Can I Make My AI Prompting More Efficient?
Let's see if I can figure out how to create better AI content faster.
Creating content with AI. Is it good? Is it bad? Is it going to ruin the world?
The jury is still out on whether it’ll ruin the world, but I’ve found that that good ol’ adage of “garbage in, garbage out” rings especially true for AI-generated content. A mediocre, vague prompt will get create forgettable content. A well-crafted, focused prompt with some human touch in the editing and revising process though? Well, that may just be where the magic is found.
This week, I experimented with ways to be more efficient while still creating specific, context-relevant content.
The Challenge: Creating Industry-Specific Content
While AI can write for specific personas and industries, I often need content that aligns with our particular messaging. I can direct AI tools to reference our website, but this doesn't cover everything. What if I am working on content about a new product that isn’t launched yet? Or working on content for a very specific subset of customers?
Initially, I solved this by creating increasingly long prompts packed with relevant messaging and information. But writing comprehensive prompts repeatedly wasn't efficient. By the time I crafted the prompt, generated results, and made necessary human edits, I could have written the content myself.
Since one of my main goals with AI is to reduce repetitive tasks, I figured there had to be a better way.
The Solution: Creating Company-Specific Resource Docs
To solve for this, I started creating company-specific resource docs. That way I could just attach the doc as a PDF and instruct the AI tool to use that as the foundation for messaging, tone, etc.
I did consider making a Gem/GPT/Project, but I have found that I prefer to use different tools for different use cases. So having a pdf that I can attach to any tool helps me be more efficient but means I’m not limited to just one tool.
Most of my experimentation here has been with specific sub-verticals, so here are the things I included in these resource docs:
Top 5-7 business needs this audience has that are related to our solution
What challenges may be contributing to those needs
Industry research/stats that demonstrate the long-term impact of solving or not solving for those needs
Our internal POV on how you can solve those challenges
Individual personas
Top pain points for each persona
Top 3-4 value props for each persona
Additional explanation for each value prop
Additional market breakdown information that could include:
Sub-groups within this market
Some big names in the market that can be a good touchpoint for AI
Publicly available customer logos/customer stories that align with this market
That feels like a long list looking at it, but most of the ones I’ve made are 3-5 slides long in Google Slides!
The Results: What Worked and What Didn’t
As with so many AI-related things, nuance was key here. I’ve attempted a few different versions of this, but everything I’ve done is so contextual that I would just say to take any advice here with a grain of salt.
That being said….
What Worked?
This experiment definitely helped me achieve my goal of being able to create content faster while ensuring it aligned with the overall messaging. It was easy to just attach the document and prompt the AI to base the results on the document. I’d recommend adding in something like:
Based on the industry and persona information in the attached document, please [task description].
What Didn’t Work?
I found that there is a sweet spot for the amount of information to put in these resource documents. Too little information, and the AI doesn’t have enough to work off of. Too much information, and it seems to focus in on details that I considered to be more of a side dish and not the main entree.
A helpful question to ask yourself is, “would I be happy if the AI really leaned into this information?” If the answer is no, don’t include it.
Things To Watch Out For
While the use cases are contextual, here are a few things I learned in my experimentation:
Make sure the rest of your prompt is still specific. Something like “write a sales email based on this document” still won’t get you the best results. Give the tool the same context that you would give a new hire who was doing this task for the first time.
Avoid layers of AI resources. By that, I mean don’t dump a bunch of documents into AI and have it create the resource doc and then use that resource doc without checking it. A concise, curated resource doc will generate better results.
Avoid numbers. For some reason I’ve found if the resource doc has percentages or numbers, the AI tools use that as inspiration to make up other numbers instead. It did that even when I prompted it to not do that, so as of right now, avoiding numbers seems to be the best option.
Conclusion
The reality is that AI content creation isn't a set-it-and-forget-it solution, but with the right foundation in place, it can genuinely speed up your workflow while maintaining quality. These resource docs won't work for every situation, and you'll likely need to tweak the approach based on your specific needs and tools.
BUT if you're tired of writing the same context-heavy prompts repeatedly or getting AI output that misses the mark, this method is worth experimenting with. Start small with one key audience or product area, see what works, and iterate from there. The goal isn't perfection. It's finding a sustainable way to create better content faster.