I've Mocked AI Agents for 2 Years. January 2026 Changed My Mind.

By Dan Sanchez · Episode published

AI agents are finally becoming useful inside everyday marketing tools. Learn how to connect work, choose models by task, and move from isolated tasks to repeatable systems.

Get the next practical AI marketing episode wherever you listen.

Apple PodcastsSpotifyYouTube

I have been hearing about AI agents since the end of 2024. For a long time, the promise was ahead of the usefulness. That changed in January 2026. Agents are beginning to show up inside the tools marketers already use, and they can now take enough steps on their own to matter.

The shift is not one magical product. It is a change in how work gets done.

Agents are starting to connect the steps

Anthropic’s Claude Cowork is an early example. The Mac app can access the terminal, which means it can inspect files, create folders, move documents, research a question, and write files for you. It can turn a conversation into a document, then return to that document in a later conversation when the original thread is too long.

The first demonstrations are simple, such as cleaning up a desktop or creating standard operating procedures. That is fine. New technology often starts with an obvious use case before people discover the work it is actually good at. I tried asking Cowork to download and install WordPress and MAMP, and it failed. The promise is real, but the reliability is still uneven.

The important idea is that an agent can connect several actions that used to require copying, pasting, uploading, and organizing by hand. A marketer might ask it to research a topic, create a brief, save the brief in a shared folder, and update the working document. That is the beginning of an agentic workflow.

Choose tools by the work, not the brand

ChatGPT Codex and Claude Code illustrate a useful difference. Claude Code often works side by side with someone who understands development. Codex is designed to take a larger assignment and keep moving without checking in as often. For a marketer who does not know exactly how the code should be structured, that difference matters.

The same pattern shows up in image work. One model may be more precise with a small edit, while another is better at improvising a complete design. I now use Claude, Gemini, and ChatGPT consistently because their strengths are different. The winning habit is to match the tool to the job instead of turning model loyalty into a workflow constraint.

Google’s Gemini integration in Drive is another small but meaningful step. Drive search is often frustrating when you did not create the folder structure. Gemini can read across files, find the document by meaning, create a new file, or move an existing one after asking for confirmation. These small maneuvers are where agent behavior becomes useful in everyday marketing operations.

Move from tasks to systems

The bigger change is not that an agent can finish a task. It is that an agent can execute a repeatable process. Marketers who only collect tasks will become the bottleneck when agents can do more of the work. Marketers who design systems will have something valuable for agents to operate inside.

Turn a recurring task into a sequence: what starts it, what information it needs, what decisions it makes, what gets produced, and how you know it worked. Then use an agent to run that sequence and improve it. A landing page, a research report, a content brief, or a small audience tool can all become practice grounds.

Vibe coding is still early. In a poll of my audience, only 11 percent said they used it every day and 25 percent weekly. That is an opportunity. Marketers can use it to build landing pages, microsites, calculators, and small ungated tools without waiting for a full development project. The point is not to become a software engineer overnight. It is to learn how to describe a useful system clearly enough for an agent to help build it.

AI is becoming more technical and more human at the same time. The tools handle more of the technical execution, which makes judgment, taste, communication, and problem framing more valuable. The marketers who learn to combine those human skills with repeatable systems will be ready for the agentic workflow that is finally arriving.