AI Marketing Research: Agency Survival, Churn AI & Disclosure

When AI can write your ads, predict who's about to leave, and draft your strategy decks — what are humans still for, and when does it matter that you say so? This week's research batch lands on a single uncomfortable truth: the closer AI gets to judgment calls, the more it needs a human watching over its shoulder. In this Research Radar Brief, Dr. Eva Wolf reviews 3 recent AI marketing research papers covering the future of advertising agencies, deep learning models for customer churn prediction, and the growing gap between how AI is actually used in professional work and what anyone discloses about it. What you'll learn: - Why AI automating ad campaigns on Meta and Google may be eroding the most billable parts of your agency — and what to offer instead - How a deep learning model achieved 96% accuracy predicting customer churn on e-commerce data — and why that number needs careful context before you act on it - Why customer satisfaction scores appear to be a stronger loyalty predictor than purchase history, and what that means for your CRM setup - How AI disclosure norms are broken even in academic research — and what that gap reveals about transparency risks building up inside marketing teams - What a genuinely useful AI disclosure statement looks like versus a generic line that tells nobody anything Papers covered: 1. A Discussion on the Future of Advertising Agencies in the Impact of Artificial Intelligence - Source type: Peer-reviewed journal article (Intermedia International e-journal) - Access: Full text reviewed - DOI: 10.56133/intermedia.1740804 - Radar verdict: Test this week 2. Generative AI for Personalized Marketing and Customer Experience in E-Commerce - Source type: Peer-reviewed journal article (International Journal of Emerging Research in Engineering and Technology) - Access: Full text reviewed - DOI: 10.63282/3050-922x.ijeret-v7i1p103 - Radar verdict: Use cautiously 3. Expectations and Practices around AI Disclosure in CS Research - Source type: Preprint (arXiv) — not yet peer-reviewed - Access: Full text reviewed - URL: https://arxiv.org/abs/2608.23271v1 - Radar verdict: Test this week Full show notes, transcript, and citations: https://bigplans.media/episodes/ai-marketing-agency-survival-churn-prediction-disclosure-2026-08-25 Disclaimer: This is a first-pass research briefing produced by an AI-generated research avatar trained on Dr. Eva Wolf's methodology. It is not a substitute for reading the original papers. Preprints have not undergone peer review and findings may change. Source quality and study limitations are noted for each paper. Nothing here constitutes financial, legal, or professional advice. -- This is a first-pass research briefing, not a final academic review. Read the original papers before making major marketing or business decisions. AI & Marketing Research Radar is produced by BigPlans Media. Subscribe wherever you listen to podcasts.

Thanks for listening to AI & Marketing Research Radar by Big Plans Media.

I’m Dr. Eva Wolf, and I help marketers, educators, consultants, and business owners turn AI marketing research into practical strategy, smarter workflows, and real business opportunities.

More episodes: https://bigplans.media/ai-marketing-research-radar/
Consulting: https://bigplans.media/ai-marketing-consulting/

Big Plans Media — Where Big Ideas Meet Smart Marketing.

