The smartest question in performance media right now isn’t “how much AI should we use?” It’s “where exactly should AI stop and humans start?” The brands getting this right are pulling ahead. The ones handing everything to automation, or nothing, are leaving real money and real meaning on the table.
AI media buying is not the future. It’s the present. And if you’re a CPG or DTC brand still debating whether to use it, you’re already behind.
The Job Split Nobody Is Talking About Clearly Enough
There’s a reason smart operators trust AI with bidding and pull humans back in for brand storytelling. These are two fundamentally different jobs and confusing them is expensive.
Bidding is a math problem. It involves signals, probabilities, speed, and scale. A human cannot process 50,000 auction decisions per second. An AI can. That’s not a threat to your team. That’s just physics.
Brand storytelling is a human problem. It requires cultural context, earned intuition, and the ability to make a judgment call about what your brand stands for when the data doesn’t give you a clean answer. No model trained on historical performance can tell you what your brand should mean to someone who’s never heard of you.
Blurring these two jobs doesn’t make you more innovative. It makes you less effective at both.
What “AI-Augmented” Actually Looks Like in Practice
Here’s how this split should work inside a performance media program — not theoretically, but operationally:
AI owns:
- Real-time bid adjustments across channels
- Budget pacing and reallocation based on performance signals
- Audience segmentation and lookalike modeling
- A/B signal processing at scale (which variant is winning, not why)
Humans own:
- Creative strategy and messaging hierarchy
- Brand voice decisions and campaign framing
- Interpretation of anomalies (“why did ROAS spike on Tuesday?”)
- Channel strategy and the relationships between them
The handoff zone — where both work together:
- Creative testing frameworks (humans design the test, AI reads the results)
- Audience insights (AI surfaces patterns, humans decide what to do with them)
- Performance reporting (AI aggregates, humans contextualize)
This isn’t a philosophical exercise. It’s an operating model. And most brands, and most agencies, don’t have it documented anywhere.
Why Most Agencies Get This Wrong
The holding company model defaults to two failure modes. Either they over-automate because it’s cheaper to run (fewer bodies, more margin), or they under-automate because their senior people built careers before algorithmic buying existed and don’t trust what they can’t manually control.
Neither is a client-first position. Both are internally motivated.
At Junction 37, we build media programs where the automation does the repetitive, high-frequency work, and our team does the thinking that requires actual judgment. That’s not a marketing line. It’s how we structure every engagement.
The result is that AI improves speed and efficiency while our strategists stay focused on the decisions that actually shape brand outcomes. According to McKinsey’s research on AI in marketing, companies that define clear human-AI collaboration models see 15-20% better marketing ROI than those running either pure automation or fully manual approaches.
That gap is real. And it’s growing.
The Architecture Question Every Brand Should Be Asking Right Now
If you’re a CPG or DTC brand scaling paid media in 2025, your agency should be able to answer this question with specificity: Which decisions in our media program does AI make autonomously, which does it inform, and which stay fully human?
If they can’t answer that, or if the answer is vague, you don’t have a media strategy. You have a media vendor.
Your brand strategy deserves better architecture than that.
FAQ: AI Media Buying for CPG and DTC Brands
What is AI media buying?
AI media buying refers to using machine learning algorithms to automate real-time decisions in paid media, including bid management, audience targeting, and budget pacing, based on live performance signals rather than manual human inputs.
Should AI handle creative decisions in performance media?
No. AI can process which creative variant performs better, but it cannot determine why, or make brand-level judgment calls about voice, framing, or cultural relevance. Creative strategy should remain a human-led function.
How do I know if my agency is over-automating my media program?
Ask them to walk you through which decisions are made by AI and which are made by their team. If they can’t give you a clear breakdown, or if the answer is “the platform handles most of it”, that’s a red flag.
What’s the right starting point for building a human-AI media model?
Start by auditing your current media workflow and categorizing every decision type: high-frequency and data-driven (hand to AI), strategic and brand-level (keep human), and interpretive (build a collaboration process). From there, you can design the handoffs intentionally instead of by default.
Ready to build a media program that uses AI where it wins and humans where it matters? Talk to Junction 37 about performance media built for CPG and DTC brands. →
Chris Pyne, Founder, Junction 37 – 30+ Years in Performance Media