WEBVTT 00:00:00.000 --> 00:00:03.759 Okay, let's unpack this. Have you ever had that 00:00:03.759 --> 00:00:05.780 feeling, you know, that little pang of disappointment? 00:00:05.980 --> 00:00:07.839 You're using this amazing AI, everyone's calling 00:00:07.839 --> 00:00:10.380 it revolutionary, and the response you get is 00:00:10.380 --> 00:00:14.599 just kind of flat, generic, maybe shallow, or 00:00:14.599 --> 00:00:17.820 it completely misses what you were really aiming 00:00:17.820 --> 00:00:20.760 for strategically. That frustration, feeling 00:00:20.760 --> 00:00:23.239 like you're not quite clicking with the AI, it's 00:00:23.239 --> 00:00:25.879 super common. Oh, absolutely. It really is. And 00:00:25.879 --> 00:00:29.140 I think a lot of that frustration comes from 00:00:29.140 --> 00:00:31.579 a basic misunderstanding. People often think 00:00:31.579 --> 00:00:34.520 there's some secret formula or like a string 00:00:34.520 --> 00:00:37.539 of magic words they need to find to unlock the 00:00:37.539 --> 00:00:39.780 AI. Like a cheat code or something. Exactly. 00:00:40.020 --> 00:00:42.399 And that search, it kind of traps you in the 00:00:42.399 --> 00:00:44.359 cycle of getting vague results back because you're 00:00:44.359 --> 00:00:46.439 not really addressing the core issue. And that 00:00:46.439 --> 00:00:47.799 core issue, that's what we're really digging 00:00:47.799 --> 00:00:49.880 into today, isn't it? Yes. Because here's the 00:00:49.880 --> 00:00:52.320 thing, the really empowering part. Getting great 00:00:52.320 --> 00:00:55.579 results from AI isn't mainly about what specific 00:00:55.579 --> 00:00:57.560 words you write. It's much more fundamental. 00:00:57.759 --> 00:00:59.740 It's about how you think. How can you think? 00:01:00.060 --> 00:01:02.439 It's about translating all that complexity in 00:01:02.439 --> 00:01:05.140 your head, your goals, the context, your assumptions 00:01:05.140 --> 00:01:08.859 into a language the AI can genuinely grasp and 00:01:08.859 --> 00:01:10.980 collaborate with. So this deep dive is basically 00:01:10.980 --> 00:01:13.819 about learning that translation process. Precisely. 00:01:14.159 --> 00:01:16.819 Our mission today is to give you a practical 00:01:16.819 --> 00:01:19.810 roadmap. Not just theory, but actual techniques 00:01:19.810 --> 00:01:23.049 you can use right away to seriously level up 00:01:23.049 --> 00:01:25.930 the quality, the relevance, the strategic value 00:01:25.930 --> 00:01:28.349 of what you get from AI. So moving it beyond 00:01:28.349 --> 00:01:31.329 just a simple task doer. Exactly. Transforming 00:01:31.329 --> 00:01:33.790 it into a real thinking partner. Doesn't matter 00:01:33.790 --> 00:01:36.010 if you're in marketing, consulting, research, 00:01:36.349 --> 00:01:38.629 creating content. This applies across the board. 00:01:38.829 --> 00:01:40.390 OK, so to make that happen, you're saying we 00:01:40.390 --> 00:01:42.989 need a mindset shift. A crucial paradigm shift, 00:01:43.150 --> 00:01:45.549 yeah. We need to stop seeing AI like a vending 00:01:45.549 --> 00:01:47.560 machine. You know, put in a command, get out 00:01:47.560 --> 00:01:49.640 a predictable product. That's not it. OK, not 00:01:49.640 --> 00:01:51.599 a vending machine. Then what? Think of it more 00:01:51.599 --> 00:01:55.920 like a highly knowledgeable, incredibly fast 00:01:55.920 --> 00:01:59.379 junior colleague. But, and this is key, a colleague 00:01:59.379 --> 00:02:01.980 who is completely lacking context unless you 00:02:01.980 --> 00:02:05.280 provide it. Ah, OK. I like that analogy. Yeah. 00:02:05.420 --> 00:02:08.259 Smart intern, maybe? Sort of. Think about briefing 00:02:08.259 --> 00:02:11.319 a human colleague. You wouldn't just dump a task 00:02:11.319 --> 00:02:13.639 on them, right? No, definitely not. You'd give 00:02:13.639 --> 00:02:15.680 background, explain why we're doing it, maybe 00:02:15.680 --> 00:02:18.080 mention things we're assuming. Exactly. Assumptions, 00:02:18.240 --> 00:02:20.900 goals, constraints, what success even looks like. 00:02:21.360 --> 00:02:25.819 AI needs that exact same rich layer of context 00:02:25.819 --> 00:02:28.939 to go from spitting out generic text to providing 00:02:28.939 --> 00:02:31.400 something genuinely useful, maybe even insightful. 00:02:31.840 --> 00:02:35.120 So those prompt generators you see online, they're 00:02:35.120 --> 00:02:36.819 not the whole answer. They can be a starting 00:02:36.819 --> 00:02:38.520 point, like a phrase book when you're traveling, 00:02:38.759 --> 00:02:42.860 but for a real deep conversation. For true collaboration, 00:02:43.360 --> 00:02:45.539 you need fluency. Fluency in the language of 00:02:45.539 --> 00:02:48.139 thought and context. Okay, fluency. I like that. 00:02:48.479 --> 00:02:50.199 So how do we get fluent? This work gets really 00:02:50.199 --> 00:02:51.539 interesting, right? You've got some techniques 00:02:51.539 --> 00:02:54.680 for us. Yes. We've got seven essential practical 00:02:54.680 --> 00:02:57.750 techniques. Think of them as your lessons in 00:02:57.750 --> 00:03:00.590 achieving that fluency, your blueprint for transforming 00:03:00.590 --> 00:03:02.569 how you interact with AI. All right, let's get 00:03:02.569 --> 00:03:04.569 into it. Technique number one. Technique number 00:03:04.569 --> 00:03:07.430 one is transparent reasoning, sharing your mental 00:03:07.430 --> 00:03:09.669 blueprint. Mental blueprint. OK, explain that. 00:03:09.849 --> 00:03:13.409 Well, think about most AI prompts. They often 00:03:13.409 --> 00:03:16.629 start out technically correct, like write a blog 00:03:16.629 --> 00:03:18.770 post about the benefits of remote work for tech 00:03:18.770 --> 00:03:20.930 companies. Seems reasonable enough. It's clear, 00:03:20.930 --> 00:03:24.180 sure, but it's strategically empty. The AI will 00:03:24.180 --> 00:03:25.939 probably give you something perfectly readable, 00:03:26.120 --> 00:03:30.680 maybe even SEO friendly, but ultimately, forgettable. 