietly shifting from TV and audio to Google, Meta, and Amazon — and what that means for everyone who isn't one of those three — why agencies arbitraging AI tokens is really just the Wall Street research bundling model with a new coat of paint, and why Meta doesn't actually care about advertising even though it funds everything they do.

Ready to level up your AI skills? Join the U of Digital AI Accelerator.

New episodes every two weeks. Get your AI Edge here.

A new Chinese AI model called Kimi K3 just matched frontier model performance at half the cost — and nobody saw it coming. Again. Google launched three new Gemini models in one week, each optimized for different tasks. OpenAI is reportedly building a screenless device with sensors, a camera, and a microphone. And Meta's AI creative tools are generating bikes with two sets of handlebars and telling advertisers it's their problem.

This week, Shiv and Myles sit down with Jamie Domenici, CMO of Klaviyo — the B2C CRM powering over 200,000 brands and eight billion messages a day. Jamie brings a rare perspective to the show: she's a marketer leading marketing for a marketing platform, which means she's living the AI transformation from both sides of the table simultaneously.

Jamie breaks down why the autonomous marketing future is closer than most people think — and why human judgment is still the thing that separates the brands that win from the ones that just move fast. She explains why you can't just turn on an agent and let it rip, why brand guidelines need to be trained into your AI from day one, and why the companies embracing AI right now are already building more, personalizing better, and outrunning the competition. She also shares how Klaviyo used AI to automate their entire product release workflow — from engineering ship to blog post — because the pace of AI-driven releases made the old way completely unsustainable.

They also get into why ChatGPT's contextual targeting is fundamentally different from Meta's blunt retargeting, why voice AI hasn't disrupted marketing yet but is about to, and why every marketer needs to start thinking about how their brand shows up in places they've never had to optimize for before — including image search, LLMs, and whatever screenless device OpenAI ships next.

Ready to level up your AI skills? Join the U of Digital AI Accelerator.

New episodes every two weeks. Get your AI Edge here.

AI-generated content is back on the rise in ad inventory — and it's not just MFA farms anymore. Reddit deleted 25,000 spammy AI posts in Q1. Google took down 50,000 AI slop networks. And the ANA's latest brand safety report shows the problem is getting worse after years of improvement. Meanwhile, Mark Zuckerberg logged onto X for the first time in years to announce a frontier AI model, an image generator, a video generator, and a cloud compute business — all in one week.

This week, Shiv and Myles sit down with Sam Bloom — founder of Camelot Strategic Marketing and Media, now Head of Partnerships at PMG — for a conversation that spans brand safety, Meta's AI ambitions, the future of the agency business, and why the most interesting real estate in AI right now might be something nobody is talking about.

Sam breaks down why AI slop isn't a new problem — it's a new weapon in a very old arms race — and why the best defense is still doing basic things well, going deep with trusted partners, and de-averaging your data to find where the fraud is actually hiding. He also makes the sharpest observation in the episode: everyone knows who owns the search box. Nobody owns the voice box. And whoever gets there first changes the whole game.

They also get into why Meta embedding AI directly into the advertiser console — and potentially not charging for tokens — could be the most disruptive business model move in the industry right now, why fully agentic end-to-end campaigns are more Tony Stark than science fiction, and why the agency business is quietly moving upstream from media reselling to business transformation.

🎙 Next U of Digital AI Edge Live Session — July 22nd at 12PM PT

ChatGPT now influences a billion people and counting. But how do you influence ChatGPT? Join us for a live look at both GEO and ChatGPT Ads — including an exclusive demo of the Evertune platform — featuring Evertune CEO Brian Stempeck, who spent 14 years at The Trade Desk before launching Evertune to tackle the next frontier of digital advertising.

This one is FREE for everyone.

Amazon just launched Alexa Plus Agentic Ads — native checkout through voice, shoppable ads across third-party publishers, and a monetization machine that just keeps accelerating. WPP and Horizon are building agent-to-agent infrastructure. Databricks announced an agentic CDP. And Hightouch made a public offer for LiveRamp's identity spine while simultaneously raising at a $2.75 billion valuation.

