AI agents now control marketing campaigns, and the infrastructure shift happened faster than expected. December 2025 brought three turning points that will define how campaigns execute in 2026.
Infillion launched the first agent-native advertising platform on December 16. Gartner and WARC project $100 billion in programmatic spend will flow through AI agents by 2028. Dover Fueling Solutions introduced retail media to gas station fuel pumps.
How AI Agents Control $100B in Ad Spend
The Agent-Native Platform Launch
Infillion Agent Connector operates entirely through Model Context Protocol (MCP). No dashboards, no manual clicking. AI agents send campaign instructions in natural language. The platform handles budget allocation, audience targeting, creative optimization, and bidding automatically.
Stagwell Marketing Cloud connected their Palantir intelligence layer within days. CTO Mansoor Basha: "We send campaign parameters, partners execute, and results flow back. Infillion's MCP-powered platform enables true interoperability."
Traditional DSPs added AI features to existing workflows. Agent-native platforms eliminate the workflows entirely.
The $100B Market Split
The programmatic advertising market is dividing into two approaches:
- Closed proprietary systems lock you into one AI agent
- Open execution layers work with any AI you choose
Infillion's platform is AI-agnostic. ChatGPT, Claude, and custom agents all connect through the same MCP interface.
Full lifecycle coverage:
- Campaign planning and budget pacing
- Identity resolution and audience building
- Creative optimization and A/B testing
- Real-time bidding and supply management
- Attribution and performance reporting
Each module operates independently or chains together for autonomous execution.
What Changes for Marketers
Campaign managers become AI orchestrators. The job shifts from platform navigation to setting constraints that agents execute.
Speed increases dramatically. Weeks of planning become minutes of natural language instructions.
Integration complexity disappears. One MCP connection replaces dozens of platform-specific APIs.
$100 billion represents roughly Meta and Google's combined display ad revenue. That's the amount Gartner and WARC project will flow through AI agents by 2028.
Learn more: How Model Context Protocol enables AI agents to access marketing data
Why Composable CDPs Grew 4x Faster Than Traditional CDPs
The Growth Signal
Composable CDP vendors (Hightouch, RudderStack, Census) grew nearly 4x faster than traditional CDPs in 2025. Their growth rate was 12.9% versus 3.4% for traditional vendors. They still represent less than 5% of the CDP market by size.
The overall CDP market reached $8.53 billion in 2025, projected to hit $37.11 billion by 2030. That's a 30.7% compound annual growth rate.
How Composable Architecture Works
Traditional CDPs copy your customer data into new systems. Composable CDPs connect directly to your existing data warehouse. Snowflake, BigQuery, or Databricks.
The data stays put. The CDP becomes a decision layer on top.
Hightouch co-founder Tejas Manohar: "The data warehouse is here to stay. While many CDPs now claim 'Composability,' true warehouse-native solutions are winning as businesses prioritize unbundled, warehouse-first architectures."
The Practical Difference
Implementation: Weeks instead of 6-12 months
Cost: No duplicate storage fees
AI-ready: Agents query your entire data universe, not pre-defined segments
The limitation: You need a data warehouse first. For teams without Snowflake or BigQuery, traditional CDPs still make sense through 2026.
Retail Media Expands to Gas Pumps and Social Platforms
Beyond Amazon and Walmart
Retail media grew 20% in 2025, but the category now extends far beyond the traditional players.
Walmart Connect grew 33% last quarter. That's six times faster than Walmart's overall sales. The platform offers closed-loop measurement with TikTok, major TV networks, and Meta. Advertisers can now track whether someone who saw a TikTok ad purchased at Walmart.
4CourtMedia launched December 18 with 13,000 screens across 1,500 gas stations. Independent operators gain access to national advertising dollars they couldn't reach before.
