The Prompt Playbook: How To Turn Your Marketing Team into Prompt Engineers
Regardless of your job, company, or industry, Artificial Intelligence (AI) is dominating the conversation. Its impact is probably felt nowhere more than in the marketing arena. Many marketers are being asked to implement AI tools or practices immediately, often without a clear plan or starting point. It’s both an incredibly exciting and equally daunting time to be a marketer.
I would argue that the most important new skill for today’s marketer is not AI, but instead prompting, with AI coming in a close second.
When you look at all of the components and use cases for AI, it quickly becomes clear why prompts matter so much. The prompt is really at the start of most AI journeys.
Whether your AI journey is a short one, maybe getting help on crafting AEO-focused social content, or a much longer journey, building complex workflows to tie your CRM in with your email marketing platform, and in turn with your analytics and reporting tools, the journey still starts with the prompt.
What are prompts?
The prompt is the instruction that you give to AI to eventually guide its output. It sounds very straightforward, but it’s one of the easiest things to get wrong.
Many marketers are excited to begin using AI but often approach it without a clear plan. Think of your AI tool as a member of your team. Much like when working with an employee, the more informed they are at the start of the project, the better the chance of a desirable outcome. The same principle applies to prompts.
Crafting the Prompt
Some things to consider when crafting your prompt:
Who is the audience?
What is the specific output you would like to receive?
What is the goal of the project?
Where will you be using the output?
How should the output sound or look?
What should the tone of the output sound like?
The more background, information, context, and style guidance you can provide, the better.
Prompt Types
There tend to be two schools of thought regarding prompt types. The first is the single-turn prompt, or one-shot prompt. This type of prompt has different names, but basically, you feed the AI one complete prompt. It could be as short as a few keywords or as long as several pages. You enter your prompt, attach any necessary documents or links, and click submit.
An example of this prompt type could be as simple as attaching a new report and asking the AI “Can you please write a 100-word summary of the attached report, and provide the top takeaways from the report?”
The second school of thought is the conversational prompt or “chained prompt.”
In this type of prompt, you might start by asking the AI “Can you write a 100-word summary of the attached report, and please provide the top takeaways from the report?” Then you would look at the output and next ask, “Can you make the summary targeted for my Chief Marketing Officer?” Then you could look at the output and maybe add another layer: “I like the summary that you provided. Could you also recommend some action items that my CMO should take based on this new report?”
I tend to be a fan of the one-shot prompt but will often end up being a conversational prompter if I think that the output may have missed the mark the first time around.
Pro tip. If you seem to be going in circles with your AI teammate, and the results are not getting better, it can help to start fresh with a brand-new prompt. Sometimes the AI teammate just gets more and more confused and is actually building on results that you don’t like or don’t find helpful. Resetting with a clean prompt gives you a chance to provide better directions and will often yield better results.
Prompt Frameworks
Prompt frameworks are structured methods for writing prompts that force the author (prompt engineer) to give the clearest set of directions and hopefully get the best results.
There are many different frameworks out there, but below I will break down three common ones.
RTF Framework (Role, Task, Format)
Role – Who should the AI pretend to be? (You are the CMO of a tech startup in the cybersecurity industry.)
Task – What do you want the AI to help you with? (Summarize a new white paper from a competitor.)
Format – What should the output look like? (Short written recap of the white paper with key points in bullets.)
Resources you should provide to help your prompt deliver good results: You should attach the new white paper, you could provide a link to your company’s website, you could also provide your competitor’s website, and it is always a good idea to include your style guide.
CTF Framework (Context, Task, Format)
Context – What is the background on the request? (This will be a LinkedIn post to promote an upcoming webinar.)
Task – What do you want the AI to help you with? (Draft a post for our company’s LinkedIn page to encourage customers to attend our webinar.)
Format – What should the output look like? (The post should be no more than 3–4 sentences long.)
Resources you should provide: a link to the webinar, your style guide, sample copy from other LinkedIn posts, and possibly the original abstract or slides from the webinar if available.
RASCEF Framework (Role, Action, Steps, Context, Examples, Format)
Role – Who should the AI pretend to be? (Act as a Senior B2B marketing strategist at Company XYZ.)
Action – What should the AI do? (Help me craft a compelling white paper based on the below inputs.)
Steps – What are the steps that you would like the AI to follow?
Clarify the goal of the creation of the new white paper (lead generation for top of the funnel).
Identify the audience and help explain their key pain points (cybersecurity professionals).
Lay out the structure of the desired white paper (samples are attached).
Draft the content of the white paper.
Provide two versions where you vary the tone so we can choose which fits best (our style guide is attached).