This week, Shiv and Myles sit down with Kashish Gupta, co-founder and co-CEO of Hightouch, to make sense of all of it. Hightouch works with some of the largest B2C brands in the world — Domino's, Spotify, and others — helping them do their marketing end-to-end in an AI native way. And Kashish has a clearer view than almost anyone of where the technology actually is versus where the hype says it is.

Kashish breaks down the number one complaint he hears from CMOs right now — AI still can't reliably stay on brand, no matter which foundation model you use — and why getting from 80% to 99.9% accuracy is the hard engineering problem nobody is talking about. He also explains why AI probably shouldn't own media buying yet — you can't fire your AI, and having a throat to choke for the metric still matters — and why the right model is a proactive insights engine with human discretion at the center, not autonomous agents running everything unchecked.

They also get into why agentic voice advertising isn't ready yet — the AI still can't think fast enough while you're talking — why Databricks building audiencing capabilities is actually good news for Hightouch, and why Hightouch's offer for LiveRamp's identity spine is less about competition and more about keeping identity neutral and accessible for the brands that need it most.

Learn more about the AI Literacy Alliance at uof.digital/alliance

New episodes every two weeks. Get your AI Edge here.

Apple announced its long-awaited Siri overhaul — and the reviews are basically "welcome to 2023." It's built on Gemini under the hood, while Apple simultaneously runs a privacy campaign aimed squarely at Google. Meanwhile, SpaceX went public at a two-trillion-dollar valuation, Anthropic and OpenAI both filed for IPOs with wildly different burn rates, and Anthropic dropped a new model called Mythos that's apparently too dangerous for the public to touch.

This week, Shiv flies solo with Chris Kelly — founder and CEO of Upwave, his neighbor and friend — for a conversation that ranges from Apple's AI catch-up to the economics of the AI arms race to what real, concrete AI transformation looks like inside a measurement company.

Chris breaks down why nobody should hold a year-old AI opinion against a current AI model — a year is "light years" in this space — and why Apple's massive iPhone distribution and head start on voice could let it make up ground faster than its brand reputation suggests. He also makes the case that AI conversations reveal far richer upper-funnel context than search ever could, opening up entirely new categories of ad units that don't exist today. And he pushes back on one of the most common AI marketing claims out there: that AI collapses the consumer journey. Some purchases still take three years, AI or not.

They also get into why Anthropic going public might eventually mean an ads business — Super Bowl irony and all — why Google's AI strategy is really about embedding AI into every surface it owns rather than building a destination app, and why predictive brand lift could triple how much of the industry's advertising actually gets measured.

Learn more about the AI Literacy Alliance at uof.digital/alliance

New episodes every two weeks. Get your AI Edge here.

Google just announced the biggest overhaul to search in 25 years. The blue links aren't going away — they're just being buried. AI mode is now the default. Agents can be dispatched from the search bar. Native checkout is live. And five new ad formats launched alongside all of it. Meanwhile, OpenAI is quietly moving toward a CPA pricing model for ChatGPT ads, Anthropic just hit $47 billion in ARR and may turn a profit this quarter, and Meta wants you to pay a subscription fee to escape the attention economy it built. Make that make sense.

This week, Shiv and Myles sit down with Dr. Cecilia Dones, founder of Three Standard Deviations and professor of AI and strategic communications at Columbia University and NYU. Dr. Dones helps organizations move from AI adoption to AI advantage — and she brings an academic rigor to the conversation that reframes some of the biggest stories in the industry in ways most people haven't considered.

Dr. Dones breaks down why Google's AI overhaul isn't just an ad product story — it's a governance story. As more brand strategy, planning, and creative decisions migrate inside platform walls, the algorithms will eventually know a brand's identity better than the brand does. She also explains why ChatGPT's pixel is a fundamentally different data asset than anything Google has ever had — because it doesn't capture what people searched for, it captures how people actually think. And she delivers the most actionable advice in the episode in one sentence: audit your brand in AI today, because it looks nothing like your SEO did.

They also get into why Meta charging a subscription fee to avoid ads is an oxymoron and a trust problem wrapped in one, why the measurement industry is quietly shifting back toward econometric modeling as platform attribution gets more opaque, and why "set it and forget it" always creates a PR disaster eventually.

Learn more about the AI Literacy Alliance at uof.digital/alliance

New episodes every two weeks. Get your AI Edge here.