- Google calls AI agents the third stage of AI in marketing. An agent doesn't just suggest changes; it makes them on the team's behalf.
- In May 2026, Google introduced Ask Advisor, an agent for Google Ads, Analytics and the Marketing Platform, launching as a beta for English-language accounts.
- Before an agent touches bids or budget, we audit conversion measurement. Every decision an agent makes is based on the data that reaches the account.
- Read-only tasks are the best place to start. Tasks with access to customer data or budget need sign-off first.
- We judge agents by the incremental profit they generate alongside your team. Time saved on its own isn't a result.
Google uses the term “Infinite Marketing” for marketing in which AI agents handle analysis, ad creative and campaign management. Tarun Rathnam of Google Cloud introduced the model on Think with Google in October 2025.
Hardly anyone in Germany searches for the term itself: around 10 searches a month. “KI-Agenten” (AI agents) gets around 4,400 searches a month (our own analysis of Google Ads data via DataForSEO, Germany, average for September 2025 to August 2026, as of October 2026).
In this article, we cover what AI agents in marketing do today, where to start and what needs to be in place in your accounts first.
An AI agent makes changes on its own
Google distinguishes three stages of AI in marketing. Predictive AI calculates things like customer lifetime value or churn probability. Generative AI creates text, images and video.
According to Google, the third stage is agents: AI systems that plan, act and learn. They carry out tasks on the team's behalf and under its control.
In an ad account, that means an agent builds audiences, changes bids or shifts budget. In other words, it makes decisions about money that used to be made by a person.
Google is building agents into its ad tools
In May 2026, Google introduced Ask Advisor. According to Google, this agent coordinates specialized agents across Google Ads, Google Analytics and the Google Marketing Platform, with a Merchant Center agent to follow.
Google's example: asked to find new customers for hair care products, Ask Advisor pulls product data from Merchant Center and uses it to set up a Google Ads campaign. It also analyzes Google Ads and Analytics data and recommends next steps.
Ask Advisor launched as a beta for English-language accounts. Google plans to roll out more features over the course of 2026.
Google's announcement doesn't say whether Ask Advisor submits changes for approval before making them. Clarify this before using it in your account for the first time.
We audit measurement before an agent decides on budget
The conversions recorded in the account determine which results the agent should optimize for. If the account counts purchases twice or misses inquiries, the agent makes every decision on that flawed basis.
So we start by reviewing search queries, ads, landing pages and conversion data. If sales also happen without a click, we consolidate and verify budget, revenue and promotion data from every source.
Our article on conversion tracking in Google Ads explains how we set up measurement in Google Ads. Our page on measurement and attribution shows how we bring the impact of every channel together in one model.
Start with read-only tasks
Google recommends a phased rollout: first individual, recurring tasks, then connected agents and finally an orchestrated multi-agent system. As starting points, Google suggests the weekly report, categorizing YouTube comments and building audiences from customer data.
We also sort these tasks by whether the agent only reads data or changes something in the account:
| Task | What the agent does | Prerequisites |
|---|---|---|
| Compile the weekly report | reads data from Google Ads and Analytics | a shared definition of conversions and revenue |
| Categorize comments and reviews | reads and summarizes | spot checks reviewed by your team |
| Build audiences from customer data | uploads customer data to the ad account | user consent and sign-off from your data protection team |
| Adjust bids and budget | changes spend | verified conversion data, limits on changes and a log of every change |
The first two tasks can be tested without putting budget at risk. The last two touch customer data or spend and need sign-off first.
Account, data and approvals stay with you
According to Google, you don't need to replace existing systems to use agents. Google describes three layers: your own data, models such as Gemini with agent-building tools, and connections to work tools such as Google Ads.
With agents, the same rule applies as for any campaign: the account and the data are yours. Before deployment, we agree with you where data is processed, who has access and how every change is documented.
We judge agents by incremental profit
Google proposes three metrics. The first measures how quickly an insight turns into action in the market. The second measures the share of routine tasks that are automated, and the third the incremental growth or profit generated by the team and agents together.
We judge agents by the third metric. Time an agent saves only counts for us when it goes into work that generates more profit.
How we use AI in our own work
AI helps us work faster. We use the time saved to look more closely and find better solutions.
We remain accountable for analysis, recommendations and decisions. We never rely solely on a system producing a plausible-sounding answer.
Google sees the competitive edge in the experience a team builds working with agents. We train your team to use AI tools on your own accounts and tasks. Learn more on our training page.
Conclusion: We deploy agents where measurement is sound
Google is bringing agents into its ad tools step by step. If you want to use them, start with read-only tasks and audit your measurement first.
The benchmark is the same as without agents: the profit an action generates. Our Google Ads page shows how we manage campaigns to that standard.
