AI agents that actually run campaigns: Yahoo DSP embeds agentic AI
On January 6, 2026 at CES, Yahoo DSP switched on agentic AI that doesn't just advise-it can run parts of your programmatic workflow with your approval. Agents now watch pacing, diagnose delivery issues, recommend fixes, and execute approved changes across planning, activation, optimization, and measurement.
As Adam Roodman put it, "Agentic AI changes how media buying actually gets done." The big shift: less clicking through menus, more describing goals in plain language and letting agents handle the busywork.
What launched
Three live capabilities shipped, all under a flexible "Yours, Mine, Ours" setup so you can use Yahoo's agents, plug in your own, or run both through shared protocols.
- Campaign activation (Yours): Connect external agents to Yahoo DSP via Model Context Protocol (MCP). Newton and RPA built a trafficking agent here that has already executed programmatic guaranteed buys.
- Troubleshooting (Mine): Native, always-on agents detect underpacing and delivery issues, identify root causes across targeting, supply, and settings, then recommend or execute approved fixes.
- Audience exploration (Ours): Partners tap Yahoo audience metadata via APIs or MCP to analyze segment size, demo mix, and pricing, then auto-recommend segments aligned to your goals.
Why marketers should care
This is time arbitrage. Tasks that used to eat hours-pacing diagnosis, audience research, line-item tweaks-compress to seconds. John Goulding at MiQ says built-in agents turn under-delivery checks and audience building from hours to seconds, and interoperability with MiQ's own agents enables smarter day-to-day decisions.
Lisa Herdman at RPA says applying agents to programmatic workflows removes operational friction so teams can focus on market strategy and partnerships, not ticketing and triage.
Interoperability and standards (the practical bits)
MCP is the "USB-C" layer for AI tools to talk to ads platforms. Google released an open-source MCP server for its Ads API on October 7, 2025. Amazon launched a closed beta MCP server on November 13, 2025. Yahoo's adoption means your internal agents can connect without bespoke glue code, while Yahoo's native agents still run inside the platform.
There's also the Ad Context Protocol from a group of ad tech firms. It sparked debate over whether the industry needs another standard. Meanwhile, the IAB Tech Lab's AI roadmap extends existing specs (OpenRTB, AdCOM, VAST) rather than starting from scratch-useful if you want less fragmentation and fewer one-off integrations. See the IAB Tech Lab for context.
What's live now vs. what's next
Yahoo outlined five agent categories spanning the full lifecycle:
- Insight: Spot trends and anomalies instantly.
- Traffic: Speed up setup with guided workflows.
- Optimize: Adjust strategies as results shift.
- Improve: QA checks that catch issues before they cost you.
- Measure: Turn raw reporting into outcomes you can act on.
The activation, troubleshooting, and audience exploration pieces are live. Optimization, QA, and advanced measurement agents are slated to roll out through 2026.
Governance, control, and trust
Agents run with guardrails. You get audit trails, clear reasoning, and the ability to approve or override changes. That matters for brand safety, budget control, and compliance. As confidence grows, teams can widen the approval envelope and let agents handle more changes automatically.
How to use Yahoo's agents in your next 30 days
- Set the rules: Define which levers agents can touch (bids, budgets, supply, targeting) and when approvals are required.
- Start with pacing: Turn on the troubleshooting agent to monitor under-delivery, then approve suggested fixes for one or two test campaigns.
- Pilot activation: Use the activation agent on a low-risk programmatic guaranteed buy to cut setup hours and reduce manual trafficking steps.
- Run audience discovery sprints: Ask the audience exploration agent for 3-5 segment options per goal, then test fast and prune.
- Create an exception path: Define SLAs for human escalation (brand flags, unusual spikes, conflicting signals).
- Review weekly: Compare agent decisions to business goals. Tighten or loosen controls based on results.
Skills your team will need
- Prompting and intent design: Give agents clear goals, constraints, and success criteria in plain language.
- Agent orchestration: Decide which tasks go to Yahoo's agents vs. your internal stack; use MCP to connect them.
- Data governance: Approvals, auditability, and access boundaries across first-party, retail, and ACR data.
- Outcome-based measurement: Tie agent actions to incrementality, not just CTR or CPA.
If you're building these skills, here's a practical place to start: AI certification for marketing specialists.
Context that affects your media plans
- Retail media arms race: Walmart Connect announced an AI assistant on January 6, 2026 to build, optimize, and troubleshoot via chat.
- Supply-side autonomy: PubMatic launched AgenticOS on January 5, 2026; Magnite added seller agents in SpringServe with Ad Context Protocol support.
- Standards push: IAB Tech Lab (January 6, 2025) outlined an agentic roadmap built on OpenRTB, AdCOM, and VAST.
- Yahoo's data and reach: Netflix added Yahoo DSP as a global programmatic partner on June 16, 2025; ACR audiences from Nexxen arrived October 2025; Yahoo's Conversion API went live in April 2025; ConnectID spans 232M logged-in U.S. users.
Under the hood (why interoperability matters)
Yahoo's "Yours, Mine, Ours" structure is a hedge against a split market: some advertisers will lean on native agents; others will bring internal models and business logic; many will run both. MCP and APIs make that mix workable without custom plumbing for every connection.
This also answers a bigger question raised in 2025: if autonomous systems can plan, target, and optimize, what's a DSP's role? Yahoo's bet is to be the agent host and coordinator-not just middleware to route bids. That keeps the platform relevant as teams bring their own AI to the table.
What to watch next
- Autonomous optimization: Which decisions will move from "recommend" to "auto-approve" as results prove out?
- Cross-platform effects: How do Yahoo's agents interoperate with retail media, CTV, and walled gardens via MCP?
- Measurement quality: Do agent decisions track with incrementality and offline outcomes via Conversion API connections?
- Standards adoption: Will MCP and IAB extensions curb fragmentation, or will multiple agent interfaces persist?
Signal on market impact
McKinsey flagged agentic AI as a top trend in 2025, with $1.1B in investment and a 985% jump in related job postings from 2023 to 2024. Google Cloud projects the market could reach about $1T by 2035-2040, with over 90% of enterprises planning integration within three years. Media teams that move early will bank the time savings and redirect effort to strategy and creative.
Key takeaways for marketers
- Agents are live inside Yahoo DSP today for activation, troubleshooting, and audience exploration-use them to cut operational drag.
- Keep humans in the loop with approvals and audit logs; widen autonomy as trust grows.
- Use MCP to connect your internal agents and keep your proprietary logic intact.
- Shift your team's focus from levers and menus to goals, constraints, and outcomes.
- Prepare for a mixed market: native platform agents plus your own AI, working side by side.
Your membership also unlocks:
