AMA: How to make AI work for you?

This AMA was sparked off by a recent story by Anmol Srivastava, AI’s Best-Kept Secret: It’s Not the Tools, It’s You

The speakers:

Tanuj Bojwani is a tech entrepreneur and board member, Digital Public Goods Alliance. He also headed People + AI. He is the co-author of a book with Nandan Nilikani, The Art of Bitfulness, on how to build a better digital hygiene and habit, and a contributor to Founding Fuel.

Tanuj was in conversation with Anmol Shrivastava, who has been a part of the Founding Fuel team. He's now in stealth mode as an entrepreneur.

21 Takeaways from the AMA

[Read Time: 6 minutes]

Quotes edited for brevity

1. AI tools can help you filter social media DMs and build better digital hygiene

[3:22]

“A lot of people send you ‘Hey, how are you’ kind of DMs, but you know that they're going to ask you for something after that. A hack I tried was deploying a bot on WhatsApp to reply to the DM, say it’s a bot and ask for the purpose of the message.”

2. You never have to start from a blank page again

[5:50]

“You can put your rough notes down, add some text above and below, saying, for example, I'm trying to figure out how to build this thing. I want you to make it into a plan that I can share with my teammates. It really gets that shorthand. The first draft that you get is a tremendous starting point, even though it’s never going be complete or accurate. But there are things that you should be starting from scratch.”

3. The easiest way to start your AI learning journey: Experiment, experiment, experiment. Don't worry about the difference between AI models.

[9:36]  

“Just get out there and every time you have a problem, try doing it on AI. You don't need to stray very far from OpenAI ChatGPT, Claude, Anthropic, or DeepSeek-R1. And if you have the ability to run it locally, then even Llama models [a series of large language models (LLMs) developed by Meta AI], or Qwen models [a set of LLMs created by Alibaba Cloud] are good enough.”

4. Learning what AI is good at: Your search history can give you clues on where to start

[12:48]

“If you have saved your search history, look at it and try to ask, what led me to ask this search question? And you can take that problem to ChatGPT and see what output it gives you. This is familiarization—just you testing the waters of what these models are capable of.

In general, if something is very frequently known, like write a leave letter, it can give a perfect letter.”

5. Examples of good tools

[15:36]

“A surprising amount just happens in ChatGPT, in Claude.”

• Whisper, a speech recognition tool that goes beyond transcribing and puts it into a specific format.

• Granola, for meeting recordings. “The notes are how I like my notes.”

• Meta AI for image generation.

• Zapier. “If there’s something that I should solve with an automation once and for all.”

• Suno for music generation.

6. How to not get overwhelmed by all the information: Goal-based reading

[18:33]

“There is no way for you to digest all the information. A good way of focusing is goal-based reading for stuff that you're trying to do, a project. Try out tools that exist in that vicinity. This gives a solid foundation in understanding their limits and their abilities.”

7. The next gen, in their teens, will not have preconceived notions. But older people need to learn about AI—even if just to recognise scams.

[21:00]

“There is a growing fear that this will make you lose some faculties. But the world will shift; what we learned and were taught how to do things, will no longer be as valuable because some of this stuff can be outsourced to our machines. The young generation will get it, they're born with it. They don't have preconceived notions.

I worry more about the older generation. All those FedEx scams and digital arrest scams—how much worse does it get with AI?

8. How will it help older people?

[23:53]

• People, young or old, need to understand how deep fakes can happen

“If you know somebody around you who doesn't know about this, whether young or old, familiarize them. The next big thing is going to be some scam preying on people who don't understand how voice cloning and deep fakes can happen.”

• Automated tech-support: let ChatGPT do it

“A Google search for how to install antivirus on my computer is how all the call centre scams start. The first ads are these scam websites.”

• Software will become more inclusive.

“They're going to reach a state soon, where you no longer have mass-produced software. This becomes crucial for the older generation, because the software, the devices, everything we make today is not designed for your silver years; they're not designed for your eyesight. They're designed for the 20-35 age group, because those are the people who are making it.

With these tools you can finally make stuff that really works for you.”

9. A good prompt helps, but for a lay user it doesn’t make a difference.

[28:06]

“A good prompt helps, especially if you are doing something multi-step. Having said that, the AI labs understand this is a problem and are starting to make it easier for you. All the voodoo that we used to do, think step by step—they’ve actually gotten rid of it in some sense.

