A recap of my session at AMA St. Louis's "The Art & Science of Marketing" - and why the Q&A was the best part.
It started, as most of my best stories do, on a highway.
December 2024. I was maybe five hours into a 2,000-mile drive from St. Louis to Los Angeles - somewhere in the flat middle of Kansas - when I pulled over and did something I'd never done before. I didn't reach for a podcast. I didn't call a friend. I opened Gemini on my phone and typed a question I'd never asked a machine:
"I am a successful professional, but I am failing to scale. Here is my data. Here is my friction. Tell me why I am stuck."
That moment - that specific pivot from treating AI like a search engine to treating it like a strategic partner - is the beating heart of everything I teach. And on April 9th, I got to share it with over 100 professional marketers in one of the most beautiful rooms I've had the pleasure of presenting in: the Community Music School of Webster University, right here in St. Louis.
Exposed wood, concert acoustics, and rows of agency owners, brand-side marketers, educators, and students - it's not where you'd expect to spend a Thursday morning talking about AI governance and custom assistants, but it was exactly the right setting for what turned out to be one of the most energetic conversations I've had with a live audience in a long time.
The event was the AMA St. Louis 2026 Annual Marketing Conference, "The Art & Science of Marketing" - a half-day experience running from 7:30 AM to 1:30 PM, organized by my friend and AMA co-President Marisa Lather. The lineup was genuinely impressive: emceed by the brilliant Keisha Mabry Haymore, with sessions from Ryan Green, Savannah Westbrock, and Allie Haupt on the producer mindset in marketing; Amanda Chamov of American Express Global Business Travel on the neuroscience of memorable experiences; a sharp panel moderated by Tracy Polansky; and a closing keynote from one of marketing's most fearless voices, Katie Martell. If you've never been to an AMA St. Louis event, you're missing out. This wasn't a stuffy conference. It was a room full of people who came ready to think, debate, and occasionally groan at corporate cautionary tales.
(More on the groaning in a minute.)
The Talk: AI as Partner, Not Answer Engine
My session was titled "The Custom Assistant Revolution: How to Stop Using Generic AI and Build a Proprietary Brain for Your Business," and I had nearly an hour including Q&A. The event description captured it well: most high-performers treat AI like a high-speed search engine. They ask for a draft, spend twenty minutes editing the robotic output, and eventually realize they could have just written it themselves. That's the result of what I call Admin Drag - the 70% of the week spent moving data rather than building relationships.
My central argument is one I've been making for a while now, but it keeps landing fresh because most people are still on the wrong side of it:
You are not using AI wrong because you're using the wrong tools. You're using it wrong because you're treating it like a vending machine.
You put a question in. You get an answer out. You evaluate whether the answer is good enough. Repeat.
That's not a partnership. That's a very expensive search bar.
The shift I wanted every person in that room to make is from prompting to architecting. From asking AI for outputs to building AI into a context-rich partner that understands your goals, your voice, your clients, and your constraints. That's the difference between a tool and a true Digital Crew member - and it's the difference between being a highly-paid data entry clerk and being the Director of your own proprietary brain.
To make that concrete, I walked them through the PRICE Framework - my repeatable 5-step standard for building prompts that actually work:
- Persona - who is the AI in this context?
- Resources - what does it have access to?
- Instructions - what exactly do you need?
- Context - what does it need to know about you and the situation?
- Examples - what does "good" look like?
Context, I kept hammering, is the multiplier. Without it, AI gives you generic. With it, AI gives you yours.
The Three Levels of Memory (And Why the Room Leaned In)
One of the moments I could feel shift the energy in the room was when I broke down what I call the three levels of AI memory. This isn't jargon - it's the practical architecture that separates people who get consistent results from AI from people who keep starting from scratch every single morning.
Level 1: Chat context. Everything within a single conversation. The AI knows what you've said in this session, and that's it. When the window closes, it's gone.
Level 2: System memory. The persistent memory you can set in your account settings. This is where you tell the AI who you are, how you like to communicate, and what your baseline context is. It travels with you across conversations.
Level 3: Custom instructions and custom assistants. This is where things get powerful. Custom GPTs, ChatGPT Projects, Microsoft Copilot Agents and Notebooks - these are purpose-built AI partners loaded with specific knowledge, a specific persona, and specific instructions for a specific job. This is where you stop ending up with AI Amnesia and start building a Proprietary Brain.
I gave a live example: the difference between my Podcast Showrunner assistant - a focused agent I built to handle the operational load of producing The AI Hat Podcast - and CLU, my AI Chief of Staff, who helps me think strategically about the business as a whole. My longtime friend Vernon Ross was in the audience, and I gave him a shoutout in that moment because he's seen both in action and has been building his own versions. He was kind enough to return the favor later when he was called up to share what he'd taken from the day - which was genuinely one of those community moments that reminds you why showing up in person still matters.
Also in the room: Alicia Underwood of The Social Box, who served as a panelist later in the day and is doing seriously sharp work in the social media space; and Perry Drake - UMSL professor and founder of the Midwest Digital Marketing Conference. Having peers like that in the audience keeps you honest. You don't phone it in when your friends are watching.
If you want to go deeper on building custom assistants like these, I've written about the Custom GPT approach here and about building a full AI Chief of Staff here.
The Samsung Story (And the Groan Heard Round the Music Hall)
If you work in marketing and you haven't heard the Samsung story, you need to. I told it in the room and the collective groan was immediate and visceral.
Here's the short version: In 2023, Samsung engineers used ChatGPT to help debug proprietary source code. The code - and therefore Samsung's trade secrets - was ingested by OpenAI's servers and potentially used as training data. Samsung had to ban ChatGPT internally afterward. The damage was done.