00:03:30.840 --> 00:03:34.740 Why? Because it lacks a specific point of view, 00:03:34.819 --> 00:03:38.099 a defined audience, a real purpose. So transparent 00:03:38.099 --> 00:03:40.340 reasoning fixes that. It's the way to fix it. 00:03:40.500 --> 00:03:43.639 You explicitly state the why behind your what. 00:03:43.979 --> 00:03:46.159 You're essentially opening up your head and showing 00:03:46.159 --> 00:03:49.340 the AI your entire mental blueprint for the task. 00:03:49.680 --> 00:03:51.419 OK, so what's in this blueprint? What do I need 00:03:51.419 --> 00:03:53.979 to share? Good question. Key components are, 00:03:54.180 --> 00:03:56.939 your core objective, what's the ultimate business 00:03:56.939 --> 00:03:58.879 or personal goal here? Your target audience, 00:03:58.939 --> 00:04:00.500 who is this actually for? What are their pains, 00:04:00.599 --> 00:04:02.020 their motivations? What do they already know? 00:04:02.120 --> 00:04:03.740 Makes sense. What else? Your key assumptions, 00:04:03.939 --> 00:04:06.039 what are you taking as true about the situation 00:04:06.039 --> 00:04:08.419 or the audience? Any constraints and boundaries, 00:04:08.539 --> 00:04:11.400 things like budget, time, brand voice, word count, 00:04:12.020 --> 00:04:14.479 legal stuff? Where are the guardrails? Exactly. 00:04:14.800 --> 00:04:17.819 And finally, crucially, your definition of success. 00:04:17.980 --> 00:04:20.279 What does a great outcome actually look like? 00:04:20.600 --> 00:04:23.019 What are the specific criteria you'll use to 00:04:23.019 --> 00:04:26.160 judge the output? Wow, okay. That's a lot more 00:04:26.160 --> 00:04:28.360 detail than just asking for a blog post. It is. 00:04:28.600 --> 00:04:31.160 But let's revisit that remote work example. With 00:04:31.160 --> 00:04:34.199 transparent reasoning, it transforms. How so? 00:04:34.300 --> 00:04:36.620 You'd say something like, okay, my main goal 00:04:36.620 --> 00:04:39.800 here is to persuade skeptical VPs of engineering 00:04:39.800 --> 00:04:43.180 at, say, Series B startups to seriously consider 00:04:43.180 --> 00:04:45.680 a fully remote model. I know they worry about 00:04:45.680 --> 00:04:47.839 collaboration and engineering velocity. Okay, 00:04:47.959 --> 00:04:50.589 much more specific audience and objective. My 00:04:50.589 --> 00:04:53.149 key assumption is their real objection isn't 00:04:53.149 --> 00:04:55.329 about raw productivity. It's about maintaining 00:04:55.329 --> 00:04:58.050 an innovative culture remotely. Constraints? 00:04:58.509 --> 00:05:01.209 Keep it under 1 ,500 words. Tone should be confident, 00:05:01.350 --> 00:05:04.050 but empathetic. Absolutely no corporate jargon. 00:05:04.189 --> 00:05:06.449 And the success definition. Success for this 00:05:06.449 --> 00:05:09.410 piece means giving these VPs actionable frameworks 00:05:09.410 --> 00:05:12.009 or maybe even short case studies they could immediately 00:05:12.009 --> 00:05:14.550 use or share internally to start building their 00:05:14.550 --> 00:05:17.800 own case for going remote. OK, yeah. I can instantly 00:05:17.800 --> 00:05:20.680 see how that prompt leads to a vastly different, 00:05:20.959 --> 00:05:22.699 much more valuable output than the first one. 00:05:22.819 --> 00:05:25.439 It's a total game changer. You're not just getting 00:05:25.439 --> 00:05:27.800 words back. You're getting output that's already 00:05:27.800 --> 00:05:30.959 aligned with your strategy. It anticipates objections, 00:05:31.620 --> 00:05:34.459 uses the right language. It dramatically cuts 00:05:34.459 --> 00:05:37.279 down your editing time. It moves the AI from 00:05:37.279 --> 00:05:40.639 just doing a task to being a strategic partner. 00:05:40.779 --> 00:05:44.019 That's the essence of it. And what's really fascinating... 00:05:44.079 --> 00:05:46.660 When you force yourself to articulate this blueprint 00:05:46.660 --> 00:05:49.300 for the AI, it often clarifies your own thinking, 00:05:49.639 --> 00:05:52.139 too. How often do we really lay out our own assumptions 00:05:52.139 --> 00:05:54.680 and success criteria that, clearly, even for 00:05:54.680 --> 00:05:56.720 ourselves? That's a really good point. OK, so 00:05:56.720 --> 00:05:58.360 transparent reasoning helps us set the stage 00:05:58.360 --> 00:06:01.579 perfectly. But what if? What if I've seen something 00:06:01.579 --> 00:06:04.019 amazing already out there, like a competitor's 00:06:04.019 --> 00:06:06.220 email that's just brilliant or an image with 00:06:06.220 --> 00:06:08.800 a style I love? How do I get the AI to replicate 00:06:08.800 --> 00:06:11.399 that kind of vibe? Describing it seems hard. 00:06:11.600 --> 00:06:14.459 Ah, yes. That brings us perfectly to our second 00:06:14.459 --> 00:06:17.360 technique. Reverse prompting, deconstructing 00:06:17.360 --> 00:06:19.819 excellence to replicate it. Reverse prompting 00:06:19.819 --> 00:06:22.600 sounds intriguing, like working backwards. Exactly. 00:06:22.759 --> 00:06:25.240 This is your secret weapon when you have an existing 00:06:25.240 --> 00:06:28.459 piece of content, text, visual, whatever that 00:06:28.459 --> 00:06:30.980 you admire, but you find it hard to articulate 00:06:30.980 --> 00:06:33.920 why it works or how to get that same feel from 00:06:33.920 --> 00:06:36.500 scratch. Yeah, describing a tone or aesthetic 00:06:36.500 --> 00:06:39.699 is tough. Incredibly tough. So instead of guessing, 00:06:39.980 --> 00:06:42.160 you leverage the AI's pattern recognition skills. 00:06:42.279 --> 00:06:44.439 You get the AI to tell you what makes it great. 00:06:44.959 --> 00:06:47.620 Okay, how does that work? Let's say for text. 00:06:47.879 --> 00:06:50.360 Like that competitor's email newsletter example. 00:06:50.779 --> 00:06:53.139 Right. First, find your gold standard example. 00:06:53.300 --> 00:06:56.000 Copy the text. Then you instruct the AI. You'd 00:06:56.000 --> 00:06:57.500 say something like, OK, I want you to act as 00:06:57.500 --> 00:06:59.759 a world -class copywriter and analyst. Please 00:06:59.759 --> 00:07:02.100 deconstruct the following email newsletter. Then 00:07:02.100 --> 00:07:04.040 you paste it in. Then what do I ask it to look 00:07:04.040 --> 00:07:06.579 for? You'd ask it to identify specific patterns. 00:07:07.300 --> 00:07:09.939 Analyze its tone of voice, its sentence structure 00:07:09.939 --> 00:07:13.060 and rhythm, any persuasion framework being used 00:07:13.060 --> 00:07:16.800 like AIDA or PAS or storytelling. Who is the 00:07:16.800 --> 00:07:19.459 likely intent? audience and what core copywriting 00:07:19.459 --> 00:07:22.040 principles like social proof or scarcity are 00:07:22.040 --> 00:07:24.240 evident. Wow okay so the AI breaks it down like 00:07:24.240 --> 00:07:27.430 an expert. Exactly. It gives you a detailed analysis. 