AI Agents are too good at marketing now
Greg Isenberg · 2026-11-09
💡 Quick Take
1. According to the host and guest, marketing has transformed from manual labor into software, code, and tokens run by AI agents.
2. To prevent agents from breaking ad platforms, the speaker advises setting up a Clickhouse data warehouse paired with Airbyte data pipelines for safe querying.
3. The speaker outlines a streamlined workflow using an image generator for static ads and Seed Dance for generating low-resolution avatar videos at a cost of about five dollars.
4. According to the host, users can scale creative output by generating cohorts of thirty or more ads simultaneously to let ad algorithms find converting customers.
5. The speaker notes that platforms like Tribe or Side Shift allow agents to coordinate human creators on a per-video basis.
📊 Detailed Explanation
According to the speaker, Cody Schneider, the traditional marketing model of hiring copywriters, graphic designers, and video editors is rapidly becoming obsolete. Instead of paying human teams for every asset, marketers can now act as conductors of AI agents like Claude Code or Grockbot. To do this successfully, builders must construct a specialized digital environment or petri dish that provides the agent with clean data, direct API access tools, and a reliable execution workflow.
To solve the data access problem without triggering rate limits or hallucination errors, the speaker recommends using Airbyte as a data pipeline paired with Clickhouse as a centralized data warehouse. Instead of querying ad platforms directly with high-frequency requests that can get accounts banned for violating terms of service, the agent queries the local data warehouse. Meanwhile, separate API tokens handle one-way write actions like publishing or pausing ads, keeping platform interactions safe and controlled.
When it comes to generating visual content, the speaker demonstrates how coding agents can execute specific skill files to spin up static ad concepts and AI avatar videos. Using tools for voice cloning and video rendering, agents can generate a base video at a lower resolution and then upscale it using ffmpeg and add motion graphics. The speaker notes that these AI-generated ads often cost around five dollars to produce, allowing brands to massively scale their creative velocity.
Beyond fully autonomous AI avatars, the speaker explains how agents can function as automated creator managers. By integrating with human creator networks via platforms like Tribe or Sideshift, an agent can distribute briefs, manage scripts, and coordinate human-led video production on a per-video payout structure. This hybrid approach lets teams combine automated scale with human creative touchpoints.
🎯 Tech Expert Opinion
The core thesis of this video is that marketing is transitioning from a human-heavy labor industry into an automated, software-driven engineering discipline. According to the host and guest, the winners in this landscape will be those who master agentic workflows and creative velocity rather than relying on traditional agency retainers.
However, implementing these automated marketing stacks introduces significant infrastructure complexity and risk. Relying on raw LLM outputs without proper guardrails can lead to hallucinated data queries, compliance violations in regulated sectors, or poorly rendered visual artifacts. The speaker's emphasis on building recursive vision loops and isolated data warehouses is a crucial technical mitigation against these failure modes.
Ultimately, this workflow is best suited for tech-savvy founders, growth engineers, and lean startups looking to test ad angles and scale creative output rapidly without massive initial headcount. Traditional brand marketers or those unwilling to troubleshoot API integrations and prompt engineering loops may find the technical overhead challenging until these platforms achieve out-of-the-box stability.
Kanal: Greg Isenberg