Tag: agentic AI
-
Most European Ad Execs Expect Agentic Buying Within a Year. Most Still Keep People Steering.
An IAB Europe survey of 50 ad executives finds 58% expect agentic ad buying to be in operational use or at scale within a year. Yet 36 of 47 say their organizations have no agent operating day to day, or one that humans still direct, and agents mostly handle reporting.
-
A Mortgage Lender Is Using ChatGPT Ads to Educate Borrowers Before Speaking to a Salesperson
Mortgage Solutions Financial is running ChatGPT ads that explain loan options beneath the chatbot’s answer instead of pushing an application, so borrowers can talk through potential issues such as missed payments and credit problems before facing a salesperson. CMO Dawn Dawson is also testing sponsored agents that pass leads straight into HubSpot.
-
Agencies Are Bolting AI Clauses Onto Contracts Written for Billable Hours
AI is shrinking the hours agencies bill for, and the contracts that govern that work were written for a different arithmetic. Four agency executives describe the fix so far: addendums, one clause at a time, covering IP, data and tool disclosure. Nobody has agreed on what AI should cost.
-
Time Prices an AI Agent’s Attention at Five to Seven Times Display
Time is selling ad impressions to AI agents at roughly five to seven times its display rate, on the argument that one agent read may shape later recommendations. Chasing impact on AI agents in a different way, Expedia has staffed a business-to-agent function for a channel that carried under 1.5% of its traffic in May.
-
Agencies Want Clients to Evaluate AI Agents Like New Hires
Some of the agencies building AI agents are the ones urging clients to slow down. Mod Op’s CTO says agents should be evaluated like new roles tied to growth, and Booking Holdings’ AI lead calls pricing and contracts a partnership exercise.
-
Gemini’s Billion Users Talk More Than They Type. Its Task Automation Runs in Two Countries.
Google says a billion people now use the Gemini app each month and 63% of them talk to it rather than type. The post naming that figure also says Gemini automates tasks across 40-plus apps. A July footnote limits the feature to two countries and a short list of phones.
-
AI Roundup: Pelé’s Lost Goal, a Referee’s-Eye View, and a Cap That Hides You From Cameras
A goal from 1959 that no camera recorded, rebuilt with a stunt double and Gemini Omni. A World Cup engine that finds you a second team when yours goes out. A cap that spots a phone pointed at you and covers your face. Six campaigns, one dependency.
-
Butler/Till Pointed Its Buying Agent at iHeart’s, Skipping the SSP
Two AI agents — one built by the agency, one by the publisher — met over an MCP server and placed just under $10,000 of podcast and streaming audio inventory. No SSP in the middle. Butler/Till says CPMs came in 42% below the client’s own benchmark for direct buying.
-
Perplexity Blocks Time’s Agent-Only Ads and Puts Publishers’ Trust Scores at Risk
Time built ads only AI agents can see. Perplexity has decided they will not count, calls the practice deceptive, and says publishers who run them risk a hit to their trust score. Its objection is to the ads existing, not to how they are labeled.
-
Monks’ Co-Founder: The Billable Hour Punishes the Agency That Uses AI Well
Every vendor at Ai4 was selling the same tenfold speedup, which Wesley ter Haar flags as a pervasive issue with AI tooling – when the speed is off the shelf, it is not a moat. The Monks co-founder argues the real constraint is the legacy model: agencies need to stop selling time, because getting faster…
-
S4 Doubled First-Half Profit. Sorrell Says the Token Spending Has Been “Ill-Disciplined.”
S4 Capital’s margins widened to 12.3% and its first-half operating profit doubled. Its chair says the company has been ill-disciplined on token spending, and Goldman Sachs expects token consumption to rise 24-fold by 2030. Monks passes token costs to clients at no margin.
-
AT&T Built a Router That Sends 45 Billion Daily Tokens to the Cheapest Model
Open models already carry a quarter of AT&T’s AI work, and its chief data and AI officer expects that to reach 70% to 80%. Cost is one reason — savings of 80% to 90% in some applications. Control over the company’s own data is the other.
-
Bayer’s Predictive Audience Models Are Beating Its DSP Targeting 4x in Small Tests
Bayer is testing models that spot customers before they churn and price bids by individual rather than cohort — four times standard DSP effectiveness, in small tests its media chief calls a scientific experiment. Its AI search playbook runs through medical journals, not media buys.