00:07:27.829 --> 00:07:30.589 But you don't stop there. The next step is crucial. 00:07:30.689 --> 00:07:33.189 What's that? You follow up. Fantastic analysis. 00:07:33.750 --> 00:07:36.089 Now, based only on the patterns you just identified, 00:07:36.629 --> 00:07:40.009 create a detailed copy -paste -ready prompt template. 00:07:40.709 --> 00:07:42.889 This template should allow me to consistently 00:07:42.889 --> 00:07:45.829 replicate this exact style and persuasive approach 00:07:45.829 --> 00:07:49.220 for different email topics in the future. So 00:07:49.220 --> 00:07:52.120 the AI builds the prompt for you based on the 00:07:52.120 --> 00:07:54.600 example? Precisely. It gives you a reusable blueprint 00:07:54.600 --> 00:07:56.959 for that specific style. That's clever. Does 00:07:56.959 --> 00:07:59.819 it work for images, too? Like getting that perfect 00:07:59.819 --> 00:08:02.759 visual style? Absolutely. You'd use a multimodal 00:08:02.759 --> 00:08:05.279 AI, one that can see images like Gemini or the 00:08:05.279 --> 00:08:08.060 ChatGPT app with image uploads. OK, so I upload 00:08:08.060 --> 00:08:10.439 the image I like? Yep. And then you ask for a 00:08:10.439 --> 00:08:13.379 similar deconstruction. Analyze this image in 00:08:13.379 --> 00:08:15.779 detail. I want to create similar visuals. Identify 00:08:15.779 --> 00:08:18.579 the key elements. What's the art style? The composition 00:08:18.579 --> 00:08:21.000 and framing. Describe the color palette, the 00:08:21.000 --> 00:08:23.540 lighting, the overall mood and atmosphere. What 00:08:23.540 --> 00:08:25.939 are the specific subject matter details? So it 00:08:25.939 --> 00:08:28.410 gives me the visual recipe. Kind of. And then, 00:08:28.470 --> 00:08:30.850 just like with text, you ask it to synthesize 00:08:30.850 --> 00:08:34.049 that analysis into a usable prompt. Based on 00:08:34.049 --> 00:08:37.389 your analysis, craft a comprehensive image generation 00:08:37.389 --> 00:08:41.120 prompt for a tool like MidJourney or Deli3. This 00:08:41.120 --> 00:08:43.399 prompt should aim to produce a new image that 00:08:43.399 --> 00:08:45.620 has a very similar visual identity and mood, 00:08:46.080 --> 00:08:48.620 but feature a different subject, let's say, a 00:08:48.620 --> 00:08:51.539 lone astronaut exploring a vibrant alien jungle. 00:08:51.679 --> 00:08:53.879 So cool. Instead of me fumbling around trying 00:08:53.879 --> 00:08:56.879 to describe moody lighting or cinematic composition, 00:08:57.559 --> 00:09:00.600 the AI analyzes the example and gives me the 00:09:00.600 --> 00:09:02.740 exact language to ask for it again. You got it. 00:09:02.779 --> 00:09:05.399 You're leveraging the AI's analytical power to 00:09:05.399 --> 00:09:07.399 essentially get it to teach you its own language 00:09:07.399 --> 00:09:09.840 of style. You get the building blocks of excellence. 00:09:09.669 --> 00:09:11.590 and straight from the source. Okay. Technique 00:09:11.590 --> 00:09:13.690 three. What's next? Technique three is a classic 00:09:13.690 --> 00:09:16.570 but incredibly powerful one. Few shot prompting 00:09:16.570 --> 00:09:18.870 guiding with high quality example. Few shot. 00:09:19.029 --> 00:09:21.110 Meaning you give it a few examples. Exactly that. 00:09:21.149 --> 00:09:23.149 It's about showing the AI what you want, not 00:09:23.149 --> 00:09:26.149 just telling it. You provide two or three specific, 00:09:26.289 --> 00:09:29.190 really high quality examples of the desired output 00:09:29.190 --> 00:09:31.470 directly within the prompt itself. How does that 00:09:31.470 --> 00:09:34.889 help? It acts as incredibly powerful in context 00:09:34.889 --> 00:09:38.470 learning. The AI learns the style, format, tone, 00:09:38.809 --> 00:09:41.389 level of detail, everything, directly from the 00:09:41.389 --> 00:09:43.250 examples you provide in that moment. Can you 00:09:43.250 --> 00:09:45.470 give an example, like before and after? Sure. 00:09:46.250 --> 00:09:48.629 Let's say you want to summarize a complex scientific 00:09:48.629 --> 00:09:52.009 paper for a general audience. A basic prompt, 00:09:52.210 --> 00:09:55.049 a zero -shot prompt might be, summarize this 00:09:55.049 --> 00:09:57.190 complex scientific article into a few bullet 00:09:57.190 --> 00:09:59.129 points for a general audience. Okay, and the 00:09:59.129 --> 00:10:01.559 result would probably be? Accurate. Most likely, 00:10:01.740 --> 00:10:03.580 but probably pretty dry, maybe technical, maybe 00:10:03.580 --> 00:10:06.019 unengaging, just the facts. Right, so how does 00:10:06.019 --> 00:10:08.700 Fewshot change it? With Fewshot, you'd preface 00:10:08.700 --> 00:10:11.759 your request with examples. You'd say, I need 00:10:11.759 --> 00:10:14.120 you to summarize complex scientific articles 00:10:14.120 --> 00:10:16.700 into clear, engaging bullet points for a non 00:10:16.700 --> 00:10:19.320 -expert audience. Please follow the exact style 00:10:19.320 --> 00:10:22.000 and format of these two examples. Example 1. 00:10:22.159 --> 00:10:25.460 The big idea. Scientists found gut bacteria can 00:10:25.460 --> 00:10:28.100 influence mood by talking to brain cells. Think 00:10:28.100 --> 00:10:29.919 hotline from your stomach to your head why it 00:10:29.919 --> 00:10:32.100 matters. Could lead to new psychobiotic treatments 00:10:32.100 --> 00:10:36.299 for anxiety. Example two, the discovery. Researchers 00:10:36.299 --> 00:10:38.940 created see -through solar panels. Future windows 00:10:38.940 --> 00:10:41.539 could generate power. The impact. Imagine your 00:10:41.539 --> 00:10:44.220 office building making clean energy. Now, using 00:10:44.220 --> 00:10:46.639 this exact format and engaging analogy -rich 00:10:46.639 --> 00:10:48.879 style, summarize the following article, and then 00:10:48.879 --> 00:10:51.620 you paste the new article text. Ah, okay. The 00:10:51.620 --> 00:10:53.840 examples act like a mini style guide, right there 00:10:53.840 --> 00:10:57.580 in the prompt. The AI sees the big idea and why 00:10:57.580 --> 00:10:59.940 it matters. It sees the short sentences, the 00:10:59.940 --> 00:11:02.679 analogies. Precisely. It immediately grasps the 00:11:02.679 --> 00:11:04.899 desired level of simplification, the tone, the 00:11:04.899 --> 00:11:07.159 structure. You get much more consistent and effective 00:11:07.159 --> 00:11:09.240 output because you've shown it exactly what good 00:11:09.240 --> 00:11:11.519 looks like in this context. It feels like, well, 00:11:11.559 --> 00:11:13.809 like training wheels, in a good way. It is. And 00:11:13.809 --> 00:11:15.889 if we connect this back to that colleague analogy, 00:11:16.169 --> 00:11:18.549 this is like showing a new team member. Here 00:11:18.549 --> 00:11:20.669 are three reports that were perfect. Make your 00:11:20.669 --> 00:11:23.049 report look and feel exactly like these. You 00:11:23.049 --> 00:11:25.190 don't just tell them, you show them. That makes 00:11:25.190 --> 00:11:27.429 perfect sense. Yeah. Is there a way to make this 00:11:27.429 --> 00:11:30.049 even more powerful, maybe for ongoing tasks? 