Steven Golus delivered a comprehensive presentation on artificial intelligence in advertising, addressing the opportunities and transformations reshaping the digital media landscape. The session featured live demonstrations from two leading agentic AI companies, SeedTag and Scope 3, highlighting how AI is fundamentally changing how media is bought and sold in 2026.

Golus began by establishing a critical perspective: while AI should be embraced as a transformative tool, professionals must understand its implications. He emphasized that over his 25-year career in advertising, two major technological shifts have reshaped the industry—the shift from traditional to internet advertising (1996) and the rise of programmatic advertising (2004-2005). Today, artificial intelligence represents a similar sea change, comparable in magnitude to these previous transitions. However, unlike programmatic advertising which took years to mature, AI is advancing at an accelerated pace, requiring immediate upskilling from industry professionals.

The speaker stressed that those who don't understand AI risk falling behind exponentially. He introduced the concept of an "AI curve" where individuals lacking AI knowledge stagnate, while those who master these tools accelerate their capabilities and effectiveness dramatically.

Golus outlined four distinct categories of artificial intelligence, each with different implications for the advertising industry:

Machine Learning represents pattern recognition using unstructured data and feedback loops. Using a candy factory analogy, Golus explained how a robot learns to sort different candies correctly through repeated iterations and feedback. Machine learning has been prevalent in advertising for years through applications like Amazon's "customers also bought" recommendations and Google Ads' Performance Max (PMAX) campaigns, which optimize across multiple channels simultaneously. This technology, while powerful, is nothing new and should not inspire fear.

Deep Learning extends machine learning by processing unstructured data such as images, sounds, and textures through neural networks. Examples include SeedTag's neurocontextual advertising, which analyzes the sentiment and emotional context of content before inserting ads. Disney's "Magic Words" technology represents another deep learning application, connecting advertisements to specific moods or moments within shows. These applications use complex pattern recognition to understand not just what content is about, but the emotional context surrounding it.

Generative AI creates new content—images, videos, text, and audio—based on training data. Tools like ChatGPT, Claude, and Gemini represent Large Language Models (LLMs) that generate human-like responses. In advertising, generative AI creates copy, creative assets, and entire campaigns automatically. Google's Pompeii tool and Paramount's Ads Manager demonstrate how businesses can generate campaign creative without traditional production resources. While impressive, generative AI threatens creative professionals, particularly video producers, directors, and graphic designers. However, it democratizes creative production for small businesses.

Agentic AI is the future—systems that take independent action toward goals without human intervention. This represents the most significant disruption. Rather than users manually requesting actions, agentic systems autonomously discover inventory, negotiate deals, and execute campaigns. This is the true game-changer redefining how media is bought and sold.

Large Language Models are systems trained on vast datasets to generate human-like responses. When asked which LLM platform attendees preferred, the audience overwhelming selected Claude, reflecting its reputation for superior capabilities and integration options compared to competitors.

LLMs train on publicly available web data, books, Wikipedia, and subscription-based content (with licensing agreements, such as OpenAI's partnership with the New York Times). When used within corporate environments, LLMs can also train on company-specific data, creating personalized AI systems.

An important concept Golus introduced is "tokens"—the fundamental units measuring LLM computational effort. Users have token limits (similar to early cellular data caps), and consuming more tokens requires additional payment. As professionals increasingly integrate AI into daily workflows, they'll likely encounter token limitations, necessitating upgraded subscription tiers.

The presentation included concrete examples demonstrating AI's current impact. Amazon uses machine learning to recommend books based on what similar readers purchased. Google Ads' Performance Max campaigns employ machine learning to optimize ad placement across search, YouTube, and display networks simultaneously.

For deep learning, Golus highlighted SeedTag's neurocontextual technology that analyzes not just URLs but the emotions, interests, and commercial intent users experience while consuming content. Disney's Magic Words technology enables advertisers to connect with audiences during specific emotional moments—for instance, placing tea advertisements when viewers experience loving moments in shows.

Generative AI examples included Google's Pompeii, which analyzes business websites and generates entire marketing campaigns automatically, and Paramount's Ads Manager, enabling small businesses to create video advertisements without professional production resources.

SeedTag's Chief Product Officer Gregor Martinez demonstrated the Lease Agent, an agentic AI tool that fundamentally changes campaign planning. Rather than manually browsing through thousands of pre-defined audience segments, the system accepts natural language briefs and automatically creates custom audiences from SeedTag's daily processing of over 100 million URLs across 17 countries.

The system analyzes interests, emotional responses, and commercial intent, providing detailed audience insights. Instead of static, predefined segments, marketers now receive dynamic, custom-built audiences matching their specific campaign objectives. Once validated, these audiences activate through programmatic or IO channels across SeedTag's network and partner DSPs.

Anne Coghlan from Scope 3 demonstrated how agentic AI transforms media buying. Their system enables brands to discover inventory globally using natural language prompts within interfaces like Claude or ChatGPT. The platform built on the Ad Context Protocol (AdCP) allows "advertising agents" to speak a common language with publisher-side agents.

In the demonstration, simply stating a campaign objective returned 129 different ad products across 12 publishers. The system provides recommendations based on the campaign brief, showing pricing, floor prices, creative formats, and reporting capabilities. Buyers can then book media at scale while publishers maintain first-party data and creative control—essentially offering direct IO quality at programmatic scale.

Golus emphasized that agentic AI will disrupt traditional roles, particularly those involving routine tasks and manual planning. However, these disruptions create new opportunities for professionals who upskill quickly. He noted that individuals who previously resisted AI have transformed into power users within weeks of structured learning.

The presentation revealed a fundamental truth: AI tools already exist throughout the advertising ecosystem—most professionals simply don't recognize them as AI. Lookalike modeling, moment targeting, and concept targeting all leverage machine learning and deep learning technologies.

Golus concluded with practical advice: embrace AI immediately, learn through daily micro-learnings (ten minutes per day), set up automation for daily updates on relevant topics, and experiment with available tools. He recommended Claude specifically for its integration capabilities and superior performance compared to alternatives.

He announced a four-week "AI Sales Training" course launching in May, specifically designed for sales professionals to master AI tools for closing deals and improving customer interactions.