Consumer packaged goods (CPG) brands are using AI agents in advertising to build audiences, test and adapt creative, check work against brand and regulatory rules and, in a few live cases, to plan and run media buys from a plain-language brief. Published examples include a Clubtails campaign run by agents on connected TV, Nestlé México prompting agents to build audiences and L'Oréal brands joining an advertising pilot inside ChatGPT. Most large CPG advertisers still keep budget decisions and brand meaning with people, and some, such as Diageo and Georgia-Pacific, say so publicly.

TL;DR. The 12 examples below are grouped by job: buying, audiences and planning, creative and compliance, preparing for AI shopping agents, and deliberate restraint. Each is dated and sourced to the company involved or to trade press. The pattern is consistent: agents take on the mechanics, people keep the decisions that commit money or speak for the brand, and several brands are fixing data and supply paths first.

What counts as a CPG brand using an AI agent in advertising?

An AI agent in advertising is software that takes a goal, such as a campaign brief, and chooses the steps to reach it within limits a person has set. That separates it from a bidding script or a chatbot. Some examples below are agents in that strict sense and some are AI systems that lead towards them.

A useful way to place each example is the earned autonomy ladder: Observe, Recommend, Draft, Human-approved action and Bounded automatic action. Most CPG examples sit at Recommend or Draft, and only the buying examples reach further. Alkimi, for example, is building a neutral marketplace for agent-negotiated deals in which people on each side keep approval rights at the deal level, and agreed terms go into a DealSheet that both parties hold and either can retrieve independently.

For agencies, retailers and platforms, see the cross-industry list of 21 sourced examples by job.

How are CPG brands using AI agents to buy media?

1. Clubtails (Geloso Beverage Group), with Butler/Till: an agent-run CTV campaign

PubMatic and the independent agency Butler/Till launched an agentic connected TV campaign for Clubtails, a flavoured malt beverage brand from Geloso Beverage Group (PubMatic, 15 December 2025). Butler/Till submitted a natural language brief through Claude, and PubMatic's agents generated the media strategy, set up the campaign and managed pacing and targeting, with age-appropriate audience controls. Shared standards for how agents communicate are led by IAB Tech Lab's AAMP and the Ad Context Protocol (AdCP), published by AgenticAdvertising.org, and this workflow used AdCP.

2. Procter & Gamble: in-house planning and AI-prototyped ads

P&G moved almost all of its media-planning analysis, and in some cases buying, in-house to be more agile, Chief Brand Officer Marc Pritchard told MediaPost (MediaPost, 26 March 2026). He described AI tools that combine consumer data and insights to prototype ads "sometimes in minutes, sometimes in hours, sometimes in days". This is AI-assisted work rather than agent-run buying, but in-house planning is the foundation a buying agent would sit on.

3. Coca-Cola: agentic tools in its media review, and AI-timed TV ads in France

Coca-Cola's statement on its global media, data and technology agency review described "a shift in mindset from traditional media planning to the emerging ways we need to reach consumers through technology, including agentic tools" (Digiday, 5 June 2026). Earlier in the year, FranceTV Publicité trialled Context.IA with Coca-Cola France during the Milano Cortina 2026 Winter Olympics (FranceTV Publicité, 26 February 2026). The model analyses the images, text, audio and video of programmes, and Coca-Cola France's advert was triggered automatically during high-engagement moments in replays on francetv sport and francetv.fr.

How are CPG brands using AI agents for audiences and planning?

4. Nestlé México: prompting agents to build audiences across 70 brands

Nestlé México, which owns more than 70 brands, connects its customer data to Treasure AI, which now calls itself an "agentic experience platform" (AdExchanger, 9 October 2026). Elsa Vences, its data and personalisation manager, said a team can prompt the platform's agents with campaign parameters, such as an audience of Gen Z consumers who love chocolate, and the agent draws on data from brands such as KitKat.

5. Kind (Mars): synthetic audiences and personalised creative tests

Kind, the Mars-owned snack brand, began testing AI-powered creative in the second quarter of 2025 that showed its protein product in different settings for different customers (Digiday, 19 November 2025). It also built synthetic audiences from past audience work, campaign results and third-party syndicated data. "Now I can use my creative agencies more as brand stewards, more strategic thought partners and less like creative and production suppliers," said Osher Hoberman, CMO at Kind North America.

How are CPG brands using AI agents for creative and compliance?

6. Reckitt: automated artwork checks, with a human decision on every output

Reckitt built a tool that automatically checks campaign artwork against regulatory, brand and legal rules, so marketers can concentrate on design (Digiday, 20 November 2025). Bastien Parizot, its SVP of IT and digital, said the company uses a different large language model as a judge to reduce hallucination. "Every output is always for a human to make a decision to use or not to use," he said.

