Loading summary
Transcript55 lines
- [00:01]
A
Hey there, it's Mike Stelzner. I was just in downtown Franklin having coffee with Michael Hyatt and we were actually talking about AI Business World and how excited he is to be attending. And he's not just going to show up and teach, he's going to be hanging out the entire time. And he's really excited to be hanging out with a thousand other marketers who are committed to come and to learn. Because the truth of the matter is that AI is taking the marketing world by storm. And, and it's an incredible competitive advantage and doesn't matter whether you're 70 years old like Michael, or in your 50s like me, or in your 20s and just getting started, I want to encourage you to come and experience the best training you will ever get in your professional career, because I've recruited the absolute best of the best. It's taking place in the Anaheim Convention center, right next to Disneyland on April 29th and 30th. Grab your tickets now and commit to taking your AI marketing to the next level. Go to AIbusinessworld live and get your tickets today. I can't wait to meet you there.
- [01:12]
B
Welcome to the AI Explored Podcast, helping you put AI to work. And now, here's your host, Michael Stelzner.
- [01:23]
A
Hello, hello, hello. Thank you so much for joining me for the AI Explored podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzner. This is the podcast for marketers, creators and business owners who want to know how to put AI to work. Just giving you a heads up, this is one you're going to want to pay special attention to. AI provides access to a gold mine of information for the person who knows how to get it. In today's episode of the AI Explored podcast, we're going to explore AI deep research. My special guest is an AI educator who helps small business owners build roles and with AI that reclaim their time. Her course and community is the AI Dream Team. Her AI newsletter is the Future Thread. Natalie MacNeil, welcome to the show. How are you doing today?
- [02:09]
B
I'm doing so well. Thank you for having me, Michael. I'm really excited about this.
- [02:13]
A
Super awesome to have you here. Let's start with your story. How'd you get into AI?
- [02:17]
B
Well, I feel like I've always been drawn to be at the forefront of things. So 16 years ago I was immersed in 360 degree video when a lot of people didn't even know what that meant.
- [02:31]
A
I remember that.
- [02:33]
B
Uh huh. And I. I feel like I have spent my career placing bets on things before they enter the zeitgeist. So sometimes that's as an early investor, sometimes that's as a creative. And I'm actively always experimenting with what's coming. And then that positions me in a way to lead when something actually enters the mainstream and reaches that tipping point. So with AI, I started investing in AI back in 2017 and 2018. There were some really big breakthroughs that were happening at that time and I could sense that we were getting close to having that moment. Usually in technology there's this moment where something has a really beautiful user experience and an interface that makes it accessible to more people. And I knew that, that mom was coming for AI, so I was studying it, I was getting really fascinated by it. And then 2022 came and everything started accelerating. And I first started experimenting with Jasper first and that predated ChatGPT and then ChatGPT when it launched. And then I became an early adopter of Claude in 2023 and I started to look at how I could push the limits. That's something I love to do. And I started to think about how it would be able to transform how I work and then how we work, how we create, how we think. And I started to see where I could implement it in my business so that I could not only create leverage and get more done, but to also get more of my time back. And that's something that I'm a big advocate for, like having a deep learning, human, AI powered business. So how can I have AI doing things for me so that I focus on my personal genius and the human parts of my business, which I love, and then have AI do the rest and help me do the rest really, really well.
- [04:49]
A
Excellent. You have a community, you talk to a lot of people about AI. What are some of the biggest misunderstandings or misconceptions that you seem to be hearing quite often when it comes to AI right now?
- [05:00]
B
The biggest misconception is that AI can't handle really complex work and tasks. So I feel like most people are using it for pretty simple things. And those simple things are usually writing and content, not even always doing content very well either. But like just using it really basically that's how a lot of people are using it. And you can actually have it do so many more complex tasks. It has the ability to do deep research for you, which can amplify your work in so many different ways and help you get your work out to more people, help you grow your business with more ease. And I know that's something we're going to talk about today, diving deeper into deep research. I think the other misconception that people have is around hallucinations and having AI not actually do what you ask it to do. And to me, that all comes down to the workflow. Like if you're getting hallucinations, it's probably that you somewhere along the way have shared contrasting information with it or like contradictory instructions, and you don't have it doing just one very specific thing or step by step things for you, and then it starts to hallucinate. So, so I think these are things to be aware of, but it really comes down to how you end up designing your workflow. The limitation isn't actually AI. The limitation is how we're using it and how we're prompting it.
- [06:44]
A
I love that and I totally agree with that. And I think about how my natural evolution myself of utilizing AIs in the beginning was just to have it do things that were really simple, that I understood how to do, and now I'm doing exceptionally complex things with it. And, and I think like you said, AI is only as good as the person who has given it the instructions or given it the requests. Right. And we're going to get into this quite a bit more since we're here to talk about AI and research. And it's a phrase that's referred to as deep research in the AI world. Some people might not understand what the word deep means. We'll get into that in a little bit. But when AI research is done, well, what is the advantage, what is the benefits, what is the possibilities? Like, just help people understand what it could do for them.
- [07:30]
B
Well, you're compressing what could be days of research into hours or into minutes. So instead of you having to keep up with everything happening in your industry, every study, every article, what competitors are doing, what's working in your industry, what's not working in your industry? AI can synthesize all of that information and you know, it is exceptional at pattern recognition, right? So it's surfacing the patterns that are actually going to make a difference for you in the work that you do and in the industry that you're in. So you're gaining real competitive advantage when you're using deep research. Like, you can map your entire market. You can look at competitor positioning, you can look at customer feedback, what customers are wanting in the industry that you're in, things like emerging trends as well. How can you stay at the forefront of the work that you're doing? And then I think the other really big advantage to Working with deep research is that you, as a leader, are improving your decision quality. So instead of making decisions based on assumptions or the information that you already have, you're making your decisions based on the latest research and based on what's actually happening in the market right now. So you really open yourself to a new level of strategic foresight that, as a visionary, as a leader, is invaluable to you. Like what becomes possible when you have that, when you have your own research analyst that's going through everything, keeping you updated, looking at where you could be, making tweaks where you can optimize things and grow. And then all of that just helps you to reclaim your mental bandwidth as well, because AI is now doing that heavier lifting. So then where do you want to spend your time? For me, that's spending time on things like vision and strategy and creativity and leadership because of how I'm working with deep research. And that's just scratching the surface of what's possible with deep research.
- [10:01]
A
Well, and I just want to double down on what you said, folks. This, what we're about to talk about today, is one of the most important yet least leveraged capabilities of AI. You've heard the phrase agents. A lot of people listening to this podcast have heard the phrase AI agents or agentic AI. This is actually that. This is actually each of the models creating a special agent whose job is to go out and scour all sorts of sources of information and bring it all together in a very short period of time and to provide you insights that frankly, would just take you thousands of dollars to hire an analyst or that you could do on your own. So it's extremely, extremely valuable. And I want people to pay attention, so I just wanted to double down on that. Okay, so is there anything we need to be talking about before we get into this concept of deep research? Any foundational stuff that kind of people need to understand before we get into the how?
- [11:00]
B
Yes, a lot of people with deep research will get frustrated because they don't get instant results like a lot of us. Especially if you're using AI for something that's quite basic, you can get a response back almost instantly. But deep research doesn't work that way. When you're asking it to do more complex things and you're asking it to go through sometimes thousands of pages of documents and URLs, it can't do that instantly. So you're going to need to have a little bit of patience when you're working with deep research. And I like to start with that like, it does take a little bit of time. And we do live in this culture that's obsessed with getting things right away. But I run deep research processes and tasks that sometimes take like one to three hours because of the massive volume of information that is being accessed. So it doesn't mean it's not working. People start to get worried that it's not working and will sometimes even close the window because they're like, oh, this isn't working. But it is working. It just takes a little bit of time. It's still dramatically faster than doing it manually because you can't go through thousands and thousands of pages and websites and URLs on your own without it taking a very, very long time, like weeks. So it's still dramatically faster than you doing any of this manually. You also need to have access to the paid versions of LLMs. I know that the people who are listening to this podcast are probably all on paid versions, maybe of multiple LLMs anyway. But, but you will need to have a paid version of Claude, ChatGPT or Gemini in order to do deep research. And you may even be limited by how many deep research runs that you can do in a month. So because it's more complex, you are sometimes limited in how many times you can actually do that in a 30 day period. And then I would say the other thing that you have to keep in mind with deep research is that your prompt design needs to be really exceptional. So you need to have a well structured prompt. And with deep research, you also have to make sure you're including in your prompt, step by step instructions. So you need to get more detailed than you will with a lot of other tasks that you're using AI for.
