WEBVTT 00:00:00.000 --> 00:00:02.560 OK, so data analysis usually starts messy, right? 00:00:02.960 --> 00:00:05.820 You've got this mountain of raw data. Yeah, the 00:00:05.820 --> 00:00:08.099 cleaning, the charting, all that tedious checking. 00:00:08.539 --> 00:00:11.919 We know the drill. It just eats up hours, hours 00:00:11.919 --> 00:00:14.740 of your life. And then there's AI promising this 00:00:14.740 --> 00:00:17.339 like instant insight shortcut. But it's tricky 00:00:17.339 --> 00:00:19.460 if you just kind of toss a random spreadsheet 00:00:19.460 --> 00:00:22.399 at a large language model. You usually get fast 00:00:22.399 --> 00:00:25.000 garbage back, don't you? The real power isn't 00:00:25.000 --> 00:00:27.690 just the tool. No. It's having the right map, 00:00:27.730 --> 00:00:30.269 the right approach. That's the actual shortcut. 00:00:30.649 --> 00:00:33.509 Absolutely. So we've gathered this deep stack 00:00:33.509 --> 00:00:35.810 of guides and tutorials for you, the listener. 00:00:36.229 --> 00:00:39.109 Our mission today is really to boil this down 00:00:39.109 --> 00:00:42.409 to two key frameworks. These are the things that 00:00:42.409 --> 00:00:46.329 turn that boring data into, well, useful, actionable 00:00:46.329 --> 00:00:48.850 insights. We're basically simplifying data science 00:00:48.850 --> 00:00:52.149 here. So first up, you get achieve. Right. Achieve 00:00:52.149 --> 00:00:54.469 tells you when AI is actually the best tool for 00:00:54.469 --> 00:00:57.560 your specific job. And second, there's D. That 00:00:57.560 --> 00:01:00.020 gives you the reliable step -by -step process, 00:01:00.119 --> 00:01:02.640 how to do the analysis correctly every single 00:01:02.640 --> 00:01:06.400 time. And you don't need a PhD in computer science 00:01:06.400 --> 00:01:08.500 to get this. Let's dive deep into the sources. 00:01:09.099 --> 00:01:12.640 So achieve. Think of AI as this incredibly fast. 00:01:13.769 --> 00:01:16.510 Very obedient research assistant. Yeah, super 00:01:16.510 --> 00:01:18.989 smart, super fast, but it needs crystal clear 00:01:18.989 --> 00:01:22.849 instructions. Achieve lays out five situations 00:01:22.849 --> 00:01:26.930 where AI really transforms your data work. Starting 00:01:26.930 --> 00:01:30.799 with A, aiding human coordination. Ah, yes. We 00:01:30.799 --> 00:01:32.819 humans, we can be messy collaborators, can't 00:01:32.819 --> 00:01:35.560 we? We really can. AI helps clean up that noise. 00:01:35.900 --> 00:01:39.079 So imagine uploading, say, a dense 30 -minute 00:01:39.079 --> 00:01:41.359 meeting transcript. OK. Instead of listening 00:01:41.359 --> 00:01:43.239 back or reading it all. Exactly. You just ask 00:01:43.239 --> 00:01:45.620 the AI, summarize the decisions, list all the 00:01:45.620 --> 00:01:47.840 action items, and this is key assign responsibility 00:01:47.840 --> 00:01:50.319 and deadlines. Wow. Instant structure from just 00:01:50.319 --> 00:01:53.180 conversation chaos. Pretty much. Or a classic 00:01:53.180 --> 00:01:55.700 business headache. Comparing suppliers. Oh, yeah, 00:01:55.719 --> 00:01:57.560 you've got maybe five different vendor emails 00:01:57.560 --> 00:02:00.180 food sound whatever for an event, right? You 00:02:00.180 --> 00:02:02.799 feed them all in ask for a simple table name 00:02:02.799 --> 00:02:05.859 service price available date boom instant clarity 00:02:05.859 --> 00:02:08.139 that saves what an hour of digging through emails 00:02:08.139 --> 00:02:12.219 and attachments easily So next is C cutting out 00:02:12.219 --> 00:02:15.479 tedious tasks the repetitive stuff the soul crushing 00:02:15.479 --> 00:02:18.340 work That's where AI just shines data cleaning 00:02:18.340 --> 00:02:20.659 is probably the biggest win here. Oh, definitely. 