Based on my April 24, 2025 conversation with Sara Nay on Duct Tape Marketing. The examples come from that interview; the exercises below are practical ways to apply the ideas.

Most AI use begins with a request for output: a draft, a summary, a list of ideas, or a plan. That can be useful, but it leaves a bigger opportunity unexplored. AI can also help you examine a problem, challenge an assumption, and discover the questions you have not thought to ask.

In a 2025 conversation on the Duct Tape Marketing podcast, I described AI as a co-pilot for the work of thinking. It needs context, but can help you see that context from more angles.

A different kind of assistant

A one-shot prompt asks AI to give an answer. A strategic conversation asks AI to help improve the question.

Imagine you are planning a new offer. You may have a strong instinct about the audience and the problem, but you could still ask AI to identify assumptions in your plan, propose alternative explanations for the evidence, or list the information a skeptical buyer would need before acting. The purpose is not to let the model choose the strategy. It is to make your own reasoning more visible.

In the source interview, I described asking AI for feedback, considerations, and questions. That is a more productive use than waiting for it to take initiative. A model does not know the hidden constraints, the internal politics, or the customer nuance that shape a real decision. You have to bring those into the conversation.

Start with a real decision, not a generic topic. Explain the goal, the audience, what you already know, and what would make the decision difficult. Then ask the model to interrogate the plan before asking it to help write anything.

Test a possibility before you commit to it

One early experience changed how I thought about AI's creative potential. When ChatGPT first appeared, I wondered whether it could do more than rearrange familiar material. I searched for an idea that seemed unlikely to be an existing recipe and asked for a Mediterranean ice cream recipe.

The ingredient list seemed plausible to me. That was the useful moment: it suggested that the model could combine ideas in a way worth examining. I did not make the recipe or taste it, so it was not proof that the result would work. It was a prompt to explore what else the system might help me imagine.

That distinction matters for marketing work. AI can generate possibilities, analogies, and combinations that move a team past a blank page. Possibility is not evidence, and a plausible first pass is not a finished strategy. But it can give a skilled marketer something concrete to evaluate.

Use low-stakes tests to explore an unfamiliar direction. Ask for three contrasting approaches, the assumptions behind each, and the evidence that would help you choose among them. Keep the test separate from a customer-facing commitment.

Ask questions that improve the plan

Strategic dialogue becomes stronger when you invite disagreement. Ask AI to take the role of a skeptical customer, a finance partner, a subject-matter expert, or a competitor. Ask what would have to be true for your plan to succeed. Ask what information would change the recommendation. Ask it to find tensions between your goal, your constraints, and the evidence you have collected.

The quality of that exchange depends on the quality of the material you provide. A model can work with customer interviews, research notes, a positioning draft, a sales-call summary, or a campaign brief. It cannot infer the most important facts merely from a broad instruction.

This is where marketers retain the central role. You decide what context is accurate, which perspective is appropriate, and whether a suggested question reveals something useful or just creates noise.

Brief research like a real assignment

Research-oriented AI can gather and organize a large amount of material quickly. In the source conversation, I described it as a substantial assignment, closer to giving someone a project charter than asking an ordinary chat question.

That means the request needs boundaries. State the decision the research should inform, the sources it should prioritize, the claims it should avoid, and the form of answer that would be useful. A well-framed brief can turn a broad investigation into a set of questions, evidence, and tradeoffs a team can actually discuss.

Save the research brief with the output. Check the important citations before using them, identify gaps or contradictions, and write down the decision the research changed. That makes the work reviewable instead of treating an AI-generated report as an authority by itself.

Keep the final evaluation human

AI can help a person see options more quickly. It can create a useful first pass, organize a messy set of inputs, or reflect a plan back in a more structured form. It cannot carry the responsibility for knowing the customer, making a promise, or deciding which tradeoff is worth making.

The value comes from the loop: bring the model a real problem, ask better questions, test its suggestions against evidence, and make the decision with human judgment. Over time, that practice can improve both your use of AI and the quality of the strategic work around it.

Use AI to make your thinking more rigorous, not to outsource the part of the work that makes your marketing worth trusting.

Put the thinking into practice

When you are ready to choose a repeatable task, use my guide to choosing which marketing workflows to automate first. Then plan the workflow with this worksheet.

Get practical AI marketing field notes by email, or explore more interviews and teaching.