- [13:35]
A
And for folks that are new, the $20 or whatever it costs to get one of these tools is going to more than pay for itself with what we're about to talk about today. So if you're on the free plan and, and you haven't bothered to upgrade to the paid plan of whatever we're talking about today, I think this alone is going to be worth it. Okay, so here's what we know so far. What we know so far is that this is a time consuming process for the AI because it has to go out there and gather up a bunch of information. We also know that you need to have a paid account. So now kind of, where do we begin? Because I'm very excited to learn how I can take advantage of this because I'll be very transparent with you. I do deep research occasionally And I mostly do it when I'm looking to buy a piece of software, you know what I mean? And I want to compare that software up against the competitors and stuff, but I know that I'm just scratching the surface of what's possible here. So how do we actually begin? What do we do next? I guess is really the question.
- [14:30]
B
Yeah, I mean, using it for exactly what you're saying, like any sort of comparative analysis analysis, if you're looking to buy new product, like, definitely use research for that, use deep research for that. And then I think in terms of something that's a little more complex, it's maybe easier to understand through an example. So one of the most powerful examples that I used recently was with new tax legislation and a new federal funding bill. So there was new tax legislation that came out that was going to affect me as a small business owner. And there was a new funding bill that was pretty different than funding bills in the past in that there were a lot of industries that were either having a lot of funding pulled or having a lot of funding that they were going to receive. So what I did was I used deep research mode and I used Gemini for this one. You can do deep research though, in Gemini, in Claude, in ChatGPT. And I did this step by step process of having it go through first the new tax legislation and these documents. Like, if any of you have ever attempted to read legislation, I'm a nerd like that.
- [15:59]
A
I've never done that.
- [16:01]
B
I've read legislation in the past. These are documents that are sometimes thousands of pages long with all the notes and little amendments and things like that. So I had deep research mode go through the entire new tax legislation, over a thousand pages. I had it analyze everything that was in the bill and then produce a report. So next step, producing a report for me explaining how the tax changes were going to affect my particular small business entity, which is sure, a lot of you are in the US an S Corp entity in the state of California. And then I ran another research sequence that was focused more on the funding piece. So now I had the tax information, things I needed to ask my accountant, things that I needed to personally know about what was going to be changing and how best to relook at how I run my business. And then, then I had it go through the federal funding bill. And I approached that more from the perspective of an investor. So I asked the AI to identify things like which industries were going to be receiving the most funding, which were going to lose funding, also which sectors would see the largest capital investments, which companies were also most likely to benefit from those funding flows. Then I had IT generate a list of the companies that may be beneficiaries of all these changes to funding. And I had it go through each company individually and run a full financial analysis of each of those companies and score it in terms of growth potential and ways that I would potentially be able to make investments that had very high roi. And then it gave me a full report of its recommendations on which might be really good for investments. And I was able to make investment decisions based on all of that research. Now this process took about three hours for it to go through because again, it's complex, it's multi step. I asked it to do a lot step by step in order. But what I was able to get from that, I could not have done all of that on my own. It would have been so challenging. And so that allowed me to rebalance my portfolio, make some investments as an investor, help me to structure my business as well. Another example recently is Meta Ads changed the way that ads work on the platform, on Facebook and on Instagram and we do paid traffic in our business business. So we knew that changes were coming. And I had deep research mode actually go through and research all the changes and everything that would start working and everything that was going to break with the new updates and how to actually approach it. And then it generated a new 30 day strategy for us to adapt all of our paid traffic campaigns for the new updates, which allowed us to get ahead of what was coming and be really successful in how we've been able to convert after all of the updates that were made. And again, I couldn't have done that without deep research because it can go through such massive amounts of information and tell you exactly what you need to know. And then I get to make data driven decisions from the reports that I get.
