📊 Key Data
  • 2M+ daily bidding decisions processed by Xnurta's ad tech platform.
  • 15% YoY increase in ad click volume and 10% lift in new-to-brand shoppers for Greenworks using Xnurta.
  • 86.1% data accuracy rate achieved by Xnurta's Insight Agent, reducing AI hallucination risks.
🎯 Expert Consensus

Experts would likely conclude that Xnurta's agentic AI technology marks a significant shift in ad tech, enabling natural language execution of campaigns while emphasizing the need for robust safeguards against AI hallucinations and financial risks.

about 4 hours ago
From Chatbots to Campaign Managers: The Arrival of Agentic Ad Tech

From Chatbots to Campaign Managers: The Arrival of Agentic Ad Tech

BELLEVUE, Wash. – November 19, 2026 – For the past three years, the advertising technology sector has been trapped in an AI hype cycle characterized by conversational parlor tricks. Marketers have grown accustomed to asking chatbots to summarize campaign performance or generate ad copy, but the actual execution—the high-stakes manipulation of bids, budgets, and targeting—has remained firmly behind traditional dashboards and manual workflows.

Today, that barrier officially falls.

Xnurta, a Bellevue-based ad tech platform that processes over two million bidding decisions daily, announced a significant expansion of its Model Context Protocol (MCP) capabilities. The update transforms popular AI assistants like Anthropic’s Claude and OpenAI’s ChatGPT from passive data analysts into active execution engines. Through bidirectional read-and-write integration, Amazon advertisers can now command these LLMs to directly adjust bids, launch campaigns, and manipulate budgets using natural language.

"Connecting AI to advertising data was just the first step," said Kashif Zafar, CEO of Xnurta, in the company's announcement. "The real transformation happens when AI can act on that information. We've evolved Xnurta MCP from an analytical tool into an execution engine that understands an advertiser's strategy, operates within their established guardrails, and turns natural-language instructions into meaningful advertising actions."

This is not merely a feature update; it is a fundamental shift in how retail media is managed. But as AI moves from the analytics dashboard to the financial steering wheel, the industry is about to test whether conversational convenience can safely coexist with the unforgiving realities of million-dollar media budgets.

The Protocol Powering the Shift

To understand the significance of Xnurta’s rollout, one must look at the underlying plumbing. The Model Context Protocol (MCP) was introduced by Anthropic in November 2024 as an open standard to connect AI models to external data sources and tools. Now governed by the Linux Foundation, MCP functions essentially as a universal "USB-C for AI," allowing different systems to communicate without bespoke, brittle integrations.

The advertising industry is standardizing around MCP at a staggering pace. Google, Meta, and Amazon Ads have all rolled out their own MCP servers over the past year. However, much of this early adoption focused on data retrieval. Xnurta’s implementation bridges the gap to execution.

By exposing its campaign management infrastructure as standardized tools through an MCP server, Xnurta allows an advertiser to type a prompt like, "Return all Prime Day event budgets back to their pre-event levels," and watch the system execute the change across hundreds of campaigns. The platform supports Amazon Sponsored Products, Sponsored Brands, and Sponsored Display campaigns, drawing on 15 months of historical data with hourly reporting granularity. Read-only access for Amazon DSP is slated for December, signaling a broader push into full-funnel retail media management.

Execution Over Optimization: The New Competitive Moat

For years, retail media has been dominated by machine learning optimization. Platforms like Pacvue, Perpetua, and Quartile have built massive businesses by automating bid adjustments based on target ACoS (Advertising Cost of Sales) or ROAS (Return on Ad Spend).

What Xnurta is attempting is fundamentally different. While traditional platforms use AI in the background to optimize towards a set goal, Xnurta is bringing agentic AI to the forefront as the primary operational interface. It replaces the complex UI of nested menus and bulk upload sheets with a conversational command line.

This approach is heavily reliant on "AI Skills"—pre-built, customizable workflows designed for specific, repeatable tasks. Rather than asking an LLM to figure out the intricacies of Amazon's campaign structure from scratch, these Skills provide the agent with a rigid operational framework.

The results of this domain-specific grounding are quantifiable. In a recent deployment, outdoor equipment manufacturer Greenworks utilized Xnurta’s ecosystem to execute a full-funnel Amazon Ads strategy. By leveraging the platform to unlock new customer segments without increasing their overall budget, Greenworks achieved a 15% year-over-year increase in ad click volume, a 10% lift in new-to-brand shoppers, and a 10% increase in ROAS.

The Hallucination Hazard in High-Stakes Bidding

Despite the impressive case studies, handing the keys of a live financial system over to a large language model is inherently risky. This is where the rubber meets the road for agentic AI, and where the most significant hidden challenges lie.

The primary danger is AI hallucination. Foundation models are designed to predict the next most likely token, not to execute flawless financial math. Internal benchmarking reveals the severity of this issue: when tasked with Amazon ad workflows, the strongest general foundation models hallucinate or miscalculate data in roughly two-thirds of their responses. Xnurta combats this by utilizing its own domain-native Insight Agent, which boasts an 86.1% data accuracy rate, serving as a critical translation layer between the general LLM and the Amazon Ads API.

Furthermore, there is the challenge of API latency and rate limiting. During high-volume shopping periods like Black Friday or Prime Day, the Amazon Ads API dynamically throttles requests to manage system load, returning 429 error codes when limits are exceeded. An autonomous AI agent attempting to execute bulk adjustments across thousands of keywords could easily trigger these limits, resulting in failed updates and stranded budgets. Agentic systems must be engineered with sophisticated exponential backoff strategies and error-handling protocols that standard chatbots simply do not possess.

Guardrails and the Human in the Loop

Recognizing that a single hallucinated zero could drain a monthly ad budget in hours, Xnurta has implemented mandatory friction into its execution engine.

The platform operates on a strict "human-in-the-loop" (HITL) architecture for high-stakes actions. While an AI assistant can analyze data, propose optimizations, and queue up bulk adjustments to bids, budgets, and targeting, it cannot unilaterally push those changes live. All critical financial and structural modifications—specifically budget updates, bid adjustments, and campaign archiving—require a human preview and explicit confirmation before being executed via the API.

This is managed through configurable "action spaces," which act as persistent policy boundaries. E-commerce directors can dictate exactly which campaigns an agent is allowed to touch, and what specific parameters it is allowed to adjust. It is a necessary compromise between the promise of autonomous AI and the reality of corporate governance.

As the industry moves toward 2027, the transition from AI as an advisor to AI as an executor is inevitable. The platforms that succeed will not be those with the most conversational chatbots, but those that can build the most robust infrastructure for catching mistakes before they hit the balance sheet. Xnurta’s read-and-write MCP integration is a major milestone in this journey, proving that natural language can indeed run complex retail media campaigns, provided there is a human standing by to hit the final button.

Topics & Related

Event:
Product Launch
Theme:
Agentic AI
Sector:
Advertising & Marketing
Product:
ChatGPT
Claude

📝 This article is still being updated

Are you a relevant expert who could contribute your opinion or insights to this article? We'd love to hear from you. We will give you full credit for your contribution.

Contribute Your Expertise →
UAID: 55545