The groan in that room wasn't just about Samsung. It was recognition. Because every single person in that room either works at an agency handling client data, or manages sensitive internal information, or both. The question isn't whether your team is using AI with your clients' data. The question is whether they're doing it with your knowledge and governance in place.
I've written about this exact problem in depth - it's what I call the Shadow AI Crisis, and it's one of the most urgent issues facing marketing agencies right now. The first step toward solving it isn't a ban - it's a governance framework your team will actually follow.
The Q&A: Where the Real Learning Happened
I'll be honest - I love the Q&A more than the talk. The prepared remarks are mine. The questions are theirs, and they always reveal what's actually keeping people up at night. Here's what came up from this audience:
1. How should agencies navigate AI with clients who are skeptical or anti-AI?
This is the real agency dilemma right now, and it's more nuanced than it looks. You have clients who are worried about privacy and compliance - which is legitimate. You have clients who have blanket "no AI" policies written into contracts, which is increasingly common. And you have clients who don't know what they want yet and are looking to you for guidance.
My answer: this is a conversation about governance and transparency, not about hiding what you're doing. Build your agency's AI policy explicitly. Know which tools ingest data and which don't. And position yourself as the expert who helps clients develop their own AI posture - because that's a service they desperately need and very few agencies are offering yet. That positioning alone is a competitive differentiator.
2. What exactly is RAG, and how can we use it in ChatGPT?
RAG stands for Retrieval-Augmented Generation. The short, non-technical version: instead of relying only on what an AI model learned during training, RAG lets you plug in your own documents and data so the AI can reference them when answering questions. It's the difference between asking a smart stranger for advice and asking a colleague who has read every file in your company's shared drive.
In practical terms, ChatGPT Projects is one of the most accessible ways to implement this right now. You upload your documents - an HR manual, a client brief, a sales knowledge base - and the AI can reference them in every conversation within that project. I covered several marketing-specific use cases: building a client knowledge base so your team can instantly answer questions without digging through files, creating an HR policy assistant so staff can self-serve on common questions, or building a sales battlecard agent that knows your competitive landscape cold.
If you want a deeper dive on this, check out The AI Librarian: How RAG Can Revolutionize Your Business.
3. What's the difference between Microsoft Copilot Agents and Copilot Notebooks?
An important distinction if your organization is on Microsoft 365. The short version: Copilot Notebooks are your grounding layer - you feed them documents, meeting notes, and reference material so Copilot has rich, accurate context to draw from. Think of it as the AI's briefing binder. Copilot Agents are the action layer - they can execute tasks, move between apps, and operate with more autonomy. Notebooks inform. Agents do. I've written a full breakdown of Copilot Agents here and Copilot Notebooks here if you want the complete picture.
4. How do I feel about AI's environmental impact?
I appreciated this question because it came from a place of genuine ethical concern, not gotcha energy. My honest answer: the short-term environmental costs are real, and I don't dismiss them. But I believe the long-term gains in scientific advancement - in drug discovery, climate modeling, materials science, energy optimization - will more than offset those costs. That's not a dismissal of the concern. It's a bet on the trajectory. And I think it's the right one.
5. What are some practical marketing use cases for RAG?
Beyond what I mentioned above: a brand voice assistant trained on your style guide and best-performing content, a client-specific agent for each major account loaded with their brand guidelines and campaign history, and a competitive intelligence agent that synthesizes your research into instant briefings. The throughline is always the same - stop relying on AI's generic knowledge and start giving it your knowledge to work with.
Katie Martell's Keynote: Woke Brands and Weary Consumers
The day closed with the session I'd been looking forward to most: Katie Martell's keynote, "Woke Brands and Weary Consumers."
Katie is one of marketing's most fearless voices, and she earned that reputation on Thursday. Her talk unpacked one of the most uncomfortable truths in the profession right now - the gap between what companies say and what they do. Drawing from years of documenting this landscape in real time, she walked through how many major corporations publicly championed causes like DEI, LGBTQ+ rights, and women's rights while simultaneously donating millions to politicians and organizations actively working against those same communities. She called it what it is: performative messaging. Woke-washing. And she did it with receipts.
As a marketer, it was a useful gut-check. The question she left hanging in the air - what does authentic advocacy actually look like in practice, in an increasingly polarized world - is one the profession hasn't fully answered yet.
It also made the theme of the day, "The Art and Science of Marketing," feel genuinely earned. Science can tell you what's resonating. Art is knowing what's worth standing for in the first place.
The Bigger Picture: Why In-Person Still Wins
I want to close with something I believe deeply, and something the entire lineup on Thursday reinforced.
There are things that happen in a live room that cannot be replicated in a webinar or a LinkedIn post. When I told the story of pulling over in Kansas and asking CLU for counsel instead of content, I could see people connect it to their own moments of friction. When I broke down the three levels of AI memory, I could see phones come out for notes. When I told the Samsung story, I could hear the exact moment the implications landed.
Keisha held the whole day together with the kind of warmth and command that only comes from someone who has delivered 500+ keynotes and actually means it when she talks about connection. Ryan, Savannah, and Allie made a compelling case for why taste and creative courage still matter in a data-driven world. Amanda brought the neuroscience of memory into a room full of marketers who now can't stop thinking about emotional peaks. And Marisa and her team built something worth showing up for.
That's not content. That's learning. And it's why I keep showing up in rooms like the Community Music School of Webster University, talking to groups like AMA St. Louis, and why I'm deeply grateful to everyone who made April 9th what it was.
If you were in the room and want to keep the conversation going - or if you're an event organizer thinking about bringing this kind of session to your audience - the best next step is the same either way.
Explore keynotes, workshops, and training options at The AI Hat.
The revolution began in Kansas. But it continues right here in St. Louis.
