AI Tools for Science Communicators - Webinar

AI is rapidly changing the digital landscape, and while there are arguably some scary things out there, AI technology is also providing tools that can be helpful in managing day to day tasks. Join digital consultant and AI specialist Claire Byrne for a conversation about AI tools that can support science communicators. Claire will demonstrate key tools she uses in her day-to-day and then open up for questions.

Claire Byrne is a product management consultant who specialises in making AI tools accessible and practical for everyday use. With experience leading product development teams in Australia and New Zealand, she's passionate about helping people discover how AI can genuinely support their work without becoming overwhelmed. Claire works with clients across various industries - including scientific organisations - to implement straightforward AI solutions that make communication and research tasks more efficient. 

Members can log in to watch the recording and view Claire’s supplemental notes. Please email hello@scanz.co.nz if you have any problems logging in.

SCANZ Session Recap - 20 August 2025

From Claire Byrne's presentation to 50+ science communicators (delivered from her office while a robot dog wandered around in the background). 

First Things First: Building New Muscles

I don't profess to be an expert in science, but what I've found is that AI has allowed us all to expand into new areas and build capabilities we didn't have before. That's really what this is about - developing new muscles, not replacing the brilliant work you already do as science communicators.

The Core Shift: From Prompts to Context

The game-changer isn't prompt engineering—it's context engineering. I think of AI like a junior coworker who knows a lot of stuff, is really proactive, but I need to check their work constantly. The more specific context I give them about what I'm trying to achieve, the better outcome I get.

Want to dive deeper? Watch the CRISPE prompting method: https://www.youtube.com/watch?v=4SeVAjVoTuY

Instead of: "Write a blog post about climate change"
Try: "Write a 800-word blog post for potential and current tiny home owners about preparing for climate-related weather events, using my clients' practical, solution-focused tone"

(And yes, I have checked their work constantly - more on how they do that below)

Your AI Toolkit by Use Case

🔬 Research & Knowledge Gathering

My top pick: Perplexity (and I've chosen to pay for quite a few of these)

• It chooses which AI model to use based on your question - so it knows which is best for different tasks

• Lightning-fast search with source citations

• The AI browser is coming - to see what's ahead: https://www.youtube.com/watch?v=KRmDj6RXXMg

• Get started: perplexity.ai

One technique that's been game-changing for me: Instead of just asking for research, I ask it to write the research brief first, then execute it. What I found is that it writes a much better research brief than I could ever write, especially compared to the time I'd dedicate to that task. You can mention the things you find unacceptance, although you’ll find the brief the AI will write for you will include guards against common hallucinations and pitfalls.

Example from my weekend startup exploration: "Write a research brief for understanding the charitable startup market in New Zealand, then execute that research."

Quick reality check: Some people asked about hallucination. Yes, it happens. That's why I always check the work and use multiple techniques to cross-check its work (more on that below).

📚 Knowledge Management & Document Analysis

Primary tool: NotebookLM (free from Google)

• Upload your authoritative documents (PDFs, Google Docs, website links)

• Creates an AI that only draws from YOUR provided sources

• Eliminates hallucination risk for domain-specific queries

• Bonus: Creates podcast-style audio overviews of your content

• Get started: https://notebooklm.google.com/

Try now: Upload 3-5 key documents about your current project and ask specific questions, it will respond drawing from only that knowledge.

🏗️ Creating Digital Twins (This Is A Secret Weapon)

Where I do this: Claude Projects (or ChatGPT Projects or Perplexity Spaces if that's your preference)

This is the fast track to Context Engineering - save yourself time on remembering to write specific prompts by giving a good amount of background information up front.

The key to creating a digital twin is naming it a persona, to give it a good amount of scope to “think”. For you, that might be "science communicator," but it could also be "data scientist" or "senior software engineer" - literally, the world is your oyster. Then, populate

Then I start feeding in the authoritative content I have by adding it to my files - just like briefing a new employee.

The workflow:

1. Create the project with your chosen persona

2. Drop in everything relevant: marketing briefs, company info, style guides

3. Use deep research to gather more context

4. Save all accepted outputs back into the project as it grows stronger

A question that came up: "How do I make it remember what we discussed?"
The tools are getting better at this - Claude literally just added conversation memory this week.
One thing I do when I've worked through a complex issue, is ask it to summarise our conversation, then bake that summary into the project. 

