How to Launch AI-Native Marketing Campaigns

Introduction

Brendan Farnand, co-founder and chief evangelist at Knak, spent the past year talking with more than 100 enterprise marketing teams—including OpenAI, Google, Stripe, and AT&T—to understand where AI actually helps campaign production and where it breaks down. With a career spent as an enterprise marketer who led the global implementation of a marketing automation platform at a $1B+ organization, Brendan brings a practitioner's view of what separates companies experimenting with AI from those actually shipping better marketing. His core argument: the definition of a marketing campaign has fundamentally changed, and most marketing teams haven't restructured their operations to catch up.

The Bottleneck Has Shifted—And So Has the Definition of a Campaign

For years, the constraint on enterprise campaigns was production. Building anything meaningful meant coordinating with 20+ people, stitching together 10 to 12 tools, and cycling through endless agency revisions. Teams settled for their "top five" ideas because throughput was the ceiling. AI has lifted that lid—but the constraint didn't disappear, it moved. As Brendan puts it, "the problem was never we don't have enough ideas. Everyone has so many ideas. But getting those ideas to the audience was always the challenge." Now the challenge is launching campaigns people actually care about, at a cadence that stays always-on, relevant, and specific to the person you're talking to.

Five Factors That Make a Campaign Worth Shipping

Because production is no longer the bottleneck, marketers face a new discipline problem: not every campaign should be launched. Brendan outlined five filters that historically enterprises could only satisfy one or two of at a time. Is it relevant? Is it interesting to your audience? Is it timely? Is it personalized? And does it have enough scale to be worth executing? The saturation point—how many campaigns your audience can absorb—should be determined by engagement, not volume for its own sake. "You have to maybe feed some of your audience a few more pickles than they're used to eating" before you find the right rhythm. Engagement is the signal that tells you whether you've hit the ceiling on quantity or the floor on quality.

Inside OpenAI's Human-in-the-Loop Campaign Workflow

The most concrete example Brendan shared was OpenAI's campaign infrastructure, built to maximize throughput without losing what the team repeatedly calls "taste." The process starts in Slack, where marketers already work. A marketer writes an unstructured message describing a campaign—the new product, the target audience, the timeline. An AI agent responds conversationally, pulling out the missing details and populating a structured brief into a project management tool called Linear. A second agent then reads that brief and calls other platforms—including Knak's MCP—to generate production-ready campaign assets like emails and landing pages. Finally, the marketer is dropped back into the platform to review, tweak, and add the human judgment that AI can't provide.

Why the Human Element Is Non-Negotiable

The OpenAI team, Brendan noted, is explicit about refusing to push "AI slop" to their audience. They deliberately insert marketers at specific checkpoints to protect the creative touch AI lacks. This reflects a broader shift he's seeing: the modern marketer is becoming "full stack"—able to move from idea to execution without handing off to developers or technical specialists. Where marketers were historically "T-shaped" (broad knowledge, one deep specialty), the right infrastructure now lets every marketer own the entire production process. That, in Brendan's view, is where the remaining friction lives in most organizations—legacy tools and processes built around the limitations of older technology.

Think Like a Factory: Find the One Constraint

For marketing leaders wondering where to start, Brendan pointed to Goldratt's Theory of Constraints from manufacturing: every production line has exactly one constraint, and solving for it unlocks throughput. The lesson for marketers is to audit the mechanics of how ideas travel from concept to audience, identify the single bottleneck, and fix that first. Throughput, importantly, isn't about producing more—it's about getting the right campaigns all the way into customers' hands. YouTube illustrates the payoff: when promoting Super Bowl coverage, minutes matter. Getting timely, relevant content live within hours of a preceding game ending means "literally millions more people watching on their platform."

Conclusion

The central mindset shift Brendan advocates is deceptively simple: stop asking "how do I produce this campaign?" and start asking "how do I produce campaigns?" Campaigns are no longer one-off, flashy events—they're an always-on, repeatable production line built for speed, relevance, and personalization. As AI floods every cha el with more content, the teams that stand out will be the ones that are timely, personalized, and grounded in what's happening in the real world right now. That requires treating campaign creation like a manufacturing process, protecting human judgment at deliberate checkpoints, and building AI directly into daily workflows rather than using it as an occasional answer engine. The infrastructure exists—the differentiator is whether your team is organized to use it.