Other 2025 expansions:
- Casey's General Stores partnered with GSTV for pump advertising
- Instacart signed deals with TikTok for shoppable formats
- BP and Jacksons Food Stores joined Axonet's c-store network
The Social Commerce Connection
Ulta Beauty's retail network ran multi-brand campaigns on TikTok during holiday 2025. Results: 40% better ROAS versus benchmarks. Top creative delivered nearly 3x the ROAS of standard campaigns.
By 2028, global retail media spending will reach $175 billion, with the United States accounting for nearly 60% of growth.
The Data Consolidation Challenge
Marketing teams now run campaigns across 4-10 different retail media networks. Each requires separate data integration.
Without consolidated reporting, comparing Amazon to Walmart to TikTok Shop becomes guesswork.
Standardize campaign naming across platforms now. Different naming conventions make cross-platform analysis impossible regardless of tools.
Related: How to consolidate marketing data from multiple platforms
Why These Three Trends Connect
Agent-native platforms need clean, consolidated data. Composable CDPs provide warehouse-native foundations. Retail media networks multiply execution destinations.
The marketing stack is splitting into two layers:
Intelligence layer: Where data lives and decisions happen (data warehouse + composable CDP + AI agents)
Execution layer: Where campaigns run (agent-native platforms + retail media networks + social platforms)
Teams building the intelligence layer in 2026 will control costs as execution becomes commoditized. Teams that skip it will pay for manual campaign management that AI could automate.
What to Do in Q1 2026
For multi-platform campaigns:
- Audit your data warehouse situation
- Standardize campaign naming today
- Test one composable CDP for a single use case
For AI agents:
- Identify one repetitive task (budget adjustments, reporting)
- Research MCP-compatible platforms
- Run a 30-day pilot with measurable time savings
For retail media:
- Map which networks your audience uses
- Test closed-loop measurement with one network
- Build cross-platform dashboards before adding more
FAQ
What is an agent-native platform?
Agent-native platforms are advertising infrastructure designed for AI systems to operate through APIs rather than user interfaces. AI agents send instructions via Model Context Protocol (MCP), and the platform executes campaign setup, optimization, and reporting automatically. Infillion Agent Connector, launched December 16, 2025, is the first example built for programmatic advertising. Unlike traditional platforms with AI add-ons, agent-native platforms eliminate manual workflows entirely.
How do composable CDPs differ from traditional CDPs?
Traditional CDPs copy customer data into new systems, creating duplicate storage and vendor lock-in. Composable CDPs connect directly to existing data warehouses without moving data. This cuts implementation time from 6-12 months to weeks and eliminates duplicate storage costs. The trade-off: you need a cloud data warehouse (Snowflake, BigQuery, Databricks) first. Composable vendors grew nearly 4x faster than traditional CDPs in 2025.
Why is retail media growing faster than other ad channels?
Retail media solves three problems simultaneously: advertisers access first-party purchase data for better targeting, retailers generate high-margin revenue from existing traffic, and brands measure actual sales not just clicks. The channel expanded in 2025 to gas stations, convenience stores, and streaming TV. By 2028, global retail media spending will reach $175 billion. Walmart Connect grew 33% last quarter, six times faster than Walmart's overall sales.
What is Model Context Protocol (MCP)?
Model Context Protocol is an open-source standard created by Anthropic that allows AI systems to connect with external data sources through a universal interface. Instead of building custom integrations for every AI model and platform, MCP provides one connection that works across ChatGPT, Claude, and other systems. For marketing, this means AI agents can access campaign data, execute changes, and pull performance metrics without manual API development.
How do AI agents control advertising campaigns?
AI agents receive high-level objectives through natural language and autonomously execute detailed work: audience selection, bid adjustments, creative rotation, budget pacing, and optimization. They operate through protocols like MCP that connect to advertising platforms' APIs. Unlike rigid automation rules, agents adapt based on real-time results. According to Gartner and WARC, over $100 billion in programmatic spend will flow through AI agents by 2028.