The reason I call it a digital twin is once it's got all that context, you can ask it to do whatever task as a starting point, and it's coming from a much more educated position and this works great for standing up a first draft for me to review and iterate on. 

📊 Data Visualization & Prototyping

Primary tools: Claude Artifacts, ChatGPT Canvas

• Build quick prototypes to test concepts before full development

• Create data visualisations to workshop with stakeholders

• Use "build" or "visualise" as trigger words

Pro tip: Get it to describe in text what to build first, then say “build” or “visualise” —saves time on iterations as it uses less computation power to work in text.

🎨 Image Enhancement (Proceed Thoughtfully)

Primary tool: ChatGPT

• Best for upcycling your own photos/content

• Use reference images to create detailed prompts

• Save successful prompts for consistent styling

Copyright note: As Susie pointed out during our Q&A, NZ law states that you (the person inputting parameters) own copyright of AI-generated work. But I don't have solid information on all the complexities - it's still a bit Wild West, to be honest.

Ethical approach: Only enhance content you own or have rights to use.

Checking Their Work: My Quality Control Methods

Remember, I treat AI like a junior colleague whose work I need to check constantly. Here's how I actually do that:

The "Are You Sure?" Method

Honestly, sometimes I just ask it: "Have a think about what you just gave me. Are you sure that's right?" Then I build on whatever they say back. It's surprisingly effective.

The Rubric Approach 

Use some version of this instruction:

“How would you grade this work? Score this work using your rubric, give actionable feedback and then implement it, till you get 10/10 on all fronts.”

Pro tip: You can target this at specific concerns. Someone asked about hallucination signs - you can tell it "I'm worried about factual accuracy" and it'll create rubrics focused on that.

Cross-AI Auditing

Take what you got from one AI and ask another to critique it. I quite often jump between tools when one's busy anyway, so this fits my workflow.

The Persona Lens

"I'm going to put this in front of my CEO. What would they be thinking?" It gives you targeted feedback through that lens. For you, that might be "What would a senior researcher think of this methodology?"

The Honest Conversation: What Keeps Me Awake at Night

Look, I'm both incredibly excited about this technology and terrified by it. I had to make a decision whether I wanted to be on the edge of it or under it, and I chose to be on the edge. But there are real things we need to talk about.

Environmental impact: Devastated. That would be a nice way of putting how I feel about the energy footprint. It's devastating.

What it's doing to our brains: I watched an interview with Sam Altman recently where he said some users are exercising more brain function than ever, while others are doing the opposite. We're in a very interesting experiment and honestly, as someone raising a son, this is what I'm most wanting to understand the impacts of, as soon as possible.

The privacy reality:

• Claude: Doesn't train on your data by default (why I prefer it)

• ChatGPT: You have to actively opt out of data training

• Gemini: Professes to be safe, but has a proven track record of selling your data

Bonus Tools Worth Exploring

AI Studio by Google 

• Can watch videos for you and produce transcripts automatically - very handy for building knowledge quickly

• Free!

• Try it: https://aistudio.google.com/

Want more tools? 

For those really keen to dive into AI tooling like me, I recommend the lennysnewsletter.com product pass - that's where I got annual subscriptions for a lot of handy tools.

Give It a Go: Building Your AI Muscles

Really, what I'd like you to take away is just to get on there and give it a go. The more you do it, the more you build this capability for yourself. About every other weekend, I have an idea for a new business that I validate using AI - the tooling just keeps getting better.

Your first experiment:

1. Pick one frustrating task from your current work

2. Choose one tool from what I've shown (I'd start with NotebookLM if you have documents to work with)

3. Set up a digital twin with relevant context

4. Ask for a research brief before diving into complex tasks

5. Check the work using one of my quality control methods

My Current Tool Stack (The Quick Reference)

Since someone asked about my top paid recommendation: Perplexity, hands down. It chooses the best model automatically and combines functionality from multiple tools.

Thank you to all SCANZ members and public attendees for such engaging and challenging questions. I hope this helps you to build new muscles and expand into areas you might not have explored before.

Connect with me:
📧 claire @ theelements.co.nz
💼 LinkedIn: Claire Byrne