Seventy percent of German advertisers already use generative AI in their daily work. Only 9% have governance rules for what happens when those AI systems start making decisions on their own. That gap is the entire story of the OWM/Accenture Song study published March 17, 2026, the first German industry governance study focused specifically on autonomous AI agents in marketing.
The Organisation Werbungtreibende im Markenverband (OWM), Germany’s largest advertiser association, partnered with Accenture Song to survey their member companies and approximately 1,000 German consumers. The findings paint a picture of an industry sprinting toward adoption while barely glancing at the guardrails.
What the OWM Study Actually Found
The headline numbers are stark. Among OWM member companies:
- 68% use generative AI for content creation on owned channels (websites, CRM communication)
- 64% use it for idea development and brainstorming
- 55% use it for image and video creation
- 46% report noticeable workflow changes from GenAI adoption
Those numbers describe generative AI usage. When the study shifts to agentic AI, the systems that act autonomously rather than just generating content on demand, the picture changes dramatically.
Only 23% of OWM members count agentic AI among their top three strategic marketing topics. About half are working on pilot projects or concrete use cases. And here is the number that matters most: 91% report that agentic AI currently influences less than 10% of their marketing decisions.
That 91% figure will not last. 60% expect agentic AI to influence up to 30% of their decisions by 2027. In agency work specifically, respondents project that 30% of all decisions will be at least partially influenced by AI within one year.
“Künstliche Intelligenz hat das Marketing bereits heute grundlegend verändert. Mit Agentic AI beginnt jetzt eine neue Phase dieser Entwicklung,” said Susanne Kunz, CEO of OWM. AI has already fundamentally changed marketing. With agentic AI, a new phase begins.
What Consumers Think
The consumer survey of approximately 1,000 German respondents adds a critical dimension. 73% believe AI agents are already being used in advertising today. They are mostly right. But their expectations come with conditions:
- 43% expect improved customer outreach and more efficient advertising content through AI
- Consumers accept supportive AI roles: recommendations, content preparation, personalization
- They demand human control over autonomous decisions, especially in sensitive areas
- Transparency and traceability are non-negotiable expectations
The disconnect between what marketers plan (increasing autonomy) and what consumers expect (human oversight) is where governance becomes unavoidable.
The Autonomy Question: Where Human Control Must Stay
The study’s most practical contribution is its framework for thinking about where autonomy boundaries should sit. Karin Libowitzky, Managing Director and Data & AI Practice Lead at Accenture Song, put it directly: “Agentic AI fundamentally shifts the role of marketing teams. While AI systems can increasingly make operational decisions, the strategic framework, goals, brand management and governance, becomes even more strongly a task for humans. Companies must define today where they allow autonomy and where human control remains indispensable.”
This is not abstract advice. The study identified specific governance gaps that companies need to close:
1. Autonomy boundary definitions. Most companies have no documented policy on which marketing decisions an AI agent can make independently versus which require human approval. When an AI agent adjusts ad spend, retargets a campaign segment, or generates customer-facing copy, who signs off? In most organizations right now: nobody, because the question has not been formally asked.
2. Agency transparency. Advertisers explicitly told OWM they want to know what AI does inside their agencies and how it affects campaign results. One unnamed brand manager quoted in the study said: “We want to know what AI does in the agencies and how it affects our results.” That demand for visibility will reshape agency-client relationships.
3. Brand safety codification. Traditional brand guidelines are written for human interpreters. AI agents need machine-readable policies: explicit rules about tone, imagery boundaries, prohibited associations, and escalation triggers. Few marketing organizations have created these.
4. KI-Kompetenz as a standard. The study recommends that AI competence should become standard across agencies and marketing teams. This aligns with Article 4 of the EU AI Act, which requires AI literacy training for all staff working with AI systems.
What Other Markets Are Doing
The OWM study does not exist in a vacuum. Singapore released the world’s first agentic AI governance framework at Davos in January 2026, built on four dimensions: risk assessment, human accountability, technical controls, and end-user responsibility. The IAB Tech Lab released its Agentic RTB Framework (ARTF) v1.0 in January 2026, specifically addressing how AI agents interact in programmatic advertising.
Germany’s contribution through the OWM study is narrower but deeper: it asks the governance question specifically from the advertiser’s perspective, not the regulator’s or the platform’s.
Marketing AI Agents Under the EU AI Act
The governance question is not optional. The EU AI Act’s transparency obligations under Article 50 and high-risk system requirements become fully enforceable on August 2, 2026. For marketing specifically, the Act creates three categories of concern:
Prohibited practices. Subliminal AI techniques that manipulate consumer behavior without awareness are banned. So is exploiting vulnerabilities based on age, disability, or socio-economic status to distort purchasing behavior. Marketing AI agents that personalize at scale must be able to prove they do not cross these lines.