00:11:30.210 --> 00:11:33.549 Yes. There is an advanced tip. Many AI tools 00:11:33.549 --> 00:11:36.929 now, like ChatGPT Plus or Claude Pro, have features 00:11:36.929 --> 00:11:39.850 where you can upload reference materials or create 00:11:39.850 --> 00:11:43.320 a knowledge base. You could upload say, dozens 00:11:43.320 --> 00:11:46.139 of your best ever marketing emails or blog posts 00:11:46.139 --> 00:11:48.980 into one of these knowledge files, then in your 00:11:48.980 --> 00:11:51.279 prompt, you can just refer the AI to that file. 00:11:51.580 --> 00:11:53.559 Summarize this report in the style and format 00:11:53.559 --> 00:11:55.820 exemplified in my best marketing emails knowledge 00:11:55.820 --> 00:11:59.179 file. It gives the AI a huge library of your 00:11:59.179 --> 00:12:01.220 preferred examples to draw from consistently. 00:12:01.460 --> 00:12:03.200 Brilliant. So you build your own internal style 00:12:03.200 --> 00:12:06.580 guide for the AI. Exactly. OK, moving on to technique 00:12:06.580 --> 00:12:09.720 number four. This one is really clever, I think. 00:12:09.940 --> 00:12:13.080 It's meta prompting. turning the AI into your 00:12:13.080 --> 00:12:15.399 prompting coach. Minute prompting, like prompting 00:12:15.399 --> 00:12:17.940 about prompting. Precisely. Instead of jumping 00:12:17.940 --> 00:12:20.580 straight to asking the AI for the final answer 00:12:20.580 --> 00:12:23.500 or the final piece of content, you use the AI's 00:12:23.500 --> 00:12:25.740 intelligence to help you figure out what the 00:12:25.740 --> 00:12:28.100 best possible prompt would be in the first place. 00:12:28.200 --> 00:12:30.960 Whoa, okay, so you collaborate with the AI on 00:12:30.960 --> 00:12:33.220 crafting the instructions for the AI. Exactly. 00:12:33.440 --> 00:12:36.159 You leverage its vast knowledge to ensure you're 00:12:36.159 --> 00:12:38.279 asking the right questions and providing the 00:12:38.279 --> 00:12:40.539 right information before it even starts the main 00:12:40.539 --> 00:12:42.740 task. Can you walk me through an example that 00:12:42.740 --> 00:12:44.940 sounds a bit abstract? Sure. Let's go back to 00:12:44.940 --> 00:12:47.960 that strategy scenario. Imagine you're a strategy 00:12:47.960 --> 00:12:50.940 lead at a premium fitness apparel brand, maybe 00:12:50.940 --> 00:12:54.259 like Lululemon. You need to create a market entry 00:12:54.259 --> 00:12:57.580 plan for Southeast Asia, focusing on, say, Vietnam, 00:12:57.779 --> 00:13:00.559 Thailand, and Singapore. That's a complex task. 00:13:00.919 --> 00:13:02.879 Definitely. Lots of unknowns. Where would I even 00:13:02.879 --> 00:13:05.059 start prompting for that? That's where meta prompting 00:13:05.059 --> 00:13:07.440 comes in. Your first prompt isn't asking for 00:13:07.440 --> 00:13:09.820 the strategy. It's asking for help designing 00:13:09.820 --> 00:13:12.720 the prompt. OK. So what would I say? You'd say 00:13:12.720 --> 00:13:16.200 something like this. Hi, chat GPT. I'm the strategy 00:13:16.200 --> 00:13:18.700 lead for a premium fitness apparel company similar 00:13:18.700 --> 00:13:21.759 in positioning to Lululemon. My objective is 00:13:21.759 --> 00:13:23.960 to develop a comprehensive market entry strategy 00:13:23.960 --> 00:13:27.220 for Southeast Asia, specifically Vietnam, Thailand, 00:13:27.399 --> 00:13:31.179 and Singapore. This is a complex strategic task. 00:13:31.580 --> 00:13:34.299 To ensure I get the most insightful and actionable 00:13:34.299 --> 00:13:37.080 output from you, I need your help crafting the 00:13:37.080 --> 00:13:40.259 master prompt for this analysis. Please act as 00:13:40.259 --> 00:13:42.600 a world -class management consultant specializing 00:13:42.600 --> 00:13:45.740 in APAC retail expansion. What specific information, 00:13:46.139 --> 00:13:48.620 context, data points, and internal company details 00:13:48.620 --> 00:13:51.440 do you need for me to formulate the absolute 00:13:51.440 --> 00:13:53.860 best, most comprehensive prompt for generating 00:13:53.860 --> 00:13:56.700 the strategy? Ah, I see. You're asking the AI 00:13:56.700 --> 00:13:59.320 what it needs to do its best work. Exactly. The 00:13:59.320 --> 00:14:01.399 AI will then shift into that consultant persona 00:14:01.399 --> 00:14:03.259 and start asking you clarifying questions. It 00:14:03.259 --> 00:14:04.980 will essentially conduct a diagnostic interview. 00:14:05.139 --> 00:14:07.200 What kind of questions would it ask? Things like, 00:14:07.600 --> 00:14:10.360 OK, to help you craft the best prompt, please 00:14:10.360 --> 00:14:13.980 tell me, what is your brand's core value proposition 00:14:13.980 --> 00:14:17.909 and specific target demographic? Or what is your 00:14:17.909 --> 00:14:20.269 preliminary budget range for this market entry? 00:14:20.570 --> 00:14:23.009 Or what are your key success metrics for the 00:14:23.009 --> 00:14:25.529 first two years? Are there any known constraints 00:14:25.529 --> 00:14:27.750 like specific retail partners you must use or 00:14:27.750 --> 00:14:30.129 avoid? So it probes for all the critical details 00:14:30.129 --> 00:14:31.570 I might have forgotten to include initially. 