- [20:01]
A
Love it. Okay, so this was the Meta Ads Andromeda change, right, that they had made just because I know we were talking about this offline. Also, the federal tax funding bill is a fascinating example. You've heard me talking about AI business world for a little while and I know some of you that are listening probably don't have the budget to travel to Anaheim, California. And I want to let you know that AI business world is a subset of social media marketing world and we happen to have something called a virtual ticket. And with the virtual ticket to social media marketing world, you get all of the content from our AI business world content live streamed. So we're Talking more than 20 AI sessions, multiple main keynotes, including myself. But you also get the recordings of all the other content that's taking place at Social Media Marketing World. So if you really, really deeply want to learn, but you just cannot afford to travel, grab your virtual ticket to Social Media Marketing World by going to AIbusinessWorld live. Go to the register page, you will see the virtual ticket option there and you have a chance to experience some of the best training that you'll ever experience. It'll be live and it'll be recorded again. AIbusinessworld live. So we've got these two great examples. Let's also talk about some other examples. We already mentioned that you could, if you're about to invest in a maybe piece of online software, you could do a competitive analysis. Right. To see kind of the pros and cons of particular software. What are some of the things? Because I know we briefly mentioned as we were riding into this, some other options, but I want to spend a little bit more time going through some of the common things you recommend people consider doing if they're not looking to do something as complicated as what you just mentioned.
- [21:48]
B
Yeah, so anything having to do do with competitor analysis, industry analysis. So what's happening in your industry right now? And can you use Deep Research mode to analyze how your industry might change? Because we're moving through a time of radical shifts in business with AI and with other emerging technologies. So how does that affect your business? Can you give Deep Research your offer and who you serve and more about your industry and have it run a full analysis of how your industry will be affected over the next few years or affected by AI and where the opportunities live for you? What's the opportunity for you to expand your offer or make some tweaks to your offer so that it is future proofed. How can you look at what's working really well in your industry and run a full competitive analysis with Deep Research so that you get a sense of where you should be positioning yourself? So it's really great with things like that. Also Buyer psychology as well, providing you with a deep research report on buyer psychology in your space and how that might be shifting. Because a lot of these changes that are happening in our industries, they are shifting like AI is shifting how people buy as well. And there are new buyer behaviors, new buyer psychology to be aware of. So Deep Research could give you reports on that. And then as an interesting side note to consider, and this goes back to what you were saying, Michael, about how you use, you've used Deep Research to do a product analysis and defined what product option might be the best one for you and your business. So in that sense, an agent is supporting you in the buying, in the purchase decision. And this is something we all need to be aware of as entrepreneurs right now, because in the coming months and years, we're going to have to be selling our services and our products to an agent before it gets sold to a human.
- [24:22]
A
Because we want to show up in that research, right?
- [24:24]
B
Yes, we want to show up in that research. And it's something to really be aware of. And it's another thing that you can actually use deep research to help you to understand right now in your industry and why that's so important. So this whole conversation, it is very critical that we're each thinking about how we end up selling to an agent. And this is actually part of the Andromeda update as well, because Andromeda is now going out and finding you ideal buyers, finding you the people that it thinks are most likely to buy your product or service. So all of the paid traffic, paid ads, updates, the updates, are partially learning how to make sure that Andromeda is going to choose to show your product to the right person. So that's another thing to just start thinking about. As an entrepreneur, is there anything that
- [25:24]
A
it can do to help marketers? Because we have a lot of marketing people listening. Any thoughts on how marketers could leverage deep research to help them be better at their work outside of just understanding the competition?
