How to Personalise ChatGPT or Copilot Into Your True AI Assistant

AI is powerful, but out of the box it’s just… generic. In this reflection episode of the Dumb Monkey Podcast, Aamir Qutub explains how to personalise ChatGPT and Microsoft Copilot so they understand your preferences, tone, and guardrails. By teaching your AI assistant who you are — what you like, what you avoid, how you want things phrased — you can transform these tools from basic chatbots into trusted business partners. Whether you want professional reports, cheeky social posts, or clear internal updates, AI can adapt to your style. This is how leaders save time, reduce misfires, and finally get practical value from AI.

Transcript

00:00:04:03 - 00:00:36:23
Aamir
Hey, this is Aamir Qutub from the Dumb Monkey Show. When we are doing podcasts, we sometimes let the conversations flow naturally, and there are some really important learnings and nuggets from the podcast that we want to focus on an emphasis on and go a bit, deeper to explore those things. So that's why I am doing these mini episodes after every podcast where we explain some of the key concepts and how you can start to implement them within your business.

00:00:37:00 - 00:00:59:05
Aamir
All of these models or systems that you would use have got something called custom instructions. So if you spend 15 minutes refining those custom instructions, and you can give it some custom instruction in terms of whenever you respond back, I want the proof. Don't, hallucinate, provide me factual information. This is the tone and so on. And now copilot has introduced custom instructions as well.

00:00:59:07 - 00:01:18:17
Aamir
So by spending some time and giving it that context, that context stays within its memory, which means you do not have to provide that context again and again in the prompt. That's a great first tip. And the first podcast, we talked about custom instructions and how you can use them to personalize your experience with both ChatGPT or Microsoft Copilot.

00:01:18:19 - 00:01:48:21
Aamir
So the first question is why is those custom instructions or personalization important? So these large language models are actually trained on a huge amount of data. And sometimes that data is rubbish. Sometimes that data is a PhD level research. And it reads from the tone of a lot of different people. So the natural tone of these large language models or AI assistant might sound a bit robotic or written by AI, but also the type of information that it provides.

00:01:48:21 - 00:02:20:02
Aamir
You can sometimes be a bit misleading. So by personalizing your copilot or GPT or any other assistant that you are using, what you can ensure is that it understands who you are, what type of responses do you expect and it can be a better assistant by providing those personal, personalization instructions. And those personalization instructions are then used by ChatGPT or Copilot to provide a response to you every time it's responding to you.

00:02:20:02 - 00:02:38:22
Aamir
Which means every time you don't have to put in a huge prompt telling it who you are and what tone you would like. It would remember that and it would carry it. Like when you are entering a prompt, it would carry the custom instructions and then will provide a response that would be contextual to you and your organization.

00:02:39:03 - 00:03:10:09
Aamir
So setting up personal instructions is very simple and easy. If you are in ChatGPT, you can go to Settings Personalize and add the custom instructions over there. And if you are in copilot again you can click on the three dots, go to Settings and Personalization and add Personal Custom instructions over there. Now you can put in any instructions over there as you like, but here's a proven framework that you can use or at least get started with, and then you can continue to modify it from there.

00:03:10:11 - 00:03:33:08
Aamir
So some of the key headings that you would put in person instructions is, first of all, give your assistant a role so you can use hashtag roles. Now we use hashtags so that it understands that subheading. So you can say hashtag rule and then you can say something like you are copilot. You are, you can give it a name, a personal assistant to America.

00:03:33:08 - 00:03:54:11
Aamir
Tom who CEO of Enterprise Monkey. And give it a context about who that personal assistant is. And then you can use another heading called About Me, where you can provide context about who you are as a person. So I am, you could see of enterprise Monkey as a part of being the CEO. These are some of the my roles and responsibilities.

00:03:54:15 - 00:04:20:04
Aamir
And you can take your job description or your KPIs and queries. You can put it put it over there and provide context about what do you do in your day to day life. The next thing is, you can, do a hashtag about enterprise monkey or which means about your company, so about the name of your company, and then you can, paste a context about your organization.

00:04:20:06 - 00:04:41:10
Aamir
Now, if you are implementing ChatGPT or Copilot across your organization, it's a good practice to have a comment about your company file that everyone can access and use it to add in their custom instructions. After that, you specify hashtag tone, and there you can specify that while talking to me, I would like you to have a casual tone.

00:04:41:10 - 00:05:05:10
Aamir
Keep the answer short or long however you like it, and then you can say when responding to the emails or when doing external correspondence, I would like you to be a bit more formal, be factual and stay on point. So for example, if I'm in Australian business, I would put something like use casual Australian business tone because we don't use overly formal words here in Australia.

00:05:05:10 - 00:05:28:00
Aamir
But depending on who you are and where you are from, you can specify and set that tone. Now another important section that you can put over there is hashtag guardrails. So in guardrails you specify what are the things that you would like it to do and what are the things that you wouldn't like it to do. So for example, you can specify to fact check every information before it provides you to provide a confidence level.

00:05:28:00 - 00:05:52:08
Aamir
If it's not sure, or to ask you follow up questions if it's not sure about your query, never assume and so on. If you use UK English, you can say always use UK English instead of m dashes. Use hyphen or commas if you don't like m Russia. I love m dashes, but that's my personal choice. And the reason I'm specifying saying instead of m dash is use hyphen or comma.

00:05:52:10 - 00:06:11:19
Aamir
And I'm not saying don't use m dashes because most of these large language models are not good at understanding the negative query. That's like my four year old child. If I tell him something, don't do it, don't do it. He would eventually end up doing it. So we have to redirect the kids and say, hey, instead of doing this, how about we do this?

00:06:11:19 - 00:06:41:07
Aamir
So similarly with your large language model, instead of giving negative instructions, you say, instead of doing this, how about we do this? Or instead of using m dashes, use hyphens and commas. Now you can put more sections or headings over there like examples and examples. You can specify some of the examples of the responses that you would like, or you can specify the examples of how you write your emails as well, but it also learns from your conversation.

00:06:41:09 - 00:07:02:00
Aamir
So I would say this is a good starting point. Now I have got a video explaining how to do it properly, and you can use ChatGPT or Copilot itself to actually help you create those instructions. Then that's something that I have specified in that video that I have linked with this episode. I'll be back with some more insights and I'll see you in episode two.