00:14:31.769 --> 00:14:33.870 Precisely. You answer its questions, providing 00:14:33.870 --> 00:14:36.809 all that rich context. Then the AI takes all 00:14:36.809 --> 00:14:38.830 your answers and synthesizes them. And then gives 00:14:38.830 --> 00:14:42.980 me It gives you the final powerful custom -built 00:14:42.980 --> 00:14:46.240 master prompt. It might say, thank you. Based 00:14:46.240 --> 00:14:48.059 on the detailed information you provided, here 00:14:48.059 --> 00:14:50.299 is a comprehensive master prompt designed to 00:14:50.299 --> 00:14:53.179 generate a robust market entry strategy. Please 00:14:53.179 --> 00:14:56.149 use this as your next input. and then it will 00:14:56.149 --> 00:14:59.049 lay out a long detailed prompt like, act as a 00:14:59.049 --> 00:15:01.549 team of senior management consultants from McKinsey 00:15:01.549 --> 00:15:04.250 specializing in consumer retail and APAC market 00:15:04.250 --> 00:15:07.330 entry. Your client is a premium fitness apparel 00:15:07.330 --> 00:15:10.549 brand incorporating details you provided. Develop 00:15:10.549 --> 00:15:13.649 a phased market entry strategy for Vietnam, Thailand, 00:15:13.850 --> 00:15:16.250 and Singapore, considering specific factors you 00:15:16.250 --> 00:15:19.389 discussed. Wow. That's like having a free consultation 00:15:19.389 --> 00:15:21.610 with an expert strategist before you even start 00:15:21.610 --> 00:15:23.330 the main work. Yeah. And make sure you don't 00:15:23.330 --> 00:15:26.250 miss crucial angles. Exactly. It leverages the 00:15:26.250 --> 00:15:28.950 AI as an expert interviewer to get the best information 00:15:28.950 --> 00:15:30.889 out of you, dramatically improving the quality 00:15:30.889 --> 00:15:33.529 of the final output. It prevents those vague 00:15:33.529 --> 00:15:35.669 prompts that lead to vague answers. OK, meta 00:15:35.669 --> 00:15:37.429 prompting. I really like that one. What's technique 00:15:37.429 --> 00:15:39.830 number five? Number five is chain prompting, 00:15:39.929 --> 00:15:42.409 building a skyscraper of thought one floor at 00:15:42.409 --> 00:15:45.720 a time. Skyscraper of thought. Nice image. What 00:15:45.720 --> 00:15:48.559 does it mean in practice? It means that for really 00:15:48.559 --> 00:15:52.279 complex multi -stage tasks like developing a 00:15:52.279 --> 00:15:55.039 full business plan, outlining a book, designing 00:15:55.039 --> 00:15:57.440 a software feature, trying to do it all in one 00:15:57.440 --> 00:16:00.789 single mega prompt is usually inefficient. and 00:16:00.789 --> 00:16:03.649 leads to shallow results. Yeah, I can see the 00:16:03.649 --> 00:16:06.450 AI maybe losing track or glossing over details 00:16:06.450 --> 00:16:10.090 with one giant request. Exactly. Chain prompting 00:16:10.090 --> 00:16:12.889 breaks that massive task down into a logical 00:16:12.889 --> 00:16:15.870 sequence of smaller interconnected prompts. Each 00:16:15.870 --> 00:16:18.389 new prompt builds directly on the AI's response 00:16:18.389 --> 00:16:20.750 to the previous one. So you guide the AI step 00:16:20.750 --> 00:16:24.029 by step. Precisely. You maintain and deepen the 00:16:24.029 --> 00:16:26.669 context at each stage. You're essentially building 00:16:26.669 --> 00:16:29.210 the final output piece by piece, floor by floor, 00:16:29.529 --> 00:16:31.450 ensuring everything connects logically. Can we 00:16:31.450 --> 00:16:33.370 use an example, maybe launching something new? 00:16:33.610 --> 00:16:35.610 Perfect. Let's say you want to use AI to help 00:16:35.610 --> 00:16:38.190 you launch a new podcast called the AI Copilot 00:16:38.190 --> 00:16:40.690 aimed at business professionals. OK, sounds familiar. 00:16:41.009 --> 00:16:43.450 So prompt one wouldn't be write the whole podcast 00:16:43.450 --> 00:16:47.820 plan. No way. Prompt one, ideation. I'm launching 00:16:47.820 --> 00:16:50.080 a podcast in the business and technology space. 00:16:50.360 --> 00:16:53.440 Generate five unique podcast concepts. For each, 00:16:53.539 --> 00:16:56.019 include a target audience, a unique angle or 00:16:56.019 --> 00:16:58.519 hook, and a catchy name. Focus on current AI 00:16:58.519 --> 00:17:00.899 trends relevant to business leaders. OK, gives 00:17:00.899 --> 00:17:04.099 me five ideas. Then what? You pick one. Prompt 00:17:04.099 --> 00:17:07.019 two. Deepening the concept. Okay, I really like 00:17:07.019 --> 00:17:09.940 concept, hashtag three, the AI copilot. Let's 00:17:09.940 --> 00:17:12.720 flesh this out. Develop a detailed listener persona 00:17:12.720 --> 00:17:15.019 for this show. Who are they? What's their job 00:17:15.019 --> 00:17:17.460 title? Industry? Biggest professional challenges 00:17:17.460 --> 00:17:20.119 related to AI? What kind of content do they crave? 00:17:20.240 --> 00:17:22.799 So now we have a target listener. Next. Prompt 00:17:22.799 --> 00:17:25.980 three, content strategy. Excellent persona definition. 00:17:26.500 --> 00:17:28.200 Now, based specifically on this listener's needs 00:17:28.200 --> 00:17:31.039 and challenges, brainstorm 10 potential episode 00:17:31.039 --> 00:17:33.539 titles for the AI co -pilot. Suggest different 00:17:33.539 --> 00:17:35.940 formats, too. Maybe interviews, deep dives, case 00:17:35.940 --> 00:17:38.160 studies, Q &A sessions. Got it. We're building 00:17:38.160 --> 00:17:40.200 the content plan. And the final step in this 00:17:40.200 --> 00:17:43.579 chain. Prompt four, execution writing. Great 00:17:43.579 --> 00:17:46.680 episode ideas. Let's draft the intro script for 00:17:46.680 --> 00:17:49.619 the very first episode. Let's title it, Don't 00:17:49.619 --> 00:17:52.319 Fear the Bot, Making AI Your Strategic Partner. 00:17:52.779 --> 00:17:55.319 The script should be roughly 300 words, immediately 00:17:55.319 --> 00:17:57.799 hook the listener based on their persona, introduce 00:17:57.799 --> 00:18:00.019 the show's mission clearly, and set the right 00:18:00.019 --> 00:18:02.740 tone. I see. Each step builds on the last. The 00:18:02.740 --> 00:18:04.839 persona informs the titles, the title informs 00:18:04.839 --> 00:18:08.599 the script. It creates a much more coherent and 00:18:08.599 --> 00:18:11.500 strategically sound final product than just asking 00:18:11.500 --> 00:18:13.740 for everything. once. Exactly. You're creating 00:18:13.740 --> 00:18:16.619 a scaffolding for the AI's reasoning process. 00:18:16.960 --> 00:18:19.480 It prevents the AI from losing the plot on complex 00:18:19.480 --> 00:18:22.319 tasks. The context carries through and deepens, 00:18:22.440 --> 00:18:24.900 resulting in a far richer, more cohesive output. 