- [25:35]
B
Yes, for marketing, I think having IT do deep research on this space that you're in and who you serve, and then going through and actually producing a full marketing report, like make it really comprehensive. What are the gaps that you feel you have right now? I love using deep research to fill in gaps. So you can tell it. With my marketing, this is where I would like to go. This is where I envision going. These are some of the campaign ideas that I've had I want you to research. And it's really important. You're providing a lot of context around this. So you need to provide your. Your business, your business plan, your current marketing plan, your customer profile, all of that information, if it has that it can actually go through and look at where you're at and how you're thinking. And you can ask it to do the research and then help you to fill in gaps in your own knowledge, because there are things that you don't even know, you don't know that are affecting your marketing and affecting your conversion rates. And now you have this genius deep research agent that can go and find the things that you're missing and actually ideate with you and give you new market marketing ideas and then build you a marketing strategy based on the research that it has. So this can work for if you're designing an advertising campaign, have it go and analyze the top performing ad campaigns and then it has all the context and information about your business and it can now generate for you a marketing campaign that is going to be the highest converting campaign in your industry because it's gone out and it's looked at the highest converting campaigns and what they have in common. And because it has that pattern recognition, it can come back to you and be like, hey, I think you've been missing this. You can give it all of your campaigns to analyze as well. You can give it your funnels to analyze and ask it to show you where your gaps are. You will be shocked at how good it is at calling you out on the things that you're missing and helping you to actually fill those gaps.
- [27:57]
A
Apps and folks, as you're listening to Natalie, you might say, well, I can do that right now with Claude or Gemini or Chat GPT without deep research and we'll both say, yes you can. But the difference is it's just going to use its big old database that it has in its brain. And the difference with the deep research is it's going to go out and collect the latest and greatest stuff that might not even be in its training models in its training set because of what we're about to talk about. And we're going to get into how to do this in just a few seconds. But Natalie mentioned that there are three primary tools, ChatGPT, Claude and Gemini. And even Grok, believe it or not, has deep research, which I've used in the past, I have not used recently. Can you just kind of give us a quick high level on the pros and cons as to which ones you prefer to use or do you generally like to use multiple ones? What's your thoughts?
- [28:45]
B
I will often use multiple ones. I mean, as an AI educator, I have subscriptions to so many tools and ll because I'm constantly testing them and keeping up with the model updates. But in terms of the differences, like I do often focus on Claude, Chat, GPT and Gemini since those are the top three that people tend to be using and it is always changing. So the information that I'm going to provide you with is current to us filming this now, although with new model updates and updates to deep research all the time, there's a chance chance that you could be listening to this at a point in the future when things have shifted a little bit. So keep that in mind. But I find that Claude is really exceptional at reasoning and it's really good for complex analysis and really in depth interpretations of longer documents. It has incredible systems thinking and when I want to deeply understand a topic or go through like loads and loads of information, I do often turn to Claude first. ChatGPT is also great at structured thinking. It's good at ideation. I do find that it tends to move faster than the other platforms as well. And again, it depends on what model we're in because they're always being updated. But I increasingly use Chat GPT less and I just find that Claude and Gemini are outperforming right now. Gemini, what I really love about it is that it is integrated into the whole Google ecosystem, which includes YouTube. So when you're doing deep research, it actually will show you the work that it's doing. So it is doing the research on its own terminal as well and going through, and it is going through websites, it's going through reports, documents, like everything that it can find. URLs that you are providing it with. Gemini, I find is a lot better at things like providing it a lot of different URLs to go through if you already have a sense of where you want it to get started with the research. It has integration with YouTube as well. So it's going through through YouTube, which can be an added benefit depending on what you're actually researching.
- [31:16]
A
That's a huge plus by the way, because Gemini is multimodal, which means it can read and it can listen and it can watch, right?
- [31:23]
B
Yes, that's right. And that's a huge advantage depending on what you're doing. And it's I think the strongest for web research, you know, because of the Google ecosystem and really, really good at analytical tasks. So I love Claude for a lot of like marketing and content and like the human sides of my business. I find that for deep research I increasingly use Gemini especially for really complex things.
- [31:54]
A
And we should just explain to everybody that. And I looked while Natalie was talking. On Claude, it's just called research, it's under the plus sign. On Chat GPT it's called deep research, it's under the plus sign and in Gemini it's under tools and it's called deep research. So you have to know where to find it. Does it matter whether we use the thinking models or not? Because some of these have different models of thinking does that really matter? Have you found that to be the case when we're, when the deep research is happening?