00:02:20.699 --> 00:02:22.819 We've all seen that column, right? You should 00:02:22.819 --> 00:02:26.360 just say But you get sales team, sales, maybe 00:02:26.360 --> 00:02:28.860 with a capital S, maybe someone typed the department 00:02:28.860 --> 00:02:32.120 name and Vietnamese. Like, can Don adept? Yeah. 00:02:32.639 --> 00:02:34.759 AI just standardizes all of it, instantly turns 00:02:34.759 --> 00:02:37.560 it all into just sales. That's huge. And it handles 00:02:37.560 --> 00:02:40.099 basic data checks too, right? Like upload a CSV, 00:02:40.460 --> 00:02:43.270 say 50 workshop signups. Yep. Instantly find 00:02:43.270 --> 00:02:46.129 the number of columns. Or, find the top three 00:02:46.129 --> 00:02:48.150 most popular topics from an interests column. 00:02:48.310 --> 00:02:49.990 Stuff that used to mean fiddling with filters 00:02:49.990 --> 00:02:53.550 for ages. Exactly. Okay, so AI gives speed, but 00:02:53.550 --> 00:02:57.289 we need a safety net. Which brings us to H. Help 00:02:57.289 --> 00:02:59.830 provide a safety net. Because humans make silly 00:02:59.830 --> 00:03:02.900 errors. Especially with detailed stuff like compliance 00:03:02.900 --> 00:03:05.599 or policy checks. You know, I have to admit, 00:03:05.759 --> 00:03:08.120 I still get tangled up in confusing policy rules 00:03:08.120 --> 00:03:10.520 sometimes. Or even just prompt definitions. It 00:03:10.520 --> 00:03:12.379 happens. We all do. It's easy to miss things. 00:03:12.699 --> 00:03:14.960 So let's take an expense report. Say I submit 00:03:14.960 --> 00:03:19.150 a $160 dinner receipt. Three people and oh it 00:03:19.150 --> 00:03:21.530 includes wine. Okay, so you upload that receipt 00:03:21.530 --> 00:03:24.449 and the company is like 50 page expense policy 00:03:24.449 --> 00:03:28.110 PDF right and the policy says maybe fifty dollars 00:03:28.110 --> 00:03:31.889 per person max and Absolutely, no alcohol reimbursement. 00:03:31.909 --> 00:03:34.110 That's a perfect test. Now. What if that policy 00:03:34.110 --> 00:03:37.960 PDF is like Enormous. 100 pages. Does the AI 00:03:37.960 --> 00:03:40.240 really read it all? Or does it just skim the 00:03:40.240 --> 00:03:42.639 start and miss some crucial detail buried on 00:03:42.639 --> 00:03:46.039 page 73? That's where its large context understanding 00:03:46.039 --> 00:03:48.639 is so powerful. It doesn't just skim. It synthesizes 00:03:48.639 --> 00:03:50.719 across the whole document. You ask it, check 00:03:50.719 --> 00:03:54.120 for violations. It flags that $160 meal instantly. 00:03:54.340 --> 00:03:56.580 It says, hold on. That's over the $150 limit 00:03:56.580 --> 00:03:58.560 for three people. And there's wine. It's a perfect 00:03:58.560 --> 00:04:01.219 second checker. The compliance safety net. Exactly. 