September 11th-September 17th // Estimated Reading Time: 10 minutes
In This Edition!
AI helps Big Tech keep winning
What’s holding back fully autonomous advertising?
U of Digital Will Be Running An AI Bootcamp At The ANA AI & Technology For Marketers Conference Next Week. Will You Be There? Let’s Meet!
Top Headlines🔥
Big Tech wins as AI tools handle more ad dollars🔒
Source: Digiday
September 15th, 2026
Summary: The IAB now expects US ad spend to grow 12.3% this year, up from its earlier forecast of 9.5% due to strong spending during the Olympics and World Cup. But the growth isn't spread out evenly. AI automated ad campaigns like Google's Performance Max and Meta's Advantage+ now account for 12% of US ad spend, up from 2% in 2023; that share is projected to hit 27% by 2030. This trend will benefit the Big Tech🔒 triopoly the most. Google, Meta, and Amazon already capture 56% of US ad revenue, up from 53% last year. Research firm Madison & Wall expects their share to keep climbing every year through 2030, leaving less for everyone else.
Opinion: Marketers grumble about a lack of transparency and control with these AI tools, but they seem to like the ease-of-use and results enough to keep pouring money into them. This feeds the big tech flywheel: More spend through PMax or Advantage+ trains better algorithms, better algorithms drive better performance, and better performance pulls in even more spend. Marketers, remember, if it’s too good to be true, it probably is. Dig deeper on where your data is going (feeding their AI?), what your ads look like (AI slop?), where your ads are running (garbage inventory?), and actual business impact (questionable).
Meet U of Digital at BRXND NYC (and save 30%)
The U of Digital team will be at BRXND NYC on November 5th for a day focused on AI + marketing. BRXND is giving our readers 30% OFF with code UOFDIGITAL30.
New Products & Features 🚀
What It Does: OpenAI is testing "Sponsored Agents," a feature that lets people start a conversation with a brand-sponsored AI agent after clicking an ad in ChatGPT. Users can ask follow-up questions and get personalized answers before deciding whether to visit the business's site. AI creative tools can draft ad copy and imagery based on an advertiser's landing page. Advertisers can also manage campaigns with natural-language prompts instead of manual setup. And new integrations let marketers manage their ChatGPT ad campaigns through HubSpot and Shopify.
Quick Take: Sponsored Agent turns a single ad click into an open-ended sales conversation, which means brands now need a well-trained, chat-ready agent inside ChatGPT. Cool concept, but brands should be wary of handing over more of their customer interaction and relationship to ChatGPT.
What It Does: Business Agent, the conversational shopping assistant already live in Search, is now testing inside YouTube ads in the US. Viewers can ask complex product questions and get tailored guidance without leaving the video. Brands can also submit richer details in their product catalog so AI Mode can surface richer answers. Google is also making AI performance insights generally available in Merchant Center across the US, Canada, and other countries, letting retailers see how they show up in AI Mode and AI Overviews compared to competitors. And Google is expanding its Universal Commerce Protocol integration hub so merchants can transfer a shopper's cart from AI Mode and other Google surfaces to their own website to finish checkout.
Quick Take: The YouTube bit looks like Google trying to keep shoppers inside the video, rather than sending them to an advertiser’s landing page. Same idea as ChatGPT Sponsored Agents, same risks. Be wary!
What It Does: Meta One raises the usage limits on Meta AI, letting subscribers generate more images and videos and use in-app AI tools like Restyle on Instagram more often, plus professional tools like Meta Business Agent for businesses and creators. Individual plans start at $2.99 a month, while business plans scale up to $499 for teams. Basic Meta AI use stays free; the subscription only covers the extra usage and features on top.
Quick Take: Meta will spend as much as $145B on AI infrastructure this year, and it needs to turn some of that into recurring revenue. It’s an uphill battle for Meta in consumer-facing AI against the incumbents (ChatGPT, Gemini, Claude). Meta Business Agent top-tier plans at $499/month are interesting: Meta is essentially selling enterprise software now.
What It Does: Pinterest built a new AI infrastructure layer with NVIDIA to power visual search, discovery, intelligence, and more. Running on NVIDIA's latest AI chips, the system responds up to 85 times faster and processes results up to 7.3 times faster overall compared to Pinterest's previous setup. The upgrade also lets Pinterest Assistant factor in 25x more image detail per request for more relevant recommendations.
Quick Take: Users won’t see the AI, but they will feel it. Faster processing means Pinterest can push more AI-driven discovery without lag. This should lead to better product-matching in visual search ads and improve user engagement metrics.
What It Does: Gemini 3.8 Live powers more natural, real-time voice conversations with AI, including background task handling like coordinating a booking mid-conversation. It's rolling out in Search Live for everyone now, and in private preview for enterprise customer service, with Salesforce named as an early partner.
Quick Take: If voice agents get materially cheaper, that pulls budget and headcount out of call centers and into a platform. Major shift.