- [32:18]
B
I really like the extended thinking models, but I do find that once you've selected deep research, then it just goes into that mode anyway. Like it is going to do what you ask it to do and it's going to go through all the steps. So I'm not sure I've found there to be a really big difference, but I'm often going to extended thinking with a lot of the tasks that I'm doing. Unless it's something super simple.
- [32:49]
A
Yeah. And I would imagine, you know, most of these paid models, you get to choose the model. And I would imagine, generally speaking, if it's really important, I would choose the pro or the. Or the extended thinking options, if it's an option. Okay, so we're not getting into the prompting side of it. I have found that generally it will ask clarifying questions of you before you begin the research. But I would imagine if we do what you're about to talk about, this might not be necessary. Right. Because most people, what do they do? They just say, I'm looking to buy affiliate software, do the research. Right. And it's super broad. And obviously that's not going to result in the same kind of stuff that you've alluded to. So help us understand how to get this right, because on some of these systems, you're only allowed to do so much deep research before you run out of, you know, your tokens or whatever. So we want to get this right out of the first go round, especially because it takes so long. Right. To do it.
- [33:41]
B
Yes. And the, the prompting, the way that you prompt is so key. And of course you can have AI help you to develop a prompt. So I personally have a GPT that is a prompt engineer that will create prompts specifically for the different models that I'm using. So if I'm going to do deep research, I'll get a deep research specific prompt. And then I know that all of you listening and those of you in this community have probably got prompt engineering down. You're giving AI a very clear goal. You're providing a role that it's playing. You're providing context about your business, your audience, the problem that you solve, the offer that you have, the industry that you're in, your strategic goals. And then I'm sure you're also providing a lot of examples and reference points as well. So all of that is really important. And then there are some critical extra things that you need to Think about when you're doing deep research in particular, because it needs step by step instructions.
- [34:55]
A
Wait before you get into the step by step instructions. I don't want to take for granted what you just said because I don't think everybody does what you just said. A lot of us sometimes have developed projects in Claude or we've developed Gems and Google, or we've developed custom GPTs where we provided a lot of important context. Right. Here's who our company background is, here's who our ideal customer profile is. But I don't think many of us have thought about providing that actual data when we're doing deep research. So let's just talk about that a little bit. Like what's the standard data that you would include when you are doing deep research before we get into what you were just about to say.
- [35:32]
B
So I, I have a framework that I created called the 3C's prompt framework. So it's clarity, context and cues. And clarity is about the role that AI is playing and the goal. So in this case AI is playing some type of analyst role or researcher role where you're probably going to set it to be the most elite researcher in your industry. Right. And then the goal, you need to be able to clearly define the goal of the research and the context piece. I do feel like people miss out on the context about their business. So what is the bigger picture that if you were bringing an analyst into your company. Right. I think about the investment. Investment banks have a very, very detailed hiring process to try analysts. So off the top of my head that would actually be an interesting thing for you to go and get context from like let's say Goldman Sachs or Morgan Stanley or like one of these investment banks is hiring an analyst. It'd be interesting for you to go and see what they're looking for.
- [36:43]
A
Oh, what that job description is. I love that.