00:18:25.019 --> 00:18:27.500 It's not just breaking down tasks. It's guiding 00:18:27.500 --> 00:18:30.079 the AI's thinking along a logical path. That's 00:18:30.079 --> 00:18:32.000 a great way to put it. Guiding the entire reasoning 00:18:32.000 --> 00:18:34.700 process. Okay. Technique six sounds really useful 00:18:34.700 --> 00:18:37.180 for avoiding blind spots. Perspective shifting, 00:18:37.420 --> 00:18:39.420 stress testing your ideas with a virtual boardroom. 00:18:39.579 --> 00:18:42.289 Yes. This one is all about recognizing that our 00:18:42.289 --> 00:18:44.829 own perspective is inherently limited. We all 00:18:44.829 --> 00:18:48.049 have biases and blind spots. This technique uses 00:18:48.049 --> 00:18:51.549 the AI to challenge those. How? By making it 00:18:51.549 --> 00:18:54.390 act like different people. Exactly. You ask the 00:18:54.390 --> 00:18:57.430 AI to adopt different specific personas or roles, 00:18:57.589 --> 00:18:59.490 roles that would naturally critique or question 00:18:59.490 --> 00:19:01.450 your idea from different angles. It's like having 00:19:01.450 --> 00:19:04.309 an on -demand red team or convening a virtual 00:19:04.309 --> 00:19:06.769 board of directors to pick your idea apart. Okay, 00:19:06.769 --> 00:19:08.970 give me an example. Let's say my team is developing 00:19:08.970 --> 00:19:11.849 a new software feature. Maybe an AI tool that 00:19:11.849 --> 00:19:13.910 automatically generates project status reports 00:19:13.910 --> 00:19:16.829 inside a tool like Asana or Trello. Great example. 00:19:16.910 --> 00:19:19.130 You've got the feature concept. Now let's stress 00:19:19.130 --> 00:19:21.109 test it using perspective shifting. So who do 00:19:21.109 --> 00:19:23.490 we ask the AI to be first? Maybe start with the 00:19:23.490 --> 00:19:26.650 money person. Perspective one. Skeptical CFO. 00:19:27.569 --> 00:19:30.500 Okay, AI. I want you act as the chief financial 00:19:30.500 --> 00:19:33.140 officer of our company. You are fiscally conservative 00:19:33.140 --> 00:19:36.420 and highly focused on ROI and risk. Review this 00:19:36.420 --> 00:19:39.279 proposed AI project status report feature. What 00:19:39.279 --> 00:19:41.700 are your primary concerns? Question the projected 00:19:41.700 --> 00:19:44.220 ROI. What are the potential financial risks, 00:19:44.559 --> 00:19:47.319 hidden development costs, or risks of increased 00:19:47.319 --> 00:19:49.200 customer churn if this doesn't work perfectly? 00:19:49.799 --> 00:19:52.539 Be critical. Ooh, tough questions. Okay, who's 00:19:52.539 --> 00:19:55.039 next? Maybe the user. Definitely. Perspective 00:19:55.039 --> 00:19:57.799 2, overworked project manager target user. Right, 00:19:57.859 --> 00:20:00.400 now shift perspective. Act as a busy project 00:20:00.400 --> 00:20:02.660 manager at a fast -paced digital agency, our 00:20:02.660 --> 00:20:05.319 key target user. You're juggling 10 projects, 00:20:05.559 --> 00:20:08.079 constantly in meetings. Realistically, how would 00:20:08.079 --> 00:20:10.619 this AI status report feature fit into your actual 00:20:10.619 --> 00:20:13.119 daily workflow? What are your immediate gut reactions 00:20:13.119 --> 00:20:14.759 or hesitations? What would make you say, wow, 00:20:14.779 --> 00:20:17.299 I desperately need this versus, oh, another tool 00:20:17.299 --> 00:20:19.380 gimmick I don't have time for? Getting the user's 00:20:19.380 --> 00:20:22.059 gut check. Smart. One more. Let's bring in support. 00:20:22.180 --> 00:20:25.220 Perspective 3. Head of customer support. Okay, 00:20:25.539 --> 00:20:27.960 final perspective. Act as our head of customer 00:20:27.960 --> 00:20:30.480 support. When this feature eventually launches, 00:20:31.019 --> 00:20:33.480 what kinds of support tickets, common questions, 00:20:33.660 --> 00:20:36.380 and customer complaints do you anticipate? What 00:20:36.380 --> 00:20:38.500 aspects are likely to be confusing for users 00:20:38.500 --> 00:20:41.460 initially? What training materials or resources 00:20:41.460 --> 00:20:44.000 will your support team need beforehand to handle 00:20:44.000 --> 00:20:46.299 this successfully? This is amazing. It's like 00:20:46.299 --> 00:20:49.099 you can instantly get critiques from finance, 00:20:49.480 --> 00:20:52.410 the actual user, and the support team. all before 00:20:52.410 --> 00:20:54.890 you've even written a line of code. Precisely. 00:20:55.029 --> 00:20:58.650 It forces a 360 -degree analysis. You get financial 00:20:58.650 --> 00:21:01.509 rigor, real -world usability concerns, and practical 00:21:01.509 --> 00:21:03.930 implementation challenges surfaced early. It 00:21:03.930 --> 00:21:06.390 leads to a much more robust, well -thought -out 00:21:06.390 --> 00:21:09.569 strategy or product. Okay, wow. Six powerful 00:21:09.569 --> 00:21:11.759 techniques. What's the seventh and final one? 00:21:11.960 --> 00:21:13.960 The seventh, and arguably one of the most powerful 00:21:13.960 --> 00:21:16.700 and underutilized techniques, is the self -evaluation 00:21:16.700 --> 00:21:19.420 loop forcing critical self -reflection. Self 00:21:19.420 --> 00:21:21.799 -evaluation. You mean asking the AI to critique 00:21:21.799 --> 00:21:24.880 itself? Exactly that. After the AI gives you 00:21:24.880 --> 00:21:27.079 a response, particularly for something complex 00:21:27.079 --> 00:21:29.720 like a strategy analysis or a piece of creative 00:21:29.720 --> 00:21:32.980 writing, you don't just accept it. You push back 00:21:32.980 --> 00:21:35.980 and ask the AI to critically review its own work. 00:21:36.619 --> 00:21:39.589 Oh. Why would I do that? Because it prompts the 00:21:39.589 --> 00:21:42.349 AI to perform a kind of second -pass analysis. 00:21:42.849 --> 00:21:45.670 It forces it to reconsider its initial output, 00:21:46.309 --> 00:21:49.089 identify potential logical gaps, challenge its 00:21:49.089 --> 00:21:51.869 own underlying assumptions, and often find areas 00:21:51.869 --> 00:21:54.150 for significant improvement or innovation that 00:21:54.150 --> 00:21:56.250 it missed the first time around. Interesting. 00:21:56.390 --> 00:21:58.230 Can you give an example of how that works? Let's 00:21:58.230 --> 00:22:00.609 stick with strategy. Imagine you run a meal kit 00:22:00.609 --> 00:22:03.529 delivery service, maybe like HelloFresh or Blue 00:22:03.529 --> 00:22:06.529 Apron. You ask the AI for a competitive analysis. 