00:04:01.479 --> 00:04:05.400 OK, moving to I. Inspire better creativity. This 00:04:05.400 --> 00:04:07.300 is about challenging our own assumptions, breaking 00:04:07.300 --> 00:04:09.599 habits. Yeah. So upload your 10 -slide pitch 00:04:09.599 --> 00:04:12.500 deck, or a big presentation, and ask the AI to 00:04:12.500 --> 00:04:15.319 act like a very tough, skeptical investor. Ooh, 00:04:15.319 --> 00:04:18.160 I like that. Don't ask for praise. Ask it to 00:04:18.160 --> 00:04:20.980 find the holes. Right. Focus purely on risks, 00:04:21.240 --> 00:04:24.120 hidden costs. It breaks your confirmation bias. 00:04:24.259 --> 00:04:26.360 So it forces you to confront those tough questions 00:04:26.360 --> 00:04:28.720 you might ignore. Totally. Like, what's your 00:04:28.720 --> 00:04:31.199 current burn rate, and how many months of runway 00:04:31.199 --> 00:04:34.610 do you actually have? Or, show me the one metric 00:04:34.610 --> 00:04:37.350 that proves your model works at scale. Yeah, 00:04:37.689 --> 00:04:40.490 it finds those conceptual gaps we tend to blind 00:04:40.490 --> 00:04:43.730 ourselves to. Okay, and the last one in achieve. 00:04:44.069 --> 00:04:47.550 E, enable great ideas to scale faster. This is 00:04:47.550 --> 00:04:49.769 moving beyond just simple reporting. Think mass 00:04:49.769 --> 00:04:52.329 personalization. Okay, like those workshop signups 00:04:52.329 --> 00:04:55.389 again. with interests and experience level. Exactly. 00:04:55.470 --> 00:04:57.870 You might have hundreds of entries. Use that 00:04:57.870 --> 00:05:00.310 data to write unique personalized emails for 00:05:00.310 --> 00:05:02.829 every single attendee. Whoa. So if someone's 00:05:02.829 --> 00:05:04.350 interested in writing in their intermediate... 00:05:04.350 --> 00:05:06.730 The email gives them a specific tip, maybe about 00:05:06.730 --> 00:05:09.069 writing benefit -driven headlines. But if someone 00:05:09.069 --> 00:05:11.689 else picked design and beginner... Their email 00:05:11.689 --> 00:05:15.149 recommends focusing on the 60 -30 -10 color rule. 00:05:15.410 --> 00:05:18.490 Completely different. Tailored advice. Wow. Imagine 00:05:18.490 --> 00:05:20.449 scaling that level of personalized analysis, 00:05:20.589 --> 00:05:24.050 that outreach, to thousands of users instantly. 00:05:24.430 --> 00:05:26.490 Right. That used to take marketing teams days, 00:05:26.670 --> 00:05:29.930 maybe weeks. That's a real shift in power. So 00:05:29.930 --> 00:05:33.410 reflecting on achieve, what do you think is the 00:05:33.410 --> 00:05:36.310 core lesson about using AI effectively here? 00:05:36.730 --> 00:05:38.689 Well, I think it shows AI is really an amplifier. 00:05:39.050 --> 00:05:41.250 Your clarity of instruction is way more vital 00:05:41.250 --> 00:05:45.470 than raw technical skill. Okay, so achieve tells 00:05:45.470 --> 00:05:48.970 us when AI is most valuable. Now, how do we actually 00:05:48.970 --> 00:05:51.689 use it correctly? You know without getting bad 00:05:51.689 --> 00:05:54.569 data or wrong answers? That's where the DIG framework 00:05:54.569 --> 00:05:58.589 comes in describe introspect goal set. It's basically 00:05:58.589 --> 00:06:01.750 exploratory data analysis EDA, but really optimized 00:06:01.750 --> 00:06:05.120 for talking to an AI. Just that first crucial 00:06:05.120 --> 00:06:07.579 step of checking your data for flaws and features 00:06:07.579 --> 00:06:10.300 before you ask the big question. Exactly. And 00:06:10.300 --> 00:06:12.819 step one, the D, describe your data is the absolute 00:06:12.819 --> 00:06:14.860 most important. You cannot skip this. Do not 00:06:14.860 --> 00:06