Forrester 分析智能体广告的生产力悖论与工作流重组

Agents don’t automatically make advertising more productive. Layering AI onto broken processes merely automates inefficiencies. The bigger opportunity is reengineering advertising around a new division of labor, where people set strategy and guardrails while agents orchestrate execution and optimization. Connected data provides the context both need to make effective decisions.

We’ve seen this phenomenon before. In the 1980s and early ’90s, the Solow paradox recognized that massive investments in IT don’t automatically lead to measurable increases in labor or economic productivity. In 2000, researchers found that for every dollar a company spends on computer hardware, it needs to invest up to $9 in rewriting business processes, restructuring team hierarchies, and retraining employees. Technology only creates disproportionate value when it’s paired with organizational capital, new processes, restructuring, and new ways of working.

Bearing that in mind, here’s how agentic AI will reshape advertising’s essential workflows:

• Media planning will become continuous decisioning. Today’s media plan is a periodically refreshed artifact. Agents will turn it into a living decision system by continuously incorporating campaign history, audience intelligence, media costs, creative test results, and business outcomes to recommend where the next marginal dollar belongs. Planners’ value will shift from constructing plans to defining how and when plans should change.
• Media buying will shift from interfaces to intent. Instead of traders translating media strategies into hundreds of platform settings, they will express objectives, constraints, audiences, economics, and brand safety requirements for agents to execute on. Agents that interact directly with each other across ad platforms will reduce the importance of centralized platform interfaces. The most valuable buying skill will be knowing what instructions and guardrails to encode.
• Media optimization will hinge on business outcomes, not campaign KPIs. Agents will extend optimization across bids, budgets, audiences, and creatives simultaneously and continuously. Faster, more thorough optimization raises the stakes of choosing the right objective. Advertisers with routine access to high-quality proprietary outcome signals like lifetime value, margin, incrementality, inventory, and profitability will accrue advantages.

Prepare Media Teams For Machine-Mediated Planning And Buying

Media planners and traders shouldn’t wait for agents to mature before redesigning workflows. By then, competitors will have already turned planning into continuous decisioning, buying into machine-speed execution, and optimization into always-on learning. The work starts now:

• Reorient planning around incisive decision-making. Decompose media planning into the decisions humans make today, including shifting dollars between channels, adjusting reach and frequency targets, and deciding when deteriorating marginal returns warrant reallocating spend. Codify the thresholds and guardrails behind each one. Then determine where agents can recommend, approve, and autonomously act and rebuild planners’ roles around scenario design, experimentation, and exception handling.
• Build an agent-ready architecture. Map exactly which decisions agents can execute and under what conditions. Then establish dollar-based approval thresholds, brand safety constraints, frequency caps, and kill switches. Make first-party audiences, conversion signals, creative metadata, and unit economics machine-readable and portable across demand-side platforms, retail media networks, search and social walled gardens, and open web publishers. An agent that only sees data inside one platform will optimize for that platform’s yield, not advertisers’ return.
• Reskill practitioners for agentic governance. Train practitioners to interrogate why an agent moved $100,000 from CTV to paid social, distinguish a genuine performance signal from attribution noise, recognize when an agent is overindexing on cheap conversions, and override it when short-term return on ad spend conflicts with incrementality. Rewrite job descriptions and incentives to reward practitioners for designing effective experiments, diagnosing agents’ failures, and determining when machines should act autonomously.

Stay tuned for more research from Forrester on machine-mediated planning, buying, and optimization. As always, feel free to request a guidance session to discuss.

Blog

Amazon Ads unBoxed 2026: Amazon Makes Its Pitch For Outcome-Based Advertising

Agentic commerce and agentic advertising were on full display at Amazon Ads unBoxed 2026. Amazon demonstrated how predictive, generative, and agentic AI can take on the work of helping consumers decide what to buy and helping advertisers turn business objectives into media outcomes. Over the course of the event, we heard directly from Amazon’s leaders […]

Blog

AI Governance Has A Strategy Problem

The good news: AI governance is on the enterprise agenda. The problem: It still isn’t embedded deeply enough in enterprise AI strategy. According to Forrester’s most recent data, only 42% of AI decision-makers say governance strategy is included in their organization’s official AI and agentic AI strategy. Policies may be proliferating, but strategic integration is […]