Benji Shomair, VP of Monetization, on LinkedIn yesterday:
“From reporting on campaigns to making decisions to taking action. In our webinar, ChatGPT Work for Marketing Teams, Chad Nelson, Sahil Punamia and I share how OpenAI’s marketing team uses ChatGPT Work and ChatGPT Ads in practice…”
See: “How Our Marketing Team Uses ChatGPT Work” – OpenAI (signup required)
From the webinar landing page:
“The market moves in real time. Marketing teams should be able to do the same. But acting quickly is hard when customer insights, creative, and campaign context are scattered across tools and teams. ChatGPT Work brings that information together…”
From tipsheet: Is this the beginning of OpenAI absorbing functions now spread across martech and ad tech tools?
Related: “Inside OpenAI: How our data team uses AI to move faster” (Sept. 8) – Vijaye Raji, CTO of Applications on LinkedIn
AGENTS
Buyer, seller agents go to work in CTV
Magnite yesterday said Amnet France used its Buyer Agent to build and activate a CTV campaign for an automotive advertiser, its first agentic campaign in EMEA.
Amnet described the campaign in natural language, with Magnite’s buyer and seller agents working through Magnite Orchestration to find and activate CTV inventory. Magnite said setup time fell about 70%.
Magnite VP of Agentic Jonathan Moffie on LinkedIn:
“Mid-campaign, the agent flagged an under-delivering behavioral audience and recommended shifting budget toward a better-scaling contextual one. The team acted on it to drive better outcomes.
An agent is only as useful as the supply it can reach and the data it can access. By plugging into Magnite Orchestration, our Buyer Agent connects directly into supply-side intelligence at scale. So does any third-party buyer agent.”
Read more. (Sept. 8)
Read: “Magnite Launches First Agentic Campaign in EMEA With Amnet France” (Sept. 8) – Magnite
From tipsheet: The important next step is Magnite’s claim that the same infrastructure is open to third-party buyer agents, potentially allowing an advertiser or agency’s agent to transact directly with Magnite’s supply-side intelligence.
SELL-SIDE
PubRunner brings AI to publisher monetization
PubRunner founder William Cichowski launched his AI advertising startup yesterday. The company personalizes open web and in-app publisher experiences while optimizing for user lifetime value rather than revenue per session.
Cichowski writes on LinkedIn:
“Instead of giving every user the SAME template of cluttered content, we enable publishers to serve dynamic, tailored, and user-friendly experiences that build up a long-term audience. We can do that because PubRunner’s objective function (the targeted goal that guides our machine learning system) is lifetime value instead of revenue per session. Because lifetime value looks at user engagement holistically, our approach grows both short-term and long-term publisher revenue.”
Read more. (Sept. 8)
PubRunner bills itself as “intelligent ad placement” in the title of its website.
From tipsheet: PubRunner is claiming to apply a personalization model similar to that of ChatGPT and social feeds to the publisher experience, with AI deciding how to balance engagement and monetization around the long-term value of each user.
PROTOCOLS
How DTE filters the bidstream
Bedrock Platform Head of Engineering Damian Naglak yesterday explained on LinkedIn how Dynamic Traffic Engine (DTE) works while positioning Bedrock around emerging programmatic infrastructure.
Donated by Amazon Ads to IAB Tech Lab in April (see tipsheet), sellers currently use a DSP’s past bidding behavior to decide which requests to send it. DTE lets the DSP tell sellers directly which requests it does not want based on its live campaigns.
Naglak writes in part:
“Every no-bid gets paid for twice, in CPU at both ends. The seller spends cycles building a request nobody wanted, the buyer spends cycles unpacking it to find out. Co-location removes the distance between them and leaves both costs in place, so dropping the obvious cases is what removes the work.”
Read more from Naglak. (Sept. 8)
From tipsheet: DTE creates a path for buying intelligence, including AI-generated decisions, to move upstream and shape the bidstream itself. Advertising’s agentic future will depend on compute efficiency, not just smarter models. Anyone (marketers, publishers, agencies, etc. — not just ad tech) seeking to own advertising intelligence may need to get their hands dirty with the hardware and infrastructure beneath it.
LLMs & CHATBOTS
Developments
- Introducing Muse: Your personal AI agent that gets things done (Sept. 8) – Meta
- Mistral raises €3B to make sovereign, open-weight AI the technology frontier (Sept. 8) – Mistral
- Build commerce agents with Claude (Sept. 1) – Anthropic
LLMs & CHATBOTS
Advertisers want a look inside ChatGPT Ads
Adweek’s Trishla Ostwal reports yesterday:
“For all the new bells and whistles that OpenAI has added to its ads business, some advertisers say there’s still a pretty basic piece of the puzzle missing: They want to know what’s happening inside ChatGPT when their ads appear. Four ad buyers interviewed for this story are asking for more visibility into the prompts that trigger ads…”
Read more: “Advertisers Want More Brand Safety, Prompt-Matching, Incrementality Tracking From ChatGPT Ads” (Sept. 8) – Adweek (subscription)
From tipsheet: Meanwhile, OpenAI is looking to balance advertising with user trust. Recall the “Ad Principles” published when ChatGPT ads were first announced in January, which included: “We keep your conversations with ChatGPT private from advertisers, and we never sell your data to advertisers.”
OpenAI is breaking new ground with the conversational format (Google, too) while revisiting the same old ground in digital advertising.
Related: Google tests traditional Search campaigns in AI Mode (Sept. 7) – Search Engine Land
TECH
Former AppNexus co-founder questions ad tech
AppNexus co-founder and former ad tech product leader Mike Nolet in a rare LinkedIn post yesterday:
- “Now 10+ years out of AdTech. Curious to know from those still ‘on the inside’, is there ANY backlash to LG’s abhorrent behavior or is everyone just doing ‘business as usual’ with ‘LG Ad Solutions’?…”
Nolet was responding to reports that LG smart TVs logged audio while screens were off, scanned local networks for nearby devices and transmitted collected data to LG.
Read more from Nolet. (Sept. 8)
Read: “LG smart TVs caught logging audio with screen off and snooping on local devices” (Sept. 7) – Notebook Check
MORE
- EU Search: Google warns of lower quality as it revamps Europe search results to avoid EU fines (Sept. 8) – Reuters
- AI creative: Why Kalshi dropped generative AI in favor of human creativity (Sept. 8) – Ad Age (subscription)
- Auditing the logs: 5 Analyses You Can Perform with Google Ad Manager Log Level Data (Sept. 1) – Burt Intelligence blog
- Remedies ruminations: “Here’s your cheat sheet for getting smart about what remedies may be ‘forced’ upon Google when the ruling is unsealed.” (Sept. 8) – Scott Messer, Ad Tech Therapy on LinkedIn
- Benchmarks wanted: 6 Months Into ChatGPT Ads, Advertisers Still Don’t Know What ‘Good’ Looks Like (Sept. 8) – Search Engine Journal