- [36:45]
B
Yeah. Like when, when someone's hiring an analyst or an analyst for a really big governmental or non governmental association, like a big research association. Right. Like what's the IMF or the World bank or what is like the Federal Reserve looking for when it's looking to hire somebody to do research and analysis. And it'd be interesting for you to pull from those job descriptions and put that into your prompt because then you're giving it more about what your expectations are. And then the, the second C, the context, the context about your business, your audience like AI needs to understand a much bigger picture of where the research is fitting into. Why are you doing the research? What do you Plan on doing with the research, like is it helping you to make a very specific decision? If you are doing the research to make a decision, make sure that AI knows the decision that you're trying to make take so that the research can be really tailored. Same with information about your industry. I would provide your website, your offer, anything that you have, I mean, you may as well provide it because you have the information anyway and it's going to help it to know exactly who it's doing this research for and how it's going to be helpful to you. And then the third C is cues. So cues are reference points and documents and examples. So if you have a lot of documentation for your business on, especially whatever the topic is that you're doing deep research on, like what are documents, that might be really helpful for it. So for example, if you're going to be doing market research, you might want to provide customer feedback that you have. So if you've done surveys, if you have feedback that you've been collecting, that would be really helpful to provide it with as well. Samples of your messaging. If you're having it analyze your funnel and make adjustments there based on the research that it ends up gathering, it would be helpful for it to actually have all of that for you to upload your funnels, your messaging, your landing pages and things like that. So that would be foundational for the prompt. And this can end up looking pretty long. This can end up being like a multiple paragraph, long prompt prompt. And then with deep research, we also need to prompt in stages. So instead of asking one big question, we have to break it down.
- [39:26]
A
Yeah. Before we get into stages, I just want to reiterate what you said, important context to provide your quote unquote research agent here. Right. Which is what we're effectively doing here is like what is their role and what is the goal of the role? Right. You are elite researcher, specializes in blank industry or topic or whatever. Right. And then with the objective of finding, accumulating, whatever, use your words. Right. And then the context, right. Which is, who are you, who is the business, why are you wanting them to do the research, what's the industry, what, whatever context they need, and then any documents you referred to, examples, all that kind of stuff. And you said paragraphs. But I would imagine this could be lots of paragraphs, Right. I mean this could be multiple pages easily, could it not?
- [40:11]
B
It definitely could be. I had a prompt that I created for a client just a few days ago and it was a 12 page.
- [40:20]
A
Love it.
- [40:22]
B
By the time we had Everything in it that needed to be included. So they can get really long depending, right. The more complexity there is in what you're asking it to do, the more clarity it's going to need, the more context it's going to need. And that's just what's necessary for you to get the result that you want and the result that will be really useful for you. Because I mean, some of these things we're talking about, when I'm having it do investment research, right, and I'm finding industries that are about to get huge increases in capital expenditure, what I'm doing is really boosting my portfolio, for example, really boosting my returns. Like, one of these research projects could end up making me a lot of extra money as well. So it's worth taking the time.
- [41:14]
A
I love it. Okay, so let's get back to the steps, because this is also part of the thing that most people have no idea, right? Like, we just assume it knows what to do, but like, help people understand, like, what are the steps that it sounds like we can have it do something exceptionally complex, not something very simple, right?
- [41:29]
B
Yes. So I think one of the biggest mistakes that people make with deep research and working with AI in general is asking one massive question. So you need to think about this more in steps, similar to how you would have a standard operating procedure for an employee. Like, you're not just saying, hey, here are things that you need to do when you're creating a standard operating procedure. You're saying, this is step one. And once you've done that, then you do this, then you do this, then you do this. And that's really important when we're training humans to do work for us as well. So you need to bring. Break the research into steps. So, for example, with the prompt that I had given you earlier that I used for the, the tax and the, the funding bills, step one is reading through the tax legislation. And then step two was producing a report on the implications for a small business owner. And then it had the context for my particular business, where I'm located, the, the information about the business that I run. Then step three, I had it go through all of the funding updates and read through all of that legislation, all of those reports. And then step four, I had it go through and find the companies that would be beneficiaries potentially of the increase in spending. And then step five, I had it do research on each of the companies that would potentially receive more funding. And then step six, I had it generate a report for me on which companies I should further look into and maybe consider investing in. So I'm asking it to produce multiple reports along the way, do research along the way in a step by step order that makes sense. So each step usually is building on the previous one. And that kind of layered approach is going to dramatically improve the quality of the deep research that it's able to do.
- [43:53]
A
I think you hinted at this a little earlier, but you can use AI to help you come up with the steps before you actually do the deep research. Right. Do you want to talk about that a little bit?
