Agentic AI in Marketing in 2026: What Agents Can Do Today

In May 2026 Google introduced Ask Advisor, a single agent built with Gemini that works across Google Ads, Google Analytics and Google Marketing Platform and can set up a campaign from a plain-language request ¹. Four months later Amazon renamed its whole advertising platform Amazon Ads Agent ². Agentic AI in marketing has moved from conference keynotes into the tools teams log into every day. What these agents can actually do, and how much control marketers should hand over, is less clear than the announcements suggest.

This article separates what is available now from what is still promised, looks at the adoption data, and sets out the governance a marketing team needs before giving an agent the keys.

What "agentic" means in practice

A generative AI assistant answers questions or drafts content when asked. An agent is given a goal and takes a series of steps to reach it: pulling data, choosing actions and, in some cases, making changes in a system. Gartner has warned about "agent washing", the rebranding of existing assistants, robotic process automation and chatbots as agents, and estimated in 2025 that only about 130 of the thousands of vendors claiming agentic AI were genuine ³. Gartner analyst Anushree Verma added that "many use cases positioned as agentic today don't require agentic implementations" ³.

A useful test for any tool is to ask three questions: what goal does it work towards, what systems can it act in, and which actions need human approval?

What agents can do today in ad platforms

The large ad platforms have shipped working agents, though most still keep a human in the approval loop.

Platform	Agent	What it does	Human control
Google Ads	Ads Advisor (from late 2025)	Recommends campaign actions, generates keywords and assets, analyses performance and helps fix policy disapprovals 4	Applies changes only with the advertiser's review and approval 4
Google (cross-product)	Ask Advisor (beta, May 2026)	Coordinates specialist agents across Ads, Analytics and Marketing Platform; builds campaigns from a brief and reviews results against goals 1	Beta for English-language accounts; approval flow not detailed 1
Amazon Ads	Amazon Ads Agent (September 2026)	Unified platform for sponsored, display, video and audio ads; conversational media planning, targeting from plain-language descriptions and natural-language analytics 2	Advertisers can keep manual control, hand optimisation to Amazon's AI, or combine both 2

Amazon reported that US advertisers who applied targeting recommendations through its earlier conversational tool saw more than 25% additional unique customers and more than 10% lower cost per impression on average ². As with any platform-reported figure, it is worth testing in your own account.

Programmatic buying is also being rebuilt for agents. In October 2025 a group of ad tech companies including Yahoo, PubMatic, Scope3, Optable, Triton Digital and Swivel launched the Ad Context Protocol (AdCP), an open standard for buyer and seller agents to plan and negotiate media ⁵. Even its backers stressed that "there is always a human in the loop" for large campaigns and creative, and that full agent-to-agent workflows were not yet working end to end ⁵.

Analytics and content operations

Analytics is where agents are most immediately useful, because the risk of a wrong answer is lower than the risk of a wrong bid. Google's Analytics Advisor, rolled out to English-language accounts from late 2025, provides performance summaries, investigates "why" questions about changes in the data and suggests next steps with instructions ⁴. Ask Advisor extends this by reviewing performance across Google Ads and Google Analytics together ¹.

In content operations, personalisation is the top AI use case for marketers, according to Salesforce's 2026 State of Marketing report, which surveyed 4,450 marketing decision-makers in late 2025 ⁶. Typical agentic workflows include generating and resizing ad variants, drafting localised copy, tagging assets and routing work for approval. The constraint is less the AI than the data feeding it.

The adoption data

Interest is high, but deep deployment is uneven:

• 75% of marketers say they are using AI to help close their personalised-content gap ⁶.
• 88% have started optimising for AI-generated responses in search ⁶.
• 48% have not yet worked out how to adapt their strategies to widespread AI use ⁶.
• Marketers with fully unified data are 60% more likely to use AI agents to scale their work ⁶.
• In a January 2025 Gartner poll of 3,412 webinar attendees, 19% reported significant investment in agentic AI and 42% conservative investment ³.

Gartner expects at least 15% of day-to-day work decisions to be made autonomously by agentic AI by 2028, up from none in 2024 ³. It also predicts that over 40% of agentic AI projects will be cancelled by the end of 2027 because of rising costs, unclear business value or inadequate risk controls ³.

Risks to manage

The main risks for marketing teams are practical rather than theoretical:

1. Spend and performance risk. An agent optimising to the wrong signal, such as cheap leads rather than qualified ones, can waste budget quickly.
2. Data quality. Salesforce found only 51% of marketers have complete access to commerce data and 56% to sales data ⁶. Agents acting on partial data make confident but wrong decisions.
3. Brand and accuracy risk. Generated copy and images can drift from brand guidelines or make claims that need legal review.
4. Regulatory risk. In the EU, the AI Act's Article 50 transparency obligations, including telling people when they are interacting with an AI system, applied from 2 August 2026 ⁷. The Digital Omnibus, which entered into force in July 2026, gives generative AI systems already on the market until 2 December 2026 to meet the watermarking requirement, and postpones obligations for Annex III high-risk systems, such as recruitment and credit scoring, to December 2027 ⁷⁸.
5. Vendor risk. Agent washing means paying agent prices for assistant features ³.

A governance checklist

Before giving an agent the ability to act, put these in place:

• Define permissions by action. Allow agents to analyse and recommend freely, but require approval for budget changes, new campaigns and anything customer-facing.
• Set hard limits. Use platform caps on daily spend, bid targets and change volume.
• Keep an audit trail. Log every change an agent proposes and applies, and who approved it.
• Fix the data first. Feed agents with clean conversion, CRM and product data, ideally from a single warehouse.
• Measure incrementally. Compare agent-managed activity with a holdout or a manually managed control.
• Label AI interactions. Disclose chatbots and AI-generated content where EU rules or your own policy require it ⁷.

Key takeaways

• Google and Amazon now offer working marketing agents inside their ad platforms, mostly with human approval for changes ¹²⁴.
• Agent-to-agent media buying has an emerging open standard in AdCP, but is not yet routine ⁵.
• Most marketers use AI, but unified data is the strongest predictor of agent use ⁶.
• Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027 and warns of agent washing ³.
• EU transparency rules for AI interactions apply from August 2026, with watermarking for existing systems due by December 2026 ⁷⁸.

What this means for your brand

Start where the risk is lowest: use analytics and reporting agents to speed up insight, and keep approval steps on anything that spends money or reaches customers. Invest in unified first-party data before scaling agent use, because the agent is only as good as what it can see. Ask every vendor exactly which actions their agent takes on its own. We would also map your chatbots and AI-generated content against the EU AI Act's transparency rules now, ahead of the December 2026 watermarking deadline.

Sources

1. Meet Ask Advisor, your new AI-powered collaborator - Google, May 2026
2. Amazon Ads Launches Unified AI-Powered Ad Platform - LBB Online, September 2026
3. According to Gartner, the future isn't actually agentic - SDxCentral, June 2025
4. Google's AI advisors: agentic tools to drive impact and insights - Google, November 2025
5. The Ad Context Protocol Aims To Make Sense Of Agentic Ad Demand - AdExchanger, October 2025
6. 75% of Marketers Have Adopted AI, Yet Still Use It to Send Generic Campaigns - Salesforce, February 2026
7. EU AI Act Omnibus Agreement — Postponed High-Risk Deadlines and Other Key Changes - Gibson Dunn, May 2026
8. The EU AI Act Deal: What It Means for Consent Infrastructure - Usercentrics, July 2026

Figures and product details are as reported by the sources listed, at the time of writing. Platforms change quickly, so check the latest documentation before acting on any single number.