What It Does: Demandbase, the B2B marketing platform, launched Mojo, which turns a strategy brief into a live, cross-channel campaign across platforms like Google Ads, LinkedIn, and Salesforce. It builds audiences, launches campaigns, and flags problems like broken tracking, learning from each run to sharpen the next. Marketers approve the actions that matter and stay in control.
Quick Take: Demandbase's data shows marketers are still mostly asking AI to build lists and pull ROI numbers, not launch campaigns. What changes in marketers' minds enough to let go of the launch button? At this point, we think it’s an accountability problem (who will get fired when things go wrong?), not a technology problem.
What It Does: Salesforce announced new tools and updates this week at Dreamforce. Koa, built with NVIDIA, is a reasoning model trained on 27 years of Salesforce CRM data that handles CRM tasks like routing customer service cases. The new AIforce pipes Salesforce's data and workflows into Claude and Slack instead of requiring users to log into Salesforce directly. It's launching with two integrations: Claudeforce, which brings prebuilt sales skills into Claude, and Slackforce, which lets teams pull live Salesforce data into Slack conversations.
Quick Take: Salesforce routing CRM data into Claude and Slack instead of its UI is Salesforce admitting its interface is no longer the product. For marketers locked into a martech contract: the question at renewal is whether that platform still controls the data layer once the front end moves to chat.
And hey Marc, all the branding is getting to be a bit much…
AI Use Case of The Week💡
Cosabella used AI to autonomously run its paid media
The Setup: Cosabella, an Italian intimates and swimwear brand, had been working with agencies to run its digital marketing but wasn't seeing the results it wanted. With a new corporate strategy ready to launch, the team was looking for different ways to execute rather than another agency relationship.
The AI Solution: Cosabella brought in Albert, an AI marketing platform, and fed it high-level KPIs, campaign parameters, and creative assets. Albert identified user behavior patterns across paid search and social, then acted. It built multi-variate micro-campaigns and bought media autonomously using predictive analytics.
The Results:
• 336% return on ad spend
• 155% increase in revenue directly attributed to Albert
• 37% increase in website sessions
• 20x increase in social purchases
Why This Matters: Albert made the calls and spent the budget itself, rather than flagging opportunities for a human to execute. That's a meaningfully different value prop than most "AI insights" tools, and a major leap of faith.
Your Action: Before handing budget decisions to an autonomous tool, test on a small scale. Take an underperforming campaign and set clear guardrails such as spend caps, pre-approved creative only, and a minimum ROAS threshold that triggers an automatic pause. Let an AI tool like Albert make and execute optimization decisions without sign-off. Check in daily for red flags the first week, then run the test for four weeks, since it often takes that long for these systems to identify what’s working and scale in response. Compare performance against a campaign your team is running manually.
Other Notable Headlines📌
Marketers are using AI search fears to win bigger budgets🔒 - As consumers move toward AI assistants for shopping research, brands like State Farm and Mercedes-Benz are responding by increasing upper-funnel brand spend.
Google rolls out a pay-per-value AI licensing program to publishers🔒 - The pilot pays publishers when their content "significantly" contributes to an AI-generated response in Gemini or AI Overviews, but sources say the payouts are "peanuts" and the calculation is a "black box."
Microsoft Advertising sets rules for AI-generated ads - Advertisers must disclose when AI helped create or alter an ad and preserve watermarks and metadata showing AI involvement.
AEO startup Profound hits unicorn valuation, raises $180M Series D - The AI search optimization startup now has a $1.8B valuation seven months after its last raise. Revenue is up 3x in six months, and it now has 1K+ enterprise customers, including Comcast and Walmart.
Mediaocean's venture arm offers AI startups access to its $200B ad spend system - The new investment unit gives young AI-in-advertising startups capital plus direct integration into Mediaocean's workflow software, which handles over $200B in annualized ad spend.
Goodwipes uses AI to spoof OpenAI's Astra ad, with a butt-wiping twist🔒 - The flushable wipes brand and AI-native agency OK Future turned around the parody ad in just four days using AI tools. It spoofs OpenAI's divisive ad for its new Astra model, swapping in "Asstra". Nice one!
Former Anthropic researcher warns AI could destroy humanity by end of decade - Jacob Coxon told the Wall Street Journal that colleagues are using terms like "endgame" as the industry races toward self-improving AI, and that competition between major labs is making rapid AI development less safe. Major AI labs are now working together on AI safety and are seeking support from US lawmakers.
That’s It For This Week 👋
The U of Digital AI Edge Newsletter is intended for AI Edge Pro subscribers, but occasional forwarding is okay!
To subscribe visit Uof.Digital/Newsletters/ai-edge or AI Edge Pro, or contact us directly for group subscriptions.
And remember, U of Digital helps teams drive better outcomes through structured education on critical topics in marketing / advertising like AI, mar tech, ad tech, CTV, commerce media, and more. Interested in learning more about how we can supercharge your team? Let’s talk!
Thanks for reading!