- [44:02]
B
It yes. So you can have a conversation with it, Right. You can open ChatGPT and, or Claude or Gemini and you might even want to use the one that you're going to end up using for the deep research. So actually, let's say Gemini, you can say, I want to work with Gemini to do deep research. Here's what I'm thinking of doing the deep research on. Here's more about my business. You know, you can just use the voice functionality. It's one of my favorite ways to prompt. But if you're using voice prompting, make sure you're still covering those three Cs. Make sure you're still keeping a framework in mind for how you're prompting and what you're asking. And then you can say, I would like help crafting a deep research specific prompt for Gemini so that I can get a prompt that's going to help me to get the results I want with the deep research, which is going to help me then make the decision that I'm trying to make right now as an entrepreneur.
- [45:04]
A
And you could probably say, what data should I provide to help the research, quote unquote assistant or whatever, better accomplish its job? And it will probably say, well, here's some things that would provide context that would allow you to do it. So you can do a lot of this prep work by interacting directly with your AI model. And it's going to be so much better. So when you actually have this whole thing together, when you give a super sophisticated prompt like this, do you find it still asks clarifying questions or does it just go and immediately do the research?
- [45:34]
B
Oh, this is a good point to highlight because when I'm working with AI, whether it's filling a role in my business, like the AI team members that I've set up, or I'm having it do deep research, I will almost always include in my prompt. Please ask me any clarifying questions that will help you to. In this case, it could be that will help you to craft A prompt that I'll be able to use for deep research in Gemini, it could be ask me clarifying questions that will help you to understand exactly what I need when you're doing, you know, X task
- [46:13]
A
in my business or even the deep research itself, right?
- [46:15]
B
Yes. Like we should be asking AI in just about everything we're doing to ask us clarifying questions. And I also like it to poke holes in anything I'm doing. So I also almost always will say, and this is actually in my system prompt. So inside Claude, Gemini and Chat GPT you can have a system wide prompt and it's the highest level. So it will always take that and it supersedes the things that you've put into GPTs or projects in Claude or what you're putting into a prompt. And I ask it to always challenge me me and to act as more of a sparring partner. So when I'm prompting, I'm saying, like, ask me the clarifying questions and also challenge me because I don't know what I don't know. And you might be seeing things that I'm missing and I want to know what those things are because that allows me to expand into more of my potential. And like, that's what you're here to help me with. Like you're here to help me to actually grow, go further than I would be able to go on my own. So I want that.
- [47:31]
A
Natalie MacNeil Holy cow. This has been a true gold mine, as I kind of alluded to earlier, of incredible insights that I can't even wait to begin to use myself. If people want to connect with you on the socials, where do you hang out? Or if they want to check out you and possibly join your membership or interact with you in any other kind of way, where do you want to send them?
- [47:53]
B
So come visit me@natalie mcneil.com and from there you'll find some of my free resources, including my free AI Dream Team guide for turning AI into roles that work in your business. And my newsletter is@natalie macneil.com forward/newsletter if you want all my updates on AI and other emerging technologies and what I'm seeing. And then on Instagram. I'm Natalie McNeil and I'd love for you to send me a note or DM if you've listened, if you've had a takeaway from this episode that you're going to go off and implement.
- [48:32]
A
Natalie, thanks again for sharing all your insights with us today.
- [48:35]
B
Thanks for having me, Michael.
- [48:37]
A
If you missed anything, we took all the notes for you over@socialmediaexaminer.com A101. Be sure to follow this show on your favorite podcasting app and if you've been a longtime listener, we would love a review and let your friends know about this show. And do check out my other show, the Social Media Marketing Podcast. This brings us to the end of the AI Explored Podcast. I'm your host. Michael Stelzner will be back with you next week. I hope you make the best out of your day and may AI help you become more successful.
- [49:05]
B
The AI Explored Podcast is a production of Social Media Examiner.
- [49:13]
A
Do you want to go deeper in your understanding of AI? You've been listening to this podcast for a while, but did you know that we have a membership with lots and lots of marketers, entrepreneurs and creators who learn every single month? Every single month we do live meetups. We have professionals coming on who teach, training and this is exclusive content only available in our AI Business Society. If you're ready to begin committing to ongoing development, join the AI Business Society right now by visiting socialmediaexaminer.com AI.