-
Mod Op Built Its Own AI Layer Rather Than License Seats for 500 Employees
Building rather than licensing saved Mod Op an estimated $3 million, a number describing a bill the agency never received. Its Orion platform sits over the major LLM APIs, automating agency workflows and, eventually, client ones. Most of the usage is still internal. No client has asked for a login yet.
-
WPP’s $1.8 Billion Enterprise Unit Is Selling Operating Models, Not Campaigns
WPP Enterprise Solutions now has a product line: five AI services, 12,000 people, $1.8 billion in revenue, and Accenture and Deloitte in its sights. The offer is an operating model rather than a campaign — and its CEO says the 2027 scorecard won’t be a dollar figure.
-
Unilever Scales to 300,000 Creators by Automating the Grunt Work
A year after declaring creators core to its marketing plan, Unilever has grown its creator program from 10,000 to 300,000 — and its executives are now deciding which parts of influencer marketing to hand to automated systems. “In most of the places where we are using technology, technology is used for us to augment the…
-
e.l.f., L’Oréal, and Tarte Are Running the AI Playbook Other CMOs Are Still Writing
While many chief marketing officers are still working out how to fold AI into their workflows, beauty marketers have quietly become the reference implementation. Brands like e.l.f. Beauty, L’Oréal, and Kosé Corp. (owner of Tarte and Decorté) are using AI for data mining, content versioning, chatting with loyalists, and product discovery — and have restructured…
-
Samsung TVs Get an AI Shopping Agent. Checkout Still Requires Your Phone.
Glance and Samsung formed a partnership to turn the living room screen into a personalized, AI-powered shopping destination — an agentic commerce experience rolling out across millions of Samsung smart TVs in the United States. Built on Glance’s proprietary AI platform and running on Samsung’s Tizen OS, it lets viewers create personalized shopping feeds and…
-
This Month in AI Activations: Anime Strikers, Digital-Twin Dogs, and Ads That Take Your Order
Ad Age’s latest Top 5 AI activations roundup arrives in a month when Meta shipped Muse Image, the first model from its Superintelligence Labs, and Google pushed lightweight versions of its Omni video and Nano Banana image models to cut marketers’ inference costs — models that increasingly reach advertisers directly inside the ad platforms.
-
Dept Builds a “Multiversal Assistant” and Won’t Bill Clients for the Glue
Dept is rolling out Deptify, an AI system accessed through a persistent assistant called “D.” The pitch is orchestration: D pulls context from whatever workflow tools a client already uses — Adobe Workfront, Asana, or something else — and routes people to the right tool for the job, instead of asking them to learn another…
-
AI Is Becoming Loyalty’s Gatekeeper
Gale’s latest loyalty report is less about points and more about the chatbot now standing between a brand and the person it insists on calling a “relationship.” In “The Preference Economy,” Gale surveyed 3,000 consumers in the U.S. and U.K. and found that 56% are comfortable delegating all communications with a brand through AI. Nearly…
-
Sephora Puts Checkout Inside Google’s AI Loop
Sephora expanded its partnership with Google to become the first prestige beauty retailer to enable shopping directly inside Google’s AI-powered platform through Google Agentic Checkout. In practice, this keeps shoppers on Google. They can ask beauty questions, research ingredients, compare recommendations in real time, build a routine, add products to basket, and check out without…
-
PMG Put $500,000 Behind a Cannes Hackathon and Kept None of the Code
Up to $500,000 in prize money, roughly two dozen entrants, three briefs drawn from real client problems — and PMG keeps none of the code. The agency says the Cannes hackathon demonstrates what AI has done to technical development. It is also a look at who can build.