Transcript

Copy Transcript Copied

SPEAKER_00 0:07

You're listening to Avita, an AI-generated research briefing avatar trained on the research framework and methodology of Dr. Eva Wolfe, marketing professor, AI researcher, and founder of Big Plans Media. Every day, Evita scans emerging research in AI, marketing, consumer behavior, psychographics, and business strategy to identify the most relevant developments, opportunities, and risks worth watching. These daily radar reports are designed to help busy professionals stay informed without having to read hundreds of research papers themselves. And every Friday, join Dr. Eva Wolfe live for her personally recorded weekly AI marketing radar roundup, where she breaks down the biggest stories, explains what actually matters, and shares practical insights and strategic implications for marketers, educators, entrepreneurs, and business leaders. Now here's today's radar report. Here's the signal I can't ignore today. Your agency is using AI to write copy. Your team is using AI to clean up reports. And somewhere in all of that, nobody has written down what AI actually did or who's accountable if it's wrong. Today's papers point to the same pattern. The rules around AI and marketing work haven't caught up to the reality of AI in marketing work, and the gap is getting expensive. We screened 342 papers. Three cleared the full text bar and made the radar. Quick caveat: this is a first-pass research briefing, not a final academic review. Every paper today has full text access. I'll tell you what the papers suggest, what they don't prove, and which ones deserve a deeper read. Okay, let's get into it. Paper one. Here's the business question. If AI can now do the stuff agencies used to charge for, what exactly are you selling? A researcher reviewed existing literature and real platform examples, specifically what Meta's Advantage Plus and Google's automated ad tools are doing right now to build an argument about where agencies go from here. No original data. This is a conceptual synthesis. Keep that in your back pocket. So here's what it says AI is already doing the repetitive agency work, writing variations, targeting, media planning, the platforms are eating it. But the paper argues there are things AI consistently gets wrong. Reading cultural nuance, making a gut-call creative decision, catching the ethical landmine before it becomes a headline. That's where the human has to stay. Here's where it gets interesting. The paper doesn't just say agencies will survive. It says the agencies that survive will do it by building new specializations. AI output auditing, brand safety oversight, culturally intelligent creative strategy. Not doing what they've always done just faster. Actually doing something different. That is not a pivot in positioning. That is a rebuild of the service model. But here's the catch. There is no data here. No agency tested this hybrid model. No performance numbers. This is one researcher's argument published in a niche communications journal. I believe the logic. I do not have evidence the model works. That's the part I keep coming back to. The argument is clean. The proof is missing. Plain English payoff. AI is taking the repetitive agency work. So if you want to stay valuable, you have to get very good at the stuff AI gets wrong. Cultural judgment, creative leaps, ethical catches. Okay, here's where this becomes commercially interesting. Money move. Build an AI output audit service. A human-led quality control layer that reviews AI-generated ad creative before it goes live, checking for cultural missteps, brand voice drift, anything that should have flagged a legal review. Brands running automated meta and Google campaigns have nobody doing this right now. That is the gap. Action step. Before your next campaign review, list every deliverable your team produced last quarter. Mark which ones AI could have drafted. That's your automation exposure. Then mark which ones required a cultural call or an ethical judgment. That's your irreplaceable value. Do the audit, no the split. Evidence check. Conceptual review article. No empirical data. No tested model. One author. Use it as a thinking framework, not a validated playbook. Radar verdict. Test this week. The hybrid model it describes is worth road testing inside your own workflow, even if the paper can't prove it works yet. This next one looks like a technical machine learning paper, and it is, but don't tune out because the finding hiding inside it is one every e-commerce marketer should tattoo on their dashboard. Paper 2. Here's the business question. Which signal in your customer data actually tells you who's going to churn before they do? A researcher built a GAN-based machine learning model, a type of deep learning, and ran it against an e-commerce customer dataset from Kaggle, then benchmarked it against the simpler models most companies actually use. Logistic regression, random forest, naive bays. The GAN model correctly identified loyal versus at-risk customers about 96% of the time, be every baseline. And it was especially good at catching behavioral drift over time. A customer who used to buy regularly but has started to slow down. Standard models tend to miss that pattern entirely. But here is the finding I actually care about. The single strongest predictor of customer loyalty in this data set? Not how much they spent, not how often they bought. Customer satisfaction scores. That is not a UX problem. That is a revenue leak. But here's the catch. And this one's real. The dataset is a public Kaggle dataset. Size isn't even reported. There's no live business deployment, no A-B test. One independent researcher, no institutional affiliation. And the journal isn't one I'd hang my hat on. So 96% accuracy on a cleaned Kaggle data set does not mean 96% accuracy in your Shopify store, not even close. Plain English payoff. Your best early warning signal for customer churn might already be sitting in your satisfaction data. You just haven't connected it to your retention triggers yet. Here's the business hiding inside the research. Money move. Offer a satisfaction score audit for e-commerce clients. Dig into their existing CRM and transaction data. Find which satisfaction signals correlate with repeat purchase, and set up automated outreach triggers for customers showing early drop-off. Low