10. A hack for reliability and productivity boost: get AI to write a program.

[30:40]

“For most tasks that you can't figure out how to do, get the AI to write a code that does it for you. And everything about deploying that code, running that code, fixing bugs in that code, the AI can handle.”

11. Ask the AI tool to help refine your prompts  

[32:17]

“Ask the AI what questions would it ask experts on the subject, and probe those questions further.”

12. For better outcomes, give the GPT examples.

[33.13]

“Examples really help. If you want something to look a certain way, instead of trying to figure out what to say, just throw in a few examples. Here is an example of a good output. Here is an example of a presentation I like. Here is an example that I would want. It lets the AI figure out for itself what you actually like. This really works for writing.”

13. Improve the outcome by breaking the task into smaller chunks.

[35:37]

“If the code doesn't work on the first try, break down the task a little bit, and let it do each one at a time. If you get stuck, or if you get an error, copy-paste the error and put it into the coding agent again. It will probably solve it for you.”

14. Free yourself of this identity of ‘I am not a coder’. As you use the GPT, you will learn coding.

[38.29]

“If the latest information that you needed was in Chinese, that wouldn't stop you. You would take the Chinese language .pdf and put it in a LLLM—or Google Translate—and get it in English, and you will try to understand and extract what you need. Code is just another language.”

“You have an expert at your disposal which is reading that code and telling you exactly what it does in plain English. Just by seeing all those lines of codes over and over again, as you're reading plain English, you will learn coding. You should throw any code you get into ChatGPT and ask what it does and how it does it. Eventually you will learn coding.”

15. AI for good: It can be used for mental health as a patient and empathetic listener—without stigma.

[41:43]

“Talk to an AI once about some personal problem. You will find that it is patient and empathetic. The responses are fairly good and helpful.

The fear that it could give bad answers is not unfounded, and I wouldn’t use it to replace a therapist.”

16. You can use AI to improve as a coach

[46:15]

“Maybe with the permission of the person you are coaching, you could upload the transcript and ask it questions like what were some moments of transitions, or where you saw the tone of voice of the person change. It might make you think back on had you caught it or not and coach better.”

17. Authenticity and creativity

[47:20]

“There are two things. One is you write to think, and one is you write to tell what you think. AI is great for the ‘write to tell what you think’. For ‘write to think’ you should be doing it yourself.”

18. Biases in the tools  

[48:37]

“Every tool will come at its pros and cons, and you need to use it in a way that you are solving for that. If we completely outsource our thinking and everything else to it, we’ll surely become a victim of these tools.”

“You have no idea what you're what you're getting into. Having said that, they're still very useful. Just don’t take the output for granted.”

19. Confidentiality and strategic advantage: Is it safe to use ChatGPT? Could the documents we share be used by others?

[50:47]

“OpenAI says they don't look at your chats. DeepSeek says we look at your chats. What they do, we don't know, because it's not like they delete your chats. A lot of your data sitting on their server.

Essentially just get a GPU machine, and run a reasonable-sized model locally where you can have complete control.”

“The new kind of creepy is really creepy. If you give, let's say, your past chat history to OpenAI and ask it to say, What do you know about me? It will say things that are surprisingly accurate, and frame it in the language that you understand also.”

20. Will misinformation increase?

[54:33]

“Misinformation will increase dramatically. A review for a restaurant—with details, photos—used to make you feel, okay, this review I can trust. Today generating those kinds of reviews is easier. You can now generate an entire reddit, sub subreddit, or some online site in which you show thousands of users who are fake in actual conversations with each other. And they can go claim something about a product and you will not know the difference.”

“Whatever AI detection you create, there will be an AI whose job it is to get past that, and they'll be good at it.”

21. Where is AI headed?

[57:59]

“The history of economics shows that things that were once valuable, once they become commoditized, the world just moves on, and makes something else of value….

“Just like today in the age of machine goods, handcrafted things are of much more premium value. Similarly, outputs generated, or things done in a particularly human high touch, high labour way, become that much more of a luxury, and that much more desired, and that much more valuable. So, I don't think we will run out of jobs very soon….

“I believe the curious will survive. The curious people who keep trying will find ways of adding value.”