• Context – What is the background on the request? (We need the white paper to help drive top-of-funnel leads.)
• Examples – Reference documents or examples to guide the project.
• Format – What should the output look like?
Resources you could provide might include a link to your website, similar white papers, persona documentation, your corporate style guide, and samples of an executive’s voice if needed.
Pro tip. Frameworks can feel overwhelming and are often an area where marketers get mired down while using AI. Keep in mind that the frameworks are really intended to force you to think through every aspect of your project before you start.
If you jump in too fast and your AI teammate produces something completely off base, it’s not the end of the world. You can easily take a step back and start fresh. You would likely have the same outcome if you gave your human colleague a vague request. The difference is that your human teammate might have spent two weeks marching in the wrong direction due to the unclear assignment. Your AI teammate thankfully may have only wasted an hour.
And as we learned above, if the output is pretty close but not exactly what you are looking for, you can always add on to your prompt to improve the output.
How can you prevent having to start from scratch every time?
The answer is to begin to curate a “Prompt Library”. The prompt library is simply a well-organized list of reusable prompts that helps you and your team work faster, and it alleviates the need to reinvent the wheel every time.
When I first began learning how to prompt effectively, building a prompt library was one of the first things I did. My library lives in a Google Doc, but yours could live anywhere that works for you and your team.
I started by dividing the library into sections based on the things that I might want to use AI to help me with.
These are just some sample sections that I started with. Yours are likely to vary based on your role, team, or needs:
Prompts for Content Creation
Prompts for Marketing
Prompts for Excel
Prompts for Brand Strategy
Prompts for Copywriting
Prompts for Competitor Analysis
These are just a few examples. I currently have 32 sections in my library and it continues to grow. Every single time I see a prompt, I copy it into my library, even if it isn’t related to marketing or to my current industry. You never know when you may need that prompt in the future or may need to share it.
For example, I have a section for “Prompts for Brand Pricing Strategy.” I have never worked on pricing in my career and I have never been asked my opinion on brand pricing strategy, but I have a resource that may help a member of the team in the future. The library becomes a long-term team toolkit instead of a personal cheat sheet.
Pro tip. Many of the prompts in my library have not been battle-tested, so they may actually not be very good. Once I have used a prompt and it delivered good results, I highlight it in yellow and strike out those that did not seem to work.
I strike the prompts that are duds out instead of deleting them because the prompts will have different output depending on the AI tool or large language model (LLM) that you are using. So, for example, if your company is using ChatGPT and the prompt really didn’t seem effective, it might work with another tool or LLM, so I didn’t want to delete it entirely.
What the heck is a prompt engineer?
Simply put, a prompt engineer is the person responsible for creating the instructions (prompts) that will guide your AI tool into producing the best output.
In most organizations, the person who owns your AI tool will likely become the default prompt engineer. As the growth of AI use in the field of marketing explodes, we will see this role quickly morph into a stand-alone job. I peeked at LinkedIn for this specific job title, and there are currently 98 jobs listed. It’s unlikely that we would have seen more than a handful a year ago.
What prompt engineers actually do
A prompt engineer:
Is able to break complex asks down into clear steps
Understands the business and its needs and knows the best information to provide to the AI
Can craft templates and frameworks and, once tested, can use them to make AI more reliable
Experiments with different prompts to ensure the output is accurate and meets the project need
Collaborates internally to help speed AI adoption and create processes
Writes it all down (Prompt Library!)
Skills of a strong prompt engineer
A solid understanding of how LLMs work
The ability to choose the right AI tool for the job
Can translate business needs into clear instructions
The ability to think about processes and the steps needed to accomplish a task
Curiosity
The ability to think outside the box
Comfort working across the organization depending on the needs of the project
How Prompting Can Benefit Your Brand
AI is the hottest topic in marketing. It’s also one of the easiest places to lose time without ever getting the full value from your investment.
A well-crafted prompt can help your organization:
Save time and money
Improve accuracy
Reduce review cycles
Turn AI from a “fun toy” into a reliable part of your workflow
Why Prompting is So Important
Prompting is quickly becoming one of the most important skills in modern marketing. It forces us to pause and really think about our business, our audience, our products, our goals, and how we want to tell our story. Prompting can help you quickly gather scattered thoughts, ideas, and data and can help guide the organization.
You don’t need to be the most knowledgeable or most technical member of your team to become a strong prompt engineer. You just need curiosity and the willingness to try new prompt types and frameworks, build your own library, document it all, and help guide your team.
Most importantly, prompting can help you really harness these new AI tools that you have been given and allow you to deliver real results.