-
Gap Splits Its AI Marketing Rebuild Across Google Cloud, Zeta and Publicis Sapient
Gap has partnered with Google Cloud, Zeta Global, and Publicis Sapient to transform its marketing approach. The collaboration aims to create a scalable, real-time marketing organization using integrated data and AI. Google Cloud provides the data foundation, Zeta builds the marketing stack, and Publicis designs the operating model. The initiative focuses on enhancing content relevance,…
-
McDonald’s Tests “Archy” at Five Drive-Thrus
McDonald’s is testing an AI-powered drive-thru ordering system at five U.S. locations, moving the familiar ritual of shouting lunch through a speaker box one step closer to automation. The system is called ArchIQ, though a franchise owner says its nickname is “Archy,” because even automation apparently needs a mascot. It was shown at McDonald’s 2026…
-
JPMorgan Wants AI Agents That Don’t Need a Bathroom Break
JPMorgan Chase says it plans to deploy “long-running autonomous” AI agents later this year — systems meant to move beyond one-off prompts and instead manage multi-step workflows across different software tools. Derek Waldron, the bank’s chief analytics officer, says these agents can run for “an hour or two,” compared with the “two or three minutes”…
-
Walmart’s AI Push Runs Into the Token Meter
Walmart executives are talking about AI the way retailers have long wanted to talk about personalization: as something that understands intent and language, not just item history. At Walmart’s Associates Week in Bentonville, CEO John Furner said AI is helping teams finish technology projects faster across the company. He also kept the brakes lightly tapped:…
-
When Product Ships Weekly, Marketing Becomes the Bottleneck
HBR’s premise is blunt: AI sped up engineering first, and marketing is now stuck doing quarterly-era coordination in a continuous-release world. Citing Anthropic research, the piece notes that software engineering accounts for “nearly 50% of all agentic activity.” That helps explain why product teams are shipping continuously, and why marketing — with its cross-functional handoffs,…
-
AI Agent Sprawl: When Everyone Gets a Bot, IT Gets a Headache
As tools like Anthropic’s Claude Cowork make it easy for nontechnical employees to spin up task-specific AI “agents,” some large companies are discovering a new flavor of shadow IT: too many bots, doing too many overlapping things, in too many places. The Wall Street Journal flags “AI agent sprawl” as the problem, fueled partly by…
-
Agency Account Teams Are Turning AI Into Client-Specific Utilities
Coming back from maternity leave in 2024, Wendy Owen, (Senior Account Manager Publicis Pro) says the biggest surprise wasn’t staffing changes. It was how casually AI had slipped into day-to-day agency operations, especially in client services, where ambiguity and speed are basically the job description. In this telling, AI’s main impact isn’t “automation” so much…
-
Hershey’s Wants Its AI Agents Measuring the Candy Spend, Not Just Buying It
While AI media-buying agents hog the attention, Hershey’s has been applying agentic tech to something less glamorous: media mix modeling. According to Digiday, the confectioner worked with Mutinex and Tracer.tech and spent a year building a system of several AI agents to produce campaign media modeling in weeks rather than months. The goal is to…
-
The Trade Desk Lets Claude Fill Out the Paperwork—Just Don’t Call It “Buying” Yet
The Trade Desk is talking to agency execs about Koa Agents, a set of agentic features meant to automate large parts of programmatic setup and troubleshooting. The system connects with whatever AI model a partner is using — including Anthropic’s Claude — through an interoperability layer called Open Agentic Kit. According to an agency exec…
-
Duluth Lets AI Bid, But Keeps Humans on the Jokes
Duluth Trading Company is taking the agentic media buying trend and drawing a fairly human boundary: AI agents can handle tasks like bidding and creative iteration management, but anything resembling brand voice, humor, or “ethos” stays closer to people with pulse rates. Marketing director Ellie Uberto framed it during a live Digiday Podcast recording at…
-
Starbucks launches beta app in ChatGPT to fuel new drink discovery
Starbucks just launched a beta “app” inside ChatGPT designed to spark drink ideas when customers don’t feel like starting with a menu. To use it, people enable Starbucks via ChatGPT’s app directory, then prompt the bot with “@Starbucks.” It can suggest and customize drinks and even let users pick a store location—then it hands you…
-
$1M, 21 AI Spots, and a Ticket to Cannes: Luma’s “Dream Brief” Goes Public
Luma AI’s “Dream Brief” set out to get working creatives hands-on experience with the product. Over two months, the contest pulled nearly 400 AI-made ad submissions from 58 countries and more than 100 agencies, then a panel of heavy hitters narrowed it to 21 spots headed to Cannes Lions. The hook is simple (and very…
-
Wpromote’s AI Strategy: Let the Bots Pitch, But Keep the Keys
Kargo has rolled out “Project Kera,” a chat-style planning and buying interface (closed beta) aimed at making cross-channel activation feel less like stitching together five different dashboards. The pitch: one agentic layer that can pull together biddable and non-biddable inventory, with beta testers including Hershey, travel media network Navigator, and indie agency Wpromote. Kargo’s CEO…