00:22:06.730 --> 00:22:09.769 So the initial prompt might be, analyze the marketing 00:22:09.769 --> 00:22:11.910 strategies of my top three competitors naming 00:22:11.910 --> 00:22:15.150 them. Based on that analysis, propose a differentiated 00:22:15.150 --> 00:22:17.369 marketing strategy for my brand for the next 00:22:17.369 --> 00:22:20.109 quarter. Right. And the AI gives you a solid, 00:22:20.210 --> 00:22:22.509 maybe decent, but likely somewhat predictable 00:22:22.509 --> 00:22:24.890 strategy based on common marketing tactics. OK, 00:22:25.049 --> 00:22:27.470 a standard response. Now how do I apply the self 00:22:27.470 --> 00:22:29.349 -evaluation loop? You follow up with something 00:22:29.349 --> 00:22:31.569 pointed, like this. Thank you for that initial 00:22:31.569 --> 00:22:34.670 strategy proposal. Now, I want you to critically 00:22:34.670 --> 00:22:37.289 evaluate your own recommendation. Put on your 00:22:37.289 --> 00:22:40.609 most skeptical hat. One, what crucial market 00:22:40.609 --> 00:22:44.109 factor or emerging consumer trend, GG, sustainability, 00:22:44.470 --> 00:22:47.029 specific dietary needs, economic pressures, might 00:22:47.029 --> 00:22:49.269 your strategy have overlooked or underestimated? 00:22:49.549 --> 00:22:52.609 Two, what are the single weakest or most questionable 00:22:52.609 --> 00:22:54.809 assumptions that your proposed strategy is built 00:22:54.809 --> 00:22:58.019 upon? Three, how could this strategy realistically 00:22:58.019 --> 00:23:00.559 backfire, or how could competitors easily copy 00:23:00.559 --> 00:23:03.720 or neutralize it? Four, finally, based on this 00:23:03.720 --> 00:23:06.960 critique, propose a more unconventional or asymmetric 00:23:06.960 --> 00:23:09.059 marketing angle, something unexpected that would 00:23:09.059 --> 00:23:11.380 be harder for competitors to replicate and could 00:23:11.380 --> 00:23:13.420 create a stronger competitive advantage for my 00:23:13.420 --> 00:23:16.519 brand. Whoa, that's asking it to really push 00:23:16.519 --> 00:23:18.279 beyond its first answer. What kind of results 00:23:18.279 --> 00:23:21.190 does that yield? often quite remarkable ones. 00:23:21.670 --> 00:23:24.809 It forces the AI to move beyond simple pattern 00:23:24.809 --> 00:23:27.869 matching or first -order thinking. It might suddenly 00:23:27.869 --> 00:23:30.750 identify the rising food as medicine, trend it 00:23:30.750 --> 00:23:33.089 initially missed, and suggest partnerships with 00:23:33.089 --> 00:23:35.910 nutritionists. It might question its own assumption 00:23:35.910 --> 00:23:37.990 about the primary target audience and suggest 00:23:37.990 --> 00:23:40.289 focusing on a niche demographic. So it leads 00:23:40.289 --> 00:23:42.509 to more resilient, maybe more innovative ideas. 00:23:42.690 --> 00:23:45.549 Exactly. More resilient and often much more innovative 00:23:45.549 --> 00:23:48.369 strategies because you've made the AI become 00:23:48.369 --> 00:23:51.079 its own toughest critic. It's like enabling meta 00:23:51.079 --> 00:23:53.519 cognition for the machine. It's incredibly powerful. 00:23:53.799 --> 00:23:56.720 Seven techniques. Transparent reasoning, reverse 00:23:56.720 --> 00:23:59.119 prompting, few -shot prompting, meta prompting, 00:23:59.279 --> 00:24:01.700 chain prompting, perspective shifting, and the 00:24:01.700 --> 00:24:04.319 self -evaluation loop. It's a fantastic toolkit, 00:24:04.619 --> 00:24:06.839 but just knowing them isn't enough, right? How 00:24:06.839 --> 00:24:08.940 do we actually internalize these and make them 00:24:08.940 --> 00:24:11.039 part of our regular workflow? That's the perfect 00:24:11.039 --> 00:24:13.640 question. It leads us to the final piece. Building 00:24:13.640 --> 00:24:16.119 your personal prompt playbook, systematizing 00:24:16.119 --> 00:24:18.680 your success. A playbook. I like the sound of 00:24:18.680 --> 00:24:21.359 that. It's crucial. Knowing the techniques is 00:24:21.359 --> 00:24:24.019 step one. Making them second nature is step two. 00:24:24.480 --> 00:24:26.839 And a playbook helps you do that. So how do we 00:24:26.839 --> 00:24:28.960 build this playbook? What are the steps? It's 00:24:28.960 --> 00:24:32.160 pretty straightforward. Step one, create your 00:24:32.160 --> 00:24:35.019 central playbook document. Use whatever tool 00:24:35.019 --> 00:24:38.549 works for you. Notion, Google Docs, Obsidian, 00:24:38.809 --> 00:24:41.109 Evernote, even just a simple text file. Okay, 00:24:41.150 --> 00:24:43.410 got my document. What goes in it? For each of 00:24:43.410 --> 00:24:45.630 the seven techniques we just discussed, create 00:24:45.630 --> 00:24:48.750 a heading. Under each heading, briefly summarize 00:24:48.750 --> 00:24:52.609 in your own words. What is this technique? When 00:24:52.609 --> 00:24:54.950 is the best time to use it? And this is key. 00:24:55.349 --> 00:24:57.650 Include a go -to prompt template or starting 00:24:57.650 --> 00:25:00.069 phrase based on your own typical work. make it 00:25:00.069 --> 00:25:02.190 relevant to you. So I personalize the examples. 00:25:02.609 --> 00:25:05.170 Exactly. Step two, configure your AI workspace. 00:25:05.630 --> 00:25:08.569 Many AI tools, especially ChatGPT, have a feature 00:25:08.569 --> 00:25:10.609 often called custom instructions or something 00:25:10.609 --> 00:25:13.349 similar. This is where you embed your core principles. 00:25:13.450 --> 00:25:15.029 How does that work? What do I put there? There 00:25:15.029 --> 00:25:18.029 are usually two parts. In the first part, what 00:25:18.029 --> 00:25:20.589 would you like the AI to know about you to provide 00:25:20.589 --> 00:25:23.930 better responses? Well, you briefly describe 00:25:23.930 --> 00:25:27.519 your role, your industry, your main goals. maybe 00:25:27.519 --> 00:25:30.119 your typical audience, just the core context 00:25:30.119 --> 00:25:32.500 about you. Okay, setting my personal context. 