.png)
Everyone in marketing has an opinion on AI right now, but most of the debate stays at the surface: which tool to buy, which prompt to try. I think the meaningful line is between companies that are AI-enabled and companies that are AI-first, and most of us are still on the wrong side of it without realizing it.
I'm VP of Marketing at ReBiz, a marketing and analytics platform for multi-unit retailers across wireless, mattress, jewelry, car wash, and more. Before that, I spent close to a decade running marketing for AI-driven SaaS companies serving enterprise procurement teams, health systems, and food service chains. Every one of those roles came back to the same question: how do you build a marketing function that runs on AI?
(You can listen to Benjamin's episode on our podcast, CMO Convo, right here, watch the video version below or continue reading this episode in blog format)
AI-first vs. AI-enabled
These two terms get used loosely, so here's how I define them.
AI-enabled is a bolt-on. You're using ChatGPT with some prompts, or a few agentic workflows built natively in Claude or another platform, but AI sits as an accessory to how your team works rather than at the center of it.
If you've been running a business for years or decades and AI showed up in 2021, AI-enabled is a perfectly fine place to be right now, and honestly a bare minimum.
AI-first means embedding AI at the core of your go-to-market. If you pulled AI out entirely, your marketing function would look completely different. That means central reference documents that run your marketing:
- Your brand narrative living in an always-on file
- Your content strategy in another set of files
- Your marketing operations running through agentic workflows rather than static tools
There's a real difference between HubSpot running lead routing on its own and a native AI workflow directing HubSpot on what to do next.
AI-first is the DNA. AI-enabled is the bolt-on.
What does an AI-first tech stack look like?
I get asked about the ideal marketing tech stack constantly, in CMO communities including CMO Alliance. My answer is minimalist: consolidate as much as you can. If I had to name one platform, it's Claude. That's what I use to build my AI-first marketing engine.
I'll put it the way I put it on the podcast: "I would think of Claude as if you're running errands and you're shopping. Claude is the Amazon or the Costco of your list." You go there, and it handles nearly everything on it.
A lot of the niche AI tools out there are really just packaged solutions for problems that basic prompt engineering and workflow setup inside Claude or ChatGPT would already solve, and those niche tools add up fast in cost, individually and especially combined.
If you're weighing build versus buy, then upskilling yourself is worth it. Skip it, and your skill set as a marketing leader stays limited long-term. If you're genuinely constrained and need to ship right now, a bolt-on tool can carry you for a while. The goal is still to consolidate and build a team that can use the core tools well.
Benjamin Brown
Where do you start if you're building from scratch?
Inside Claude, you build skills, which are plain-text files that train the tool to do a specific task to your standard. Setting up an AI-first marketing engine isn't structurally different from setting up marketing at any company. You still need to understand your brand narrative, your value proposition, and your ICP, and how they connect. What's changed is how fast you can get there.
Getting there means talking to everyone: company leadership, sales (what are they seeing in the field), customer success (what are customers saying), and customers directly. All of that information goes into one central brand narrative document, a long file that's always on and always referenced. Everything else you build depends on it. In practice, it's a summary of transcripts, notes, and existing assets combined into one core reference document for the company.
From there, you build a style guide (voice, written style) and a visual guide (imagery for ads, social posts, blog covers, slide decks). These used to be nice-to-have assets that sat in a drive somewhere and got ignored by freelance writers. Now they're always-on skills that AI references for every piece of content you produce, from a single email to a 40-page white paper. AI won't replace your design team. It can get a design most of the way there so people spend less time building from scratch and more time finishing it.
Then there's product messaging: a narrative for each product you sell, your brand narrative applied at the product level, and what lets product marketing scale.
None of these are static documents. Your company changes, the market changes, competitors rise and fall. These assets need to evolve at the same pace as the rest of your go-to-market, or they stop being useful.
How do you shift your team's mindset?
Change management has been the biggest obstacle in every AI rollout I've been part of, whether I was selling the product or adopting it internally. Every customer's underlying worry is adoption and whether their team will actually use it.
My honest answer, every time, is that AI is meant to increase your impact and your output, not replace you. If AI can easily do your job, that's a signal to upskill and become a better contributor. That has to be the starting point, and it has to be genuine. Any underlying motive shows up in how you talk about it, and people pick up on it.
I'll set aside the bigger debate about AI and headcount, which plenty of people, myself included, have strong opinions on. What matters here is giving your team security first, then setting goals and a timeline. Most people who've been in the workforce a while never had formal AI training, because it didn't exist yet when they were in school.
As I said on the podcast: "AI is a race car, and great drivers obviously need training on how to operate that race car." So train them: YouTube, community forums and threads, outside consultants, whatever gets the job done.
Give people room to fail and experiment, and let them share both wins and misses with the group. This is a process, not something that happens overnight, but the team is accountable for learning it over time. Start from the AI outputs you actually want to produce, and let the skill-building follow from there.
Benjamin Brown
Where does human judgment stay nonnegotiable?
Storytelling is what got me into marketing in the first place, and it's still the clearest example of where AI hits a limit. AI can help with a lot, but it can't tell a company's story on its own, not in a way that reaches people from the heart. A person still has to lead that process. AI can help with execution. The storytelling is why the rest of us are here.
Human in the loop means having a person review some or all of an AI output before it goes out, for quality control, reliability, security, or compliance. Every customer I've sold AI to has been more comfortable once that piece was in place, and the same holds inside a marketing team.
A simple version: AI drafts a blog post, a person edits it before it publishes, instead of an agentic workflow that drafts and publishes with nobody looking at it first. At ReBiz, this shows up in what we call supervised AI.
Our system can flag something happening in a store, potential theft, for example, and a person reviews the footage before it ever reaches a store manager. A false positive there could mean an unnecessary call to the police or an unnecessary termination. That review step cuts down on costly, consequential mistakes.
Brand strategy and vision are hard for AI to build on its own, even with every competitor report and market analysis fed into it. It doesn't have eyes and ears in the field, and it can't make the creative judgment calls a person can. Someone has to start the loop, trigger it, and take responsibility for the story and the budget behind it.
Some brands run marketing entirely on AI, usually because there's no budget for a team. If it's AI or nothing, take the AI. But AI works from its existing dataset. There's no new information in what it produces, and it defaults to certain patterns. That's part of why an AI-built website is often easy to spot, the same way people have gotten tired of copy full of em dashes and short, choppy sentences. People call that AI slop, and the human in the loop is what keeps your go-to-market out of it. Skip that step, and it gets hard to grow, or even hold onto, a brand people trust.
This conversation would have looked completely different six months ago, maybe even six weeks ago. Things are moving fast, and the job for a marketing leader is to stay close to where AI is heading without feeling pressured to adopt everything at once. What makes a marketing function work hasn't changed in the age of AI. What's changed is how you use it to run that function at scale.
Become a CMO Insider
Thank you for subscribing
Get exclusive insights, frameworks, and strategies from marketing leaders driving real business impact.
An email has been successfully sent to confirm your subscription.
Follow us on LinkedIn
