(This article is part of our AI Essentials for Advisers series on LinkedIn — reposted here for easy access.)
If you’ve blinked in the last few weeks, you might have missed yet another wave of AI updates. ChatGPT keeps evolving, new AI models (like DeepSeek) are making waves, and dedicated AI apps for writing, summarising, and content creation are popping up everywhere.
But with so many options, how do they compare? More importantly, how do they handle real tasks that advisers actually need-like creating client-friendly content?
To find out, we ran an experiment with a mix of established AI tools and two very new ones:
ChatGPT 4o (widely used, known for conversational fluency)
Claude (great at structured responses and clarity)
Perplexity (adds research-based insights)
Qwen (a newer upgrade, impressive reasoning from Alibaba)
DeepSeek (a controversial new Chinese model making headlines)
Prompting vs. Refining-Why It Matters More Than Ever
I’ve said before (and many others have too) that good prompting is key when using AI. The right approach, structure, and effort in your input make a big difference in output quality.
That still holds true, but here’s something that’s becoming just as important-refining the AI’s response.
These tools are improving their reasoning abilities, meaning their first draft might be decent, but their second, third, and refined versions can be dramatically better. Learning to push AI to self-critique and improve its output is becoming one of the most valuable skills in leveraging AI effectively.
Putting AI to Work: A Real-Life Use Case
Imagine you’re crafting an easy-to-understand explanation of trauma insurance for clients. You need it to be clear, engaging, and informative-whether for an email, a flyer, or website content.
We ran this task through these five AI tools to compare their results. Here’s the prompt:
I’m a financial adviser in Australia, and I’d like a simple, client-friendly explanation around the value of trauma insurance, what sorts of things it covers, (regarding policy coverage), and the reasons you might want to have the cover (medical costs beyond whats covered, recovery time, lost earnings, covering short-term debt or repayments etc).
Round 1: First AI Drafts-A Mixed Bag
Each AI tool had a unique take:
- DeepSeek structured its response with clear headings and bullet points.
- Qwen added a real-life example, making it more relatable.
- ChatGPT used a friendly, engaging tone (and even emojis).
- Claude kept things concise but included technical elements like product disclosures.
- Perplexity backed its content with research and statistics.
These differences highlight an important reality-not all AI tools are the same. Some are great at storytelling, others excel at technical details.
Although these models vary in output – they’re all good!
Round 2: The Refinement Test (Or, AI Debating Itself)
Instead of picking one winner, we took things further.
We asked each AI model to:
1️⃣ Review the responses from all five models
2️⃣ Decide which was the best
3️⃣ Create an improved version, incorporating the strongest elements
Now, if AI had egos, this would have been an argument. Each model thought its response was the best.
However, when pressed to refine its own answer while incorporating insights from the others, the results got significantly better.
Consider yours and other responses attached, please: 1 – tell me which seems the best output, and 2 – provide me an improved version, with yours (Claude/Qwen etc) as the base, with any valuable improvements or expansions from the other platforms.
🔹 The takeaway? AI is great on the first pass, but the real magic happens in the refinement stage. Asking follow-up questions, combining insights, and guiding AI to improve its own work makes a huge difference.
Final Thoughts: What This Means for Advisers
You don’t need to run every task through five AI tools-but understanding their strengths helps you choose the right one for the job. And if you’re using AI for content creation, always refine, question, and iterate to get the best results.
🚀 Want to see the final polished version? Click here to get a formatted copy. (We may even offer a branded version for premium members!)
AI Is a Tool-Use It Wisely
A final note: AI tools are helpful, but be mindful of data privacy. Never share personal or sensitive client information unless you’re certain about data security.
As AI keeps advancing, the key takeaway remains the same: Experiment, refine, and make AI work for you-not the other way around.
Best regards,
Marc Fabris
Risk Hub Founder