00:25:32.700 --> 00:25:34.579 And in the second part, how would you like the 00:25:34.579 --> 00:25:37.779 AI to respond? Okay, you set your ground rules 00:25:37.779 --> 00:25:39.940 based on the techniques. You might write something 00:25:39.940 --> 00:25:42.819 like, always respond from the perspective of 00:25:42.819 --> 00:25:46.000 a senior strategic advisor in my field. Before 00:25:46.000 --> 00:25:48.880 tackling complex requests, always ask clarifying 00:25:48.880 --> 00:25:51.819 questions to ensure you have full context, reference, 00:25:51.859 --> 00:25:54.640 transparent reasoning. When generating strategic 00:25:54.640 --> 00:25:57.000 options, always consider multiple perspectives 00:25:57.000 --> 00:25:59.660 mentioning perspective shifting. When evaluating 00:25:59.660 --> 00:26:02.480 ideas, apply critical thinking and identify assumptions 00:26:02.480 --> 00:26:05.160 referencing self -evaluation. Strive for depth, 00:26:05.440 --> 00:26:07.859 originality, and actionable insights over generic 00:26:07.859 --> 00:26:10.339 summaries. Wow. So you're essentially preloading 00:26:10.339 --> 00:26:12.700 the AI with your preferred way of working. based 00:26:12.700 --> 00:26:14.740 on these techniques. Precisely. It sets the stage 00:26:14.740 --> 00:26:16.740 for every conversation. And the final step, step 00:26:16.740 --> 00:26:20.119 three, is integrate and iterate. Meaning, actually 00:26:20.119 --> 00:26:23.420 use your playbook. When you start a new complex 00:26:23.420 --> 00:26:26.819 task with the AI, glance at your playbook. Maybe 00:26:26.819 --> 00:26:29.859 start the conversation by saying, OK, AI, referencing 00:26:29.859 --> 00:26:32.000 my custom instructions and playbook, let's use 00:26:32.000 --> 00:26:34.019 the meta -prompting technique to figure out the 00:26:34.019 --> 00:26:36.680 best way to approach analyzing this market research 00:26:36.680 --> 00:26:39.640 report. So you actively reference your system. 00:26:39.839 --> 00:26:43.720 Yes. And crucially, iterate. When you craft a 00:26:43.720 --> 00:26:46.240 prompt, perhaps using chain prompting or few 00:26:46.240 --> 00:26:48.759 shot and it yields exceptionally good results, 00:26:49.339 --> 00:26:52.099 copy that successful prompt structure into your 00:26:52.099 --> 00:26:55.000 playbook under the relevant technique. Your playbook 00:26:55.000 --> 00:26:57.599 becomes a living document constantly refined 00:26:57.599 --> 00:26:59.519 with what works best for you. That's brilliant. 00:26:59.660 --> 00:27:01.759 It's not just about using AI anymore. It's about 00:27:01.759 --> 00:27:04.160 building and optimizing your personal interaction 00:27:04.160 --> 00:27:06.720 system with it, making it truly yours. It becomes 00:27:06.720 --> 00:27:09.759 a personalized extension of the AI's brain tailored 00:27:09.759 --> 00:27:12.680 to your specific needs and goals. OK, so as we 00:27:12.680 --> 00:27:14.319 wrap up this deep dive, let's bring it all together. 00:27:14.400 --> 00:27:16.559 What's the main takeaway here? The core message, 00:27:16.579 --> 00:27:19.359 really, is that mastering AI interaction isn't 00:27:19.359 --> 00:27:21.480 fundamentally a technical skill you need to learn, 00:27:21.619 --> 00:27:23.839 like coding. It's primarily a cognitive skill. 00:27:24.180 --> 00:27:27.289 A thinking skill. Yes. It's about learning to 00:27:27.289 --> 00:27:29.490 communicate your thoughts, your context, your 00:27:29.490 --> 00:27:32.269 intentions with much greater clarity and precision. 00:27:32.650 --> 00:27:35.329 And these seven techniques we covered. Transparent 00:27:35.329 --> 00:27:37.890 reasoning, reverse prompting, few -shot prompting, 00:27:38.069 --> 00:27:40.549 meta prompting, chain prompting, perspective 00:27:40.549 --> 00:27:44.049 shifting, and self -evaluation. They aren't just 00:27:44.049 --> 00:27:46.670 clever tricks. Think of them as powerful frameworks 00:27:46.670 --> 00:27:49.450 for better thinking and better communication 00:27:49.450 --> 00:27:52.220 with your AI partner. So for everyone listening, 00:27:52.480 --> 00:27:54.160 what's the challenge? What's one thing they can 00:27:54.160 --> 00:27:56.039 do this week? Let's make it super actionable. 00:27:56.519 --> 00:27:59.740 Your challenge is this. Pick just one of these 00:27:59.740 --> 00:28:02.619 seven techniques this week. Just one. And consciously 00:28:02.619 --> 00:28:04.759 apply it to a real task you're doing with AI. 00:28:05.220 --> 00:28:07.569 OK, like what? If you're creating any kind of 00:28:07.569 --> 00:28:10.829 content, emails, posts, reports, try few -shot 00:28:10.829 --> 00:28:13.529 prompting. Grab two or three examples of your 00:28:13.529 --> 00:28:15.829 own best work in that format and put them right 00:28:15.829 --> 00:28:18.049 into the prompt before your request. Good one. 00:28:18.150 --> 00:28:20.549 What else? If you're working on any kind of strategy 00:28:20.549 --> 00:28:23.710 or analysis, try the self -evaluation loop. After 00:28:23.710 --> 00:28:26.990 the AI gives you its first take, push back, ask 00:28:26.990 --> 00:28:29.150 it to critique its own assumptions and suggest 00:28:29.150 --> 00:28:32.029 unconventional alternatives. Or maybe if you're 00:28:32.029 --> 00:28:33.809 just starting something new. If you're kicking 00:28:33.809 --> 00:28:36.650 off a new project or tackling a complex problem, 00:28:37.029 --> 00:28:39.750 start with meta -prompting. Ask the AI to help 00:28:39.750 --> 00:28:41.990 you figure out the best way to even ask the question 00:28:41.990 --> 00:28:44.170 in the first place. Just pick one technique and 00:28:44.170 --> 00:28:47.089 try it. Excellent. And the final thought to leave 00:28:47.089 --> 00:28:50.089 people with. It's this. When you start treating 00:28:50.089 --> 00:28:52.809 your AI less like a vending machine and more 00:28:52.809 --> 00:28:56.269 like that smart, context -hungry junior colleague, 00:28:56.970 --> 00:28:59.250 when you brief it thoughtfully using these kinds 00:28:59.250 --> 00:29:02.289 of techniques, you will unlock a level of productivity, 00:29:02.809 --> 00:29:04.950 creativity, and insight that might genuinely 00:29:04.950 --> 00:29:07.619 surprise you. Ultimately, the quality of the 00:29:07.619 --> 00:29:10.160 AI's output will start to mirror the quality 00:29:10.160 --> 00:29:12.039 and clarity of your thinking and communication. 00:29:12.779 --> 00:29:14.900 Every interaction can shift from a simple query 00:29:14.900 --> 00:29:17.920 response to a powerful strategic dialogue. So 00:29:17.920 --> 00:29:20.079 what does this all mean for you listening? It 00:29:20.079 --> 00:29:22.559 means your AI is absolutely ready to become your 00:29:22.559 --> 00:29:24.779 most powerful thinking partner. You just need 00:29:24.779 --> 00:29:27.660 to learn how to have that truly meaningful, context 00:29:27.660 --> 00:29:28.599 -rich conversation.