Transparency requirements. Users must be informed when interacting with AI (chatbots on brand websites, AI-powered customer service). AI-generated content must be identifiable. Deepfakes and AI-generated text for public interest purposes must be clearly labeled. Autonomous ad-targeting systems like Google Smart Bidding, Meta Advantage+, and TikTok’s automated optimization tools fall within scope.
Penalty exposure. The stakes are real: up to EUR 35 million or 7% of global annual turnover for prohibited practices, EUR 15 million or 3% for other non-compliance including transparency violations, and EUR 7.5 million or 1% for providing inaccurate information to authorities.
The OWM study’s timing, five months before the August enforcement deadline, is not accidental. German advertisers need governance frameworks that satisfy both their own oversight needs and the legal requirements coming in August.
Who Is Already Building Marketing Agent Systems
The study arrives as marketing AI agents move from concept to commercial deployment:
Serviceplan Group launched Serviceplan Agents in March 2026, AI coworkers for SMB marketing on the Sokosumi platform. Their agent “Hannah” handles marketing research (market sizing, competitive analysis via Statista and GWI data), while “Elena” manages account and project coordination. Both operate within European data protection requirements.
Accenture Song itself has equipped 600+ professionals with agentic AI tools, reducing campaign steps from 135 to 85 and achieving 25-35% faster time to market.
NBCUniversal/FreeWheel completed the first AI-agent-led programmatic guaranteed deal including live sports programming during NFL playoff games in Q1 2026.
PubMatic launched AgentOS, an operating system for autonomous advertising execution on NVIDIA-accelerated computing, with launch partners including WPP Media and MiQ.
Viant’s Lattice Brain tested a “no-human-in-the-loop” approach for retailer MacKenzie-Childs, achieving a $15 cost per action versus $45 from human traders.
These are not experiments. They are commercial deployments generating measurable results. The governance question is not whether to adopt, but how to maintain control while adopting.
Building a Marketing AI Governance Framework
Based on the OWM study’s recommendations and the broader regulatory context, a practical governance framework for marketing AI agents needs five components:
Decision classification. Map every marketing decision an AI agent could make and classify it: fully autonomous, autonomous with monitoring, human-in-the-loop, or human-only. Ad spend allocation above a threshold might require human approval. A/B test variant selection might be fully autonomous. The key is making these boundaries explicit.
Audit trails. Every autonomous action by a marketing AI agent should be logged with enough detail for post-hoc review. What data did the agent access? What decision did it make? What was the outcome? This is not just good governance; it is an EU AI Act requirement for systems that generate content or make targeting decisions.
Escalation protocols. Define precisely when an AI agent must escalate to a human. Brand safety violations, spend anomalies, performance degradation beyond a threshold, or any decision affecting sensitive audience segments should trigger escalation. The OWM study found that most companies have not yet built these escalation pathways.
Agency contracts. Update agency agreements to require transparency about AI usage in campaign execution. The unnamed brand manager’s demand, “We want to know what AI does in the agencies,” needs to become a contractual clause, not a wish.
Competency building. The study’s call for AI competence as an industry standard means training programs, not just for using AI tools, but for governing them. Marketing teams need to understand what agentic AI can and cannot do, where risks concentrate, and how to read audit logs.
McKinsey’s State of Marketing Europe 2026 report found that 94% of European marketing organizations have not yet advanced their GenAI maturity. The 6% who describe their use as “mature” report 22% efficiency gains. The gap between those who govern AI well and those who avoid it entirely is becoming a competitive divide, not just a compliance checkbox.
Frequently Asked Questions
What is the OWM AI agent governance study?
The OWM (Organisation Werbungtreibende im Markenverband) partnered with Accenture Song to publish the first German governance study on autonomous AI agents in marketing in March 2026. It surveyed OWM member companies and approximately 1,000 German consumers about AI adoption, governance gaps, and autonomy boundaries in marketing.
How many German marketers use AI agents in 2026?
According to the OWM study, 70% of member companies use generative AI daily, with 68% using it for content creation. However, only 23% rank agentic AI (autonomous AI agents) among their top three strategic priorities, and 91% say agentic AI currently influences less than 10% of their decisions.
How does the EU AI Act affect marketing AI agents?
The EU AI Act’s transparency obligations take effect August 2, 2026. Marketing AI agents must comply with rules on consumer interaction disclosure, AI-generated content labeling, and prohibitions on subliminal manipulation. Penalties reach up to EUR 35 million or 7% of global annual turnover.
What governance framework does the OWM study recommend for marketing AI?
The OWM study recommends defining clear autonomy boundaries for AI decisions, establishing transparency requirements between agencies and advertisers, codifying brand safety rules in machine-readable formats, and making AI competence (KI-Kompetenz) a standard across agencies and marketing teams.
Which companies are already using autonomous AI agents in marketing?
Serviceplan Group launched AI coworker agents for SMB marketing in March 2026. Accenture Song equipped 600+ staff with agentic AI tools, reducing campaign steps by 37%. PubMatic launched AgentOS for autonomous ad execution. Viant’s Lattice Brain achieved $15 cost per action versus $45 from human traders in testing.