implementation costs, high retention upside. Action step. Check your current churn model. Or if you don't have one, your retention email logic. And ask, are we using satisfaction scores as an input? If the answer is no, that's your test. Add one satisfaction signal and measure its predictive lift over 60 days. Evidence check. Kaggle dataset of unknown size. No real-world deployment. Single independent researcher. The accuracy figures are not real-world benchmark. Extract the satisfaction as loyalty insight cautiously and test it in your own data before drawing conclusions. Radar verdict. Stay with me for this one. On the surface, it's about academic research disclosure, but the business implication, especially if you're running an agency or producing AI-assisted client work, is bigger than it sounds. Paper three. Here's the business question. When your team uses AI to build a strategy deck, a research report, or a campaign brief, does your disclosure actually tell the client anything useful? This is a preprint, not yet peer-reviewed. Two researchers analyzed AI disclosure policies across 65 top computer science conferences, surveyed 109 researchers about when they think disclosure actually matters, and ran a computational analysis of nearly 14,000 real AI disclosure statements from major academic conferences. Here's what they found. Most disclosure policies exist, but they're vague. They don't say what to disclose or when. So researchers fill the gap by disclosing the low-stakes stuff, editing sentences, cleaning up code, and not disclosing the high-stakes stuff. Like when AI shaped the actual analysis or the experimental design. And the thing researchers said they care most about, a clear statement that the authors take responsibility for AI-assisted work is missing from 98% of disclosures at one conference and 77% at another. Let me translate that. People are disclosing that AI fixed their grammar. They are not disclosing that AI shaped their conclusions. And almost nobody is saying, we reviewed this, we stand behind it. Right. This study is about academic publishing, not marketing. Real limitation. But the pattern it describes, I see it everywhere in agency work. AI tools were used in the production of this report. Full stop. That tells the client nothing. I bet the number in commercial marketing deliverables is higher. Plain English payoff. AI tools were used is not a disclosure. It's a dodge. A real disclosure says what AI did, at which stage, and who reviewed and takes responsibility for the output. Okay, here's where this becomes commercially interesting. Money Move. Build an AI disclosure generator. A simple tool where agency teams input what AI did at each stage of a project and get a structured, specific disclosure statement to attach to client deliverables. Liability reduction, trust signal, and a genuine differentiator for agencies that want to look like the grown-ups in the room. Action step. Pull the last three AI-assisted deliverables your team sent to a client. Read your disclosure language. Does it say what AI specifically did? Does it confirm a human reviewed and is accountable for the output? If the answer to either is no, rewrite it before the next deliverable goes out. Evidence check. Preprint, not peer reviewed. The domain is academic CS research, not marketing. The leap to commercial content contexts is conceptual, not empirical. The large-scale disclosure analysis is substantive, but apply the framework as a starting point, not a proven standard. Radar verdict. The disclosure framework the paper describes is ready to adapt for internal AI governance right now, without waiting for peer review to confirm what common sense already suggests. At first glance, these papers look separate. Ones about agencies, one's about churn prediction, ones about disclosure norms. But together they show something I can't stop thinking about. We've moved fast on AI adoption, really fast, and we've built almost no infrastructure around it. No quality control layer, no accountability statements, no clear sense of which data signals actually matter versus which ones we're defaulting to because they're easy to measure. Not more AI. Better guardrails around the AI we're already running. The agency paper says, AI can do the repetitive work, but someone has to own the judgment calls. The churn paper says the signal that actually predicts loyalty satisfaction is probably sitting in your data ignored while you watch purchase frequency. The disclosure paper says, most AI accountability statements are theater, not transparency. Here's the tension I keep coming back to. Every one of these papers is pointing at a gap between what we're doing with AI and what we should be doing. But none of them have hard proof that closing those gaps changes outcomes. The hybrid agency model isn't tested. The GAN accuracy doesn't transfer to your live data. The disclosure framework comes from academic publishing, not commercial work. So the pattern isn't act on this. The pattern is the infrastructure questions are real. The answers aren't proven yet. And the teams building the accountability layer now will be ahead of the ones scrambling to retrofit it later. Not hype, architecture. Here's the playbook from today. One, audit your team's deliverables. Map which tasks AI handled and which required human judgment. The second list is your value proposition. Protect it. Two, check whether your churn model or retention logic includes satisfaction scores as an input. If it doesn't, that's your next 60-day test. 3. Pull your last AI-assisted client deliverable and read your disclosure language out loud. If it doesn't say what AI did and who's accountable, rewrite it before the next one goes out. Evidence check on all of that? Two of today's papers are conceptual or preprint. One relies on a Kaggle data set. Use them to decide what to test, not what to blindly believe. Links to all three papers are in the show notes. Read the originals before making major decisions. Want the human expert take? Join Dr. Eva Wolfe every Friday for the AI Marketing Radar Roundup, where she extracts no nonsense, money-making tips, practical strategy, and real business opportunities from the week's research. Subscribe on Apple Podcasts, Spotify, YouTube, and wherever you listen to podcasts. This is Avita for Big Plans Media, and I'll be back in the next radar brief.