The Future Focused CMO: AI Prompt Engineering - How You Reverse Engineer Success

Listen to the companion podcast for this article HERE

Artificial intelligence in general and LLMs specifically are rapidly reshaping the marketing landscape. These technologies are helping marketers achieve deep levels of personalization, efficiency, and strategic insight. The era of AI-driven marketing is here, and CMOs must not only embrace these changes but also strategically integrate AI to unlock its full potential.

Theory is all well and good, but platitudes do not help us deliver. So I went hands-on with Google’s NotebookLM to learn more about what the marketing industry is talking about when it comes to LLMs and integrating AI into daily marketing life. There is a LOT of chatter out there. But, by uploading over 600 pages into my NotebookLM and asking it to identify themes, I was able to very quickly see where there is broad agreement, and what topics are still on being developed.?

One theme and skill emerged as the clear starting point: Prompt Engineering. This doesn't make it the most important theme or skill to understand when it comes to marketing with AI. But it does mean it’s a very good starting point. Future focused CMOs need to not only be aware of the impact AI is having on their discipline, but also learn practically how to use it, starting today.?

The Set Up.

The first practical example of putting AI to work in marketing is how I researched for this article. I downloaded 15 different eBooks and articles (over 600 pages!) on AI in marketing. You can see a spreadsheet with all the article and eBook source links here.? Then I uploaded them into Google’s NotebookLM ?and asked it to summarize the source material. Finally - and here is the magic - I conducted a chat interview with the AI to ask questions like, “what are three key themes that emerge from across this data set?”, and “what are the most important practical skills marketers should learn to effectively incorporate AI into their daily responsibilities?”?

If you’d like access to the notebook I used, including all the sources, to conduct your own conversation, just message me here on LinkedIn with your email and NotebookLM.?

This approach let me rapidly digest a huge amount of content from knowledgeable sources that I trust. Then it let me instantly identify the areas of overlap between the sources that signal agreement and thematic strength. This agreement is a good indicator of relevance and importance. Then I was able to engage with the AI system, conversationally, to tease out specifics, edge cases, and innovation hiding in the full data set.??

I was able to write this article based on an interview I conducted with AI that had consumed over 600 pages of AI for Marketers content from companies like Salesforce.com, Hubspot, Dell, Boston Consulting Group and others. This instance of AI in my NotebookLM project became my subject matter expert.?

And The Winner Is…

Prompt Engineering: The Key to Unlocking AI’s Potential

One of the most crucial yet often overlooked skills in AI-driven marketing is prompt engineering. Prompt engineering is simply the way you ask your AI system a question or instruct it to perform a task. But the way you ask is much different than typing in a search query and is more like asking a brand new employee to perform a task. A new employee knows how to do a task but? may not know where all the important data are, or what the context is. You have to explain to them the task with a bit more specificity. You engage with the LLM system the same way.?

For example, a vague instruction like "Analyze this dataset" or, “give me ideas for a blog post” will yield generic results. However, a refined prompt—"You're a data scientist analyzing our quarterly demand data. Identify key engagement patterns for each vertical, then create a visualization that shows those patterns and highlights anomalies"—will produce far more valuable insights.

Two modes of AI Engagement

I regularly use two modes for engaging with AI. I’m sure there are many more, but these two work well for me right now. These modes are:

• Conversational engagement
• Task direction

Conversational engagement is perfect for when you want to learn or consume information and you want AI to help you access more content more quickly than you can alone. Conversational engagement means you treat the LLM like a person and talk to it.?

As with any conversation, it is important to first establish context and topic. With a general online system like ChatGPT, Perplexity, Gemini, Claud or others, it’s helpful to first tell it what you’re going to be talking about. Point it to some websites, upload some documents, tell it an established fact or concept that you want to explore together. Then talk. Have a conversation. Interview the system like you would a SME. Become the student and let the system be the teacher.?

The prompts that you create for conversational engagement can be iterative as the system will keep conversational awareness. You can say things such as, “Given what we’ve just been discussing, what do you think about _______?”. The goal is to create different conversational pathways for uncovering new or unique understanding that the AI has the ability to reveal.?