7. Unilever: Brand DNAi and product digital twins

Unilever's Brand DNAi turns decades of brand equity into a living context layer that teams anywhere can query, grounding new ideas in what each brand stands for, according to Selina Sykes, its global head of digital marketing and social (Adweek, 13 January 2026). Product digital twins let teams create new pack variations instead of reshooting them in each market. Sykes also argued that as people rely on AI agents to filter options, brands must understand how those systems "see" them.

8. Heineken: a lab building agentic systems for marketing content

Heineken opened a Global GenAI Lab in Singapore with AI Singapore to build "sophisticated agentic systems that can autonomously solve complex problems", including automated marketing content creation (Heineken, 19 March 2025).

How are CPG brands preparing for AI shopping agents?

9. L'Oréal: brands in the ChatGPT ad pilot and a brand agent for La Roche-Posay

L'Oréal said SkinCeuticals, CeraVe and Garnier are part of the global ChatGPT ad pilot, and that Maybelline New York will bring a makeup virtual try-on into ChatGPT (L'Oréal, 17 June 2026). L'Oréal also describes a brand agent for La Roche-Posay on large language model platforms, built on its own Beauty Knowledge Graph (L'Oréal).

10. Mondelez: a global lead for agentic commerce

Mondelez, owner of Oreo and Cadbury, advertised for a global lead of emerging commerce platforms with agentic commerce as the primary focus, to shape strategy, run pilots and build a playbook (Digiday, 20 April 2026). Andrew Lederman, its VP of global digital commerce, described a move "from clicks and channels to intelligent agents that guide decisions, discovery and transactions". Digiday reported that Mondelez's retailer partners expect 30% of their site traffic to be agentic by 2028.

Which CPG brands are holding back on AI agents, and why?

11. Diageo: AI that advises rather than executes

"We want to be seen to be using it responsibly, and in doing so, thinking more about how it advises what we do, as opposed to executing on our behalf," said Mike Cheetham, Diageo's head of culture, entertainment, media and digital (Digiday, 18 February 2026). He added: "We will get to a point where we start to think about how AI could execute on our behalf."

12. Georgia-Pacific: clean the supply path before adding an agent

Georgia-Pacific is holding off on buying agents after two years of cutting supply-side platforms and scoring inventory before the bid (Digiday, 19 August 2026). "Before we even go to a buying agent we have to clean it up or else the buying agent is going to essentially deliver subpar results," said Paras Shah, senior director of digital media. Shah said the company would still consider buying agents later.

What do these CPG examples have in common?

Agents take on the mechanics and people keep the meaning. Nestlé México's agents assemble audiences and Reckitt's tool checks artwork, but people make the final call.

Checking is designed in from the start. Reckitt uses a second model as a judge, and Unilever's NOVA, a system for processing customer orders rather than advertising, uses five agents, with the fifth checking the work and missing information and exceptions going to a person (Microsoft UK Stories, 26 October 2026). Any agent that commits media spend deserves the same split between doing and checking.

The record matters once money moves. In the Agree. Transact. Verify. simulation, 95.3% of deals had mismatched records when each side kept its own, against 0.19% with a shared record. That is simulated rather than live market data, but it shows why a deal agreed between agents needs one record both sides can check.

Clean inputs come first. As Georgia-Pacific's supply path work shows, an agent automates whatever it is given, including waste.

How should a CPG brand start with AI agents in advertising?

Start on the lower rungs of the ladder, with tasks where a person reviews every output. Before an agent buys media, write a mandate that sets budget limits, inventory scope, audience controls such as age gating for alcohol, and the points that need human approval, using a media buying mandate template. Retail media is where many CPG budgets are likely to meet agents first, as this explainer on retail media sets out.

Then test the agent before real money moves. Alkimi's Gym, available now, lets buyers and sellers build their own agents, set the rules they work to and spar against counterparty agents first, and its Concourse product benchmarks AI media-buying agents in live conditions.

Frequently asked questions

Are CPG brands letting AI agents buy media on their own? Rarely. Published cases such as Clubtails show agents running a campaign from a brief, but most large CPG advertisers keep budget decisions and final approvals with people.

Which jobs do AI agents do most often for CPG brands? Audience building, creative variation and compliance checks are the most common in published examples. These are tasks where a person can review every output before it reaches consumers or commits money.

Why are some CPG brands holding back on AI agents? Some, such as Georgia-Pacific, want to fix data and supply paths first so that an agent does not automate waste. Others, such as Diageo, want AI to advise before it executes.

How do AI shopping agents change CPG advertising? When shoppers ask AI assistants for recommendations, brands need product data and ads that those assistants can read. L'Oréal brands joining the ChatGPT ad pilot and Mondelez hiring for agentic commerce are early responses.

How CPG Brands Use AI Agents in Advertising