BLOG

Learning by Doing: Building My First AI Framework

How I Went from Organizing Notes to Optimizing SEO with AI

I’ll be honest – I didn’t start using AI because I had some grand vision of transforming public health work. I just needed help organizing my notes so they made more sense and flowed better. That’s it. Simple request, and it worked pretty well.

Then HLN Consulting asked me to handle SEO (Search Engine Optimization) for our website. SEO helps your website content appear higher in search results, like Google or Bing, by using the right keywords and optimizing pages for searches. 

We decided on RankMath, a WordPress plugin that handles SEO metadata, that includes all the SEO data and focus keywords that are needed to be specified for each page on the website. I figured, why not ask AI to generate the metadata? Seemed logical enough.

Here’s where things got interesting – and frustrating. The AI gave me perfectly formatted SEO suggestions, but they were completely generic. Think about it: public health informatics is incredibly niche. We’re not competing for searches like “best restaurants” or “how to lose weight.” We’re targeting IIS managers, disease surveillance coordinators, and public health IT staff. The AI had very little clue about our specialized world.

That frustration changed everything for me. I realized I needed to get better at prompting (commands that tell an AI exactly how to help you) if I wanted results that actually made sense for our field. This led me down the rabbit hole of ChatGPT and Perplexity prompting techniques, which eventually got me to sign up for a weekend AI training summit.

The Weekend That Changed My Approach

During that training, I learned something crucial: there’s a huge difference between asking AI questions like you’re doing a Google search versus giving AI a specific job to do. Instead of saying “help me with SEO,” I learned to say “You are an SEO optimization expert with proven experience helping public health organizations improve search visibility.”

This role-based approach was a game changer. Suddenly, the AI understood what I wanted. Instead of suggesting we target “healthcare” (way too broad and competitive), it started recommending terms like “immunization information systems” and “disease surveillance modernization.” Now we were talking.

Testing It Out on Something Personal

My first real test case was completely unrelated to work – I needed to create wedding ceremony programs for my nephew’s Hindu-Christian wedding. I used Claude to generate step-by-step explanations for both ceremonies that guests could follow along with.

The AI did a great job providing structured content, but here’s what I learned: you can’t just copy and paste what AI gives you. It’s a starting point, not a finished product. We had to customize everything with our specific family details and preferences.

This taught me something fundamental that applies to all our public health work: AI gives you the framework, but human expertise provides the substance and accuracy. Once we provide the information and knowledge from our professional experience, AI can help format our thoughts to flow well, grab the reader’s attention, and present information systematically on the topics we’re covering. It’s like having a skilled editor who helps organize and polish your expertise, but the core content and technical accuracy still come from you.

What Actually Surprised Me

The biggest surprise wasn’t that AI could write – it was how flexible it could be compared to Google searches. With Google, I get links to explore. With AI, I can specify exactly how I want information presented, what tone to use, and even ask for multiple versions until something clicks.

For our website work, this meant I could ask for SEO recommendations, then immediately follow up with “make this more technical for public health informatics professionals” or “create a version that speaks to state health department decision-makers.”

The First Pushback I Encountered

But here’s where reality hit. When I started suggesting AI-generated improvements to our blog content – not just SEO optimization, but actual content edits to help people find our work – I met resistance from people. The feedback was that it “sounded too AI generated” and “not written by humans.”

This was my first real lesson about AI acceptance in public health. People were worried about three things: originality, accuracy, and authenticity. In our field, these aren’t just preferences – they are professional requirements. Public health professionals are trained to be skeptical of information sources, and rightly so.

My Basic Framework Emerged

By the end of these early experiments, I had a simple approach:

1. Define the AI’s role clearly
2. Specify the tone and style of writing
3. Add specific instructions to the LLM (Large Language Model which is an AI system trained on vast amounts of text data to understand and generate human-like language for a wide range of tasks, example ChatGPT, Gemini, Claude, etc.) profile that will be used as a standard for all outputs – this ensures consistency across every interaction
4. Provide specific context about public health audiences
5. Request structured outputs with clear formatting
6. Plan to edit and customize everything
7. Test results against our actual needs
8. Create reusable prompt templates – Document successful prompts that work for recurring tasks like SEO optimization, content summarization, and document review.

This foundation prepared me for more systematic applications.

Try This Yourself!

If you’re curious about trying AI in your public health work, here’s how to start safely:

Your First Steps:

1. Pick one simple task – Choose something low-risk like organizing notes, drafting emails, or summarizing meeting points
2. Choose one beginner-friendly tool from the options below
3. Set up basic guidelines – Use our AI Profile Instructions to ensure professional, and accurate outputs
4. Document your experience – Keep notes on what prompts work and what doesn’t