Our Future (and Past) Predictions for AI in Marketing

By Dan Sanchez · Episode published

Score past AI predictions honestly, then plan for what is next: vibe coding, safer cross-application agents, useful systems, and the human judgment that keeps marketers valuable.

Get the next practical AI marketing episode wherever you listen.

Apple PodcastsSpotifyYouTube

Predictions are useful when they force you to place a bet before the outcome is obvious. They are even more useful when you return a year later and score them honestly.

Looking at the last few years of AI in marketing, the pattern is clear: the tools keep becoming more capable, but the biggest advantage comes from learning how to use them before the rest of the market turns a new behavior into standard practice.

Vibe coding is becoming a marketing skill

The most important near-term shift is that marketers can increasingly build the tools they need. Code agents can create a landing page, a microsite, a scraper, or a small internal application without requiring the marketer to become a traditional developer.

That does not mean every output is perfect. I still need enough HTML and CSS knowledge to recognize when a tool has taken the wrong turn. But the barrier has dropped far enough that I can describe what I want, review the result, and keep moving.

Start with a landing page. Then try a small calculator or an ungated tool for your audience. As you get more comfortable, you can build a fuller website or connect several steps into an application. The useful skill is not memorizing syntax. It is knowing what the system should do, giving the agent clear constraints, and recognizing when the result works.

Agents need to cross applications safely

Browser agents are another sign of where work is heading. An agent can already move information between open applications, but asking it to browse arbitrary websites creates a security problem called prompt injection. A malicious page can hide instructions that try to hijack the agent and make it reveal credentials or take an action you did not approve.

OpenAI is hardening Atlas against that problem, but speed and reliability still matter too. For cross-application work to become a normal delegation layer, agents have to move through several tools quickly, understand what they see, and resist instructions that come from untrusted pages.

The direction is still important. A marketer could eventually give an agent a project and have it check a task list, update a document, make a website change, and report back. That is different from adding a “rewrite this sentence” button to a SaaS product. It is a workflow that crosses the tools where work actually happens.

Preserve the human edge

The more execution AI handles, the more valuable the human skills become. Storytelling, relationships, taste, personal values, and lived experience are difficult to copy because a model has not actually lived a human life.

That is why AI-assisted work should remain anchored in your own ideas. A book created from your conversations, a podcast built around your questions, or a brand shaped by your experience can use several AI tools without becoming generic. The tools can organize, draft, edit, and repurpose. You still supply the reason the work matters.

Discernment becomes especially important. AI can produce something that sounds plausible, but you have to decide whether it is good, whether it is true, and how it will make someone feel. The ability to recognize great work is a competitive advantage when everyone can generate a competent first draft.

Score your predictions and adjust

The point of a prediction is not to sound certain. It is to create a testable expectation. Some forecasts from the prior year held up: AI agents mostly arrived as smarter automation, marketers crossed the adoption chasm, and AI video became useful enough for real production. Other predictions were early or wrong, including hyper-personalized marketing experiences and full automation for small businesses.

That scorecard changes how you plan. Keep the predictions that matched behavior. Move the early ones farther out. Drop the ideas that never developed evidence. Then make the next bet based on what people are actually doing, not on the loudest product announcement.

AI is moving quickly, but marketers still have time to build an advantage. Learn to make small tools, design repeatable systems, protect your work across applications, and develop the human judgment that makes generated output worth using. Those skills will matter regardless of which model leads the next leaderboard.