Conversational engagement with LLMs is a very easy way for marketers to process and extract and refine insights that may be buried in and across data sets that are too unwieldy for individuals to efficiently consume.?

Conversely, task directed AI prompts are more like computer commands presented in natural language. “Given context _____? and these inputs______ [do a job | create a deliverable | write some code] that achieves objectives ______ and _____.”?

Much of the content on prompt engineering focuses this kind of task direction.?

Task direction is important for development of agentic AI - task and deliverable focused AI Agents that do a task and deliver an output. Such agents can be connected via process integration to create higher order autonomous or semi-autonomous AI agents that can complete complex tasks on their own.?

It is easy to remember when to use the two different modes of engagement by remembering:?

• Conversational engagement with AI helps you learn a thing.?
• Task engagement with AI helps AI do a thing for you.?

Power in Specificity.

The quality of your AI output is highly dependent on the quality of the prompts. It's crucial to be specific and detailed in your requests.

Let’s take a look at some specific examples of good prompting and poor prompting.

Bad Prompt: "Analyze this dataset"

Good Prompt: "Analyze marketing engagement trends in this CSV"

Excellent Prompt: "You're a data scientist analyzing our quarterly marketing campaign data. Identify key engagement patterns, create visualizations, and highlight anomalies. Focus on customer segments and regional variations"

This progression shows the value of:

Setting the Stage (Context) Tell the AI who it's talking to, define the expertise level needed, and specify any constraints or limitations.

Defining the Task (Action) State exactly what you want, include specific requirements, and break complex tasks into steps.

Specifying the Output (Format) Define the structure, set the length, and clarify the tone and style

Constraining AI

Setting clear parameters about what you want vs don’t want or what you want the AI to focus in on? helps the AI understand exactly what success looks like.

Bad Prompt: "Analyze this dataset".

Good Prompt: "Analyze customer churn patterns".

Excellent Prompt: "Analyze our customer churn dataset with these parameters: Focus on users inactive > 30 days. Compare behavioral patterns pre-churn. Include statistical significance tests. Output visualizations in google sheets".

Requesting Examples and Variations

Getting multiple versions of the same content helps explore different approaches and find the perfect fit..

Bad Prompt: "How do I implement this?"

Good Prompt: "Show different ways to implement this creative design"

Excellent Prompt: "Provide 2 implementations of an AB test for this creative ad design. Optimize one implementation for impression efficiency. Optimize the other for engagement. Include predicted analysis benchmarks for each".

Refining AI prompts is a learning process and each LLM model has different idiosyncrasies that you need to learn. Here are some more examples of refining AI prompts:

Rather than asking for “ideas for a blog post,” specify “Generate three blog post ideas on how AI enhances personalized marketing experiences for B2B companies".

Instead of a vague request, such as "Write a headline and introductory paragraph for an email promoting our lead scoring tool,” try "The tone should be insightful yet straightforward, aimed at sales and marketing managers who want to streamline their lead qualification process".

These examples demonstrate that effective prompts are clear, specific, and provide sufficient context for the AI to generate relevant and high-quality outputs.

Wrapping it up

So what should the AI interested marketer start doing, starting now??

The first skill a marketer should learn when integrating AI into daily marketing activity is prompting

Here's why:

• Essential for AI output Prompting is essential for guiding AI to produce high-quality outputs tailored to specific marketing needs. A well-crafted prompt helps AI understand marketing goals, just as a solid creative brief aligns a team around a campaign's vision.
• Shapes AI responses Effective prompting allows marketers to shape AI's responses to produce precisely what is needed. This skill enables marketers to steer AI toward outputs that align with objectives and brand voice
• Foundation for other skills Mastering prompting provides a foundation for leveraging AI in other areas, such as content creation and data analysis. The more marketers develop this skill, the more control they will have over AI's outputs, ensuring they are accurate and aligned with brand and goals
• AI can help with prompting itself AI tools can suggest starting prompts, which can then be refined to get closer to the desired response

Prompting enables marketers to use AI effectively, regardless of the specific marketing activity.

Mastering prompt engineering allows marketers to direct AI with precision, ensuring outputs align with brand voice, campaign goals, and strategic objectives.