Marketing teams are racing to scale with AI. Yet the work is getting worse. After watching Marketing Against the Grain with Kipp Bodnar and Kieran Flanagan, I’m convinced the problem isn’t the tech. It’s the teams. My view: AI doesn’t ruin marketing—lazy AI ruins marketing. The fix is roles, process, and ownership, not more prompts and more bots.
The Messy Middle of AI Marketing
Kieran called it “the messy middle,” and he’s right. Everyone can make stuff with AI, but few make anything that matters. We’ve gone from one-off prompts to fleets of agents running unsupervised. The result is noise, sameness, and brands that blend into the feed. I see it across crypto, SaaS, and social. Volume up, memory down.
“It’s more agents… but the problem is nobody owns what comes out the other end.”
Ownership is the missing role in most teams. Leaders obsess over which jobs AI might kill. That’s the wrong debate. The better question is which jobs must exist so AI actually works.
The Stakes Just Got Higher
Search is shifting to answers. That changes how buyers find products. If your brand is missing from AI answers, you are invisible. The show highlighted a surge in traffic from answer engines like ChatGPT and Perplexity, up hundreds of percent in months. Gartner projects 40% of enterprise apps will ship with task-specific AI agents by 2026. Meanwhile, 88% of marketers use AI, but only a third scale past tinkering. That gap is people and process, not tools.
“Your buyer is asking AI, ‘What product should I buy?’ And your company is not even part of the answer.”
As someone who has built online businesses for decades, I’ve seen this movie. Distribution shifts first. Brands that retool early win while others complain.
The Five Roles Worth Hiring Now
Kieran laid out five roles every team needs. I would hire these today, even in lean times. Here’s why—and how I’d run them.
- Prompt Strategist: Own a shared prompt library, train the team, and police consistency. Structured prompts cut hallucinations by 40% and improve brand alignment by 60%.
- Agent Ops Manager: Monitor, audit, and tune agent fleets. Think DevOps for marketing agents. Without this, failure states multiply and no one notices until it’s costly.
- AI Specialist for Answer Engines: Optimize to appear in AI answers. With AI-originated searches surging and most brands absent, this is open field.
- AI Content Strategist: Set style, build platform playbooks, and edit for soul. Use AI for 70% of the draft. Spend human time on the final 30% that makes it stick.
- AI Creative Director: Direct models for video, copy, and images. Slow the team when needed. Keep quality high and voice distinct.
This is not busywork. It’s how you stop brand drift and keep agents from going off the rails.
Proof From the Field
Unilever shows what happens when someone actually owns the system. Their brand DNA repository limits models to approved voice, values, and visuals. Their beauty studio ships faster and lifts completion and click-through at the same time. That’s process plus oversight, not “let the bot run.”
Contrast that with campaigns that look rushed and hollow. Speed without taste is spam. Even Coca-Cola has wrestled with this balance. As Kieran noted, the job is knowing when output is “technically correct but emotionally wrong.”
“Just because you can create more content with AI doesn’t mean you should.”
Where I’d Start This Quarter
I’d pick one product line and build a minimal version of this team. Then I’d hold weekly reviews with hard rules.
- Centralize prompts. No ad-hoc prompts in the wild.
- Instrument agents. Track errors, drift, and cost per task.
- Publish for answers. Target top five buyer questions and measure AI answer share.
- Edit for voice. Human pass on every asset that touches a customer.
- Close the loop. Feed human edits back into prompts and agents.
These steps force discipline. They also build leverage that compounds each week.
Counterpoint—and Why It Falls Short
Some argue that AI will get so good we can skip new roles. That view ignores brand risk and the speed of change. Even strong models need rules, training data, and guardrails. No model can protect your voice without human judgment. The teams winning now pair agents with clear ownership and craft.
The Job Isn’t to Type Faster
I agree with the closing thought from the show: the winners are the people who can manage AI, not the ones who fear it. Andrew Yang’s line still rings true: workers who use AI will replace those who don’t. I’d add one more: workers who design the system will replace those who just prompt it.
Here’s my challenge to you: stop shipping forgettable content. Stand up these roles, measure them, and demand distinct work. Your buyers are already asking AI for answers. Make sure your brand shows up with something worth remembering.
