Most people who say they use AI are really using it the same way they'd use a smarter Google — paste a question, read the answer, move on. That's not wrong, but it's also not the point. Growth Amplified Session 21 brought in Josefa, an AI trainer who has worked with over 1.2 lakh students and professionals, to show a room of 1,200 mutual fund distributors what actually using AI looks like in practice.
Here's what came out of it.
The One Skill That Makes Everything Else Work
Before any tool, before any feature — there is prompting. Every AI tool in the world, whether it's ChatGPT, Gemini, or Perplexity, is only as useful as the instruction you give it. The session introduced a clean three-part framework for this called PET: Persona, Environment, Task.
Persona means telling the AI what role to play. Not just "help me write a post" but "you are a mutual fund distributor with 15 years of experience helping everyday investors." That framing shifts the entire quality of the response. The AI stops being a search box and starts functioning like a specialist.
Environment is your context — the story behind the request. Who are your investors? What are their concerns? What are you trying to solve? The more detail you put in, the better the output. Think of it the way a good standup comedian tells a story: every specific detail makes it land better.
Task is the actual instruction, but specific. Not "write a post about the budget." Rather: "Analyze the key investor implications of Budget 2026 in simple language, add a confidence score to each point, take your time before answering, and format it as a LinkedIn post." That last instruction — "take your time" — actually causes the model to reason more carefully before responding. It sounds strange but it works.
One more trick worth knowing: always add a confidence score to research-heavy prompts. ChatGPT can and does fabricate information. A confidence score of 70% is a signal to verify independently. Above 90% with web search enabled is more reliable.
The Right Tool for the Right Job
This is where most people go wrong — treating ChatGPT as the answer to everything.
ChatGPT with web search enabled is solid for writing, editing, creating social posts, fixing grammar, and generating content. But its default mode has a knowledge cutoff of roughly six months. Always enable web search manually — it turns off automatically and has to be switched on each time under the "+" menu.
Perplexity is the better choice for research. It pulls from 15+ live web sources before answering, cites every claim, and gives you verifiable data. If you're writing about a market event, a budget, or anything current — use Perplexity first, then bring that content into ChatGPT to polish and format.
Gemini works well for Google-integrated tasks and reasonable image generation, though the image quality is not as sharp as dedicated tools.
NotebookLM: The Most Underrated Tool in the Room
Google's NotebookLM lets you upload up to 50 documents — fund factsheets, policy documents, research PDFs — and then interact with them in multiple ways. You can chat with the document, ask specific questions, and get answers drawn only from what you uploaded.
But the feature that genuinely surprised the room was audio overview: the tool converts any uploaded document into a natural two-person podcast-style conversation. Upload a fund presentation, and within minutes you have an audio explanation that sounds like a real discussion. That audio can then be translated into Hindi or regional languages using a tool called Sarvam AI and shared directly with clients on WhatsApp.
NotebookLM also generates slide decks, video summaries, mind maps, and study flashcards — all free, all from the same document.
Putting It Together
The practical workflow Josefa demonstrated was this: use Perplexity to get current, sourced information on a topic. Bring that into ChatGPT with a detailed PET prompt to generate a polished LinkedIn post. Use Flow by Google to create an AI image to accompany it. The whole process takes around 15 to 20 minutes.
None of this requires a designer, a copywriter, or a content agency. It just requires learning the right order of operations — and being willing to experiment beyond the first tool you heard about.
Best regards,
Written By Samyak Naik
