Everyone in this industry is either overclaiming AI or dismissing it. We’re going to do neither.
At Junction 37, we’ve been deliberate about where AI earns a seat in our workflow, and equally deliberate about where it doesn’t. After a year of real-world testing across CPG and DTC media accounts, we have a clear-eyed view of what’s working, what’s not, and what the trust gap looks like from inside a performance agency.
Where AI Has Genuinely Helped Us
The biggest win has been time, not intelligence.
AI has cut the time our team spends on reporting synthesis by roughly 40%. First-pass analysis of campaign performance data, pattern flagging across channels, and initial draft summaries of weekly pacing reports are tasks that used to take hours. Now they take minutes.
That’s meaningful. It means our strategists spend less time assembling data and more time interpreting it. A senior media buyer at Junction 37 is expensive not because of what they can type, but because of what and how they think. AI gave us more thinking time. That’s the win.
We’ve also found real value in using AI for scenario modeling during media planning. Feeding historical CPM benchmarks, audience overlap data, and flight timing into an AI-assisted model to stress-test budget allocations has made our planning conversations sharper. This isn’t because AI tells us what to do (it doesn’t), but because it forces faster iteration on assumptions.
Where the Trust Gap Is Real, and Earned
Here’s the honest part.
We’ve caught AI-generated outputs that were confidently wrong, and not wrong in an obvious way. Instead, they were wrong in a way that would sail past someone who didn’t already know the answer:
A pacing summary that misread a mid-flight budget adjustment. A competitive analysis that hallucinated a media cost benchmark. A channel recommendation that ignored a frequency cap we’d clearly set.
Each of those errors required a human to catch it. Every time.
This isn’t a reason to panic about AI, but it is a reason to be clear-eyed. AI outputs in performance media require human verification, full stop. Anyone building a workflow where AI is the final checkpoint, instead of a human, is going to learn that lesson the hard way, and probably in a client meeting.
The trust gap isn’t irrational. The stakes are real. Everything needs to be verified.
The Real Problem Isn’t AI. It’s How Agencies Are Using It.
Here’s our actual concern.
The broken agency model (built on junior staff, thin oversight, and bloated margins) doesn’t get fixed by AI. It gets hidden by it. If a holding company swaps out three junior analysts for one AI tool and pockets the difference without improving the quality of strategic thinking, the client loses. Again.
AI is a capacity tool, not an expertise tool. It doesn’t replace judgment about why a Meta campaign is underdelivering against a specific audience segment in Q4. It doesn’t know your brand’s media history, your retailer dynamics, or what your CMO is worried about this quarter.
What AI can do is give experienced people more room to think. That’s the use case. That’s the benefit.
What Performance Marketers Should Actually Do
Stop asking whether to adopt AI. Start asking where your team’s time goes and whether AI can give any of it back.
Audit your workflow for the tasks that are high-volume, low-judgment, and repeatable. These tasks might be reporting pulls, data formatting, first-draft summaries, or competitive benchmarking research. Those are your entry points.
Then build a verification layer. Every AI output in your workflow should have a named human responsible for reviewing it before it reaches a client or informs a decision. That’s not as a formality. It’s a real quality check.
And be honest with your clients about it. We are. Our clients know we use AI tools in certain parts of our workflow. They also know that every insight, recommendation, and strategic call comes from our team, not a model.
That transparency isn’t a liability. It’s a competitive advantage.
The Bottom Line
AI is a useful tool, but it’s not trustworthy on its own. In performance media, where budget decisions have direct revenue consequences, that distinction matters more than it does in almost any other marketing function.
We’re not anti-AI. We’re anti-hype, and we’re pro-accountability. That means being honest about what AI can and can’t do in a media agency context.
The agencies that figure out that balance will be sharper. The ones that don’t will be faster at being wrong.
Chris Pyne, Founder, Junction 37 – 30+ Years in Performance Media