The hardest part of putting AI to work on a marketing team isn't teaching the tool. It's rethinking how the work gets done.
Most experienced marketers have spent years building their craft: learning how to research a market, develop positioning, recognize good design, structure a website, write a strong brief, or turn an ambiguous business problem into something actionable.
Naturally, when AI enters the conversation, our first instinct is often to identify which parts of that process can be automated and which parts should remain human.
But I think that's the wrong question.
The better question is:
If we were designing this workflow from scratch today, with AI available from the beginning, how would we do it differently?
That distinction has changed the way I think about AI adoption.
Recently, I wanted to understand how far AI could go in a type of project that I had historically considered highly dependent on human expertise: developing a website's visual and UX direction. So I ran an experiment: Instead of beginning with the traditional process, I started with AI.
I asked Claude to analyze roughly 100 relevant websites, identify patterns, help define a visual direction, translate those ideas into a wireframe, explain the reasoning behind its recommendations, and eventually generate working HTML.
Within about an hour, I had something surprisingly substantial. Not a finished website. Not something I would ship without review. But a strong enough first direction that it fundamentally changed my perception of where AI could participate in the process: work that traditionally requires significant time to research, synthesize and translate into an initial direction could suddenly be compressed dramatically.
That didn't eliminate the need for expertise. It changed where I needed to apply it.
Instead of spending most of my time producing the first version, I spent it evaluating:
Does this actually fit the positioning?
Does this feel premium or generic?
Is the hierarchy right?
Does the experience reflect what the buyer cares about?
Which recommendations should we keep?
Which ones look convincing but fall apart when you apply business context?
That distinction matters. AI accelerated the exploration. Human judgment determined the direction. And that's a very different way of working.
One assumption I think many of us have carried into the AI era is that the more strategic a task is, the more important it is for a human to create it from the beginning.
I'm increasingly questioning that assumption. Strategic work may actually be where experienced people can get some of the greatest leverage from AI. Why?
Because expertise gives you the ability to distinguish good from almost-good.
AI can generate twenty directions. Experience helps you recognize the two worth pursuing. AI can analyze a hundred competitors. Experience helps you understand which patterns matter for this particular company, buyer, market, and moment. AI can suggest a positioning framework. Experience tells you when the logic looks perfect on paper but won't survive a sales conversation.
That's not AI replacing strategic thinking, that's strategic thinking moving higher up the stack.
| Traditional workflow | AI-augmented workflow |
|---|---|
| Expertise is heavily applied to producing the first version | Expertise is heavily applied to directing and evaluating |
| Research is constrained by available time | AI dramatically expands the amount of information we can explore |
| We often start from a blank page | We can start with multiple informed possibilities |
| Iteration is relatively expensive | Iteration becomes fast and abundant |
| Experience helps you create the output | Experience helps you direct, challenge and improve the output |
The expertise didn't disappear. Its leverage increased.
Of course teams need AI skills. People need to understand prompting, context engineering, model selection, tool limitations, hallucinations, privacy, and when AI simply isn't the appropriate solution. But those skills are increasingly learnable. The bigger transformation is conceptual.
For years, software helped us execute existing processes more efficiently. AI gives us an opportunity to redesign the process itself. And that requires different questions.
Instead of: “Can AI produce this deliverable?”
Ask: “Which parts can AI explore quickly so I can spend more time making the decisions that matter?”
That framing is much more interesting to me because the goal isn't to remove humans from the work, it's to remove unnecessary constraints from how humans work.
I don't think the answer is telling teams to “use more AI.” That's too vague, and it turns AI adoption into another compliance exercise.
A better approach is experimentation: take a workflow your team already knows well and run an AI-first version alongside it, not because the traditional process is wrong, but because you need a baseline to understand what's now possible.
Then compare: What became dramatically faster? Where did quality improve? Where did AI struggle? Where was human context indispensable? Which steps no longer need to exist? Which new possibilities appeared because iteration became cheaper?
Those questions turn AI adoption from a debate into a learning process and they create room for something important: the workflow can change without implying that the people doing the work were doing it wrong before.
The technology changed.
The operating model should change with it.
Pick one meaningful piece of work your team already understands well.
Before starting the usual process, ask: “What would an AI-first version of this workflow look like?”
Then run the experiment. Let AI research broadly, generate options, challenge the brief, produce an initial version.
Then bring human expertise aggressively into the evaluation, challenge the assumptions, add the context the model doesn't have, recognize what is generic, make the tradeoffs, and finally improve the final result.
Document what you learned.
You may discover that AI saves 10% of the process or you may discover it saves 80%.
You may discover that it's excellent in one part and surprisingly weak in another.
That's exactly the point.
We're still learning where the boundaries are and the teams that benefit most from AI won't necessarily be the ones with the most sophisticated tools. They'll be the ones most willing to keep experimenting with where human expertise creates the greatest leverage in an AI-augmented workflow.
That's the mindset shift I'm trying to make in my own work.
And every experiment moves the line a little further.
Thais Campos is a Fractional CMO and GTM leader for B2B SaaS. She writes about building AI-augmented marketing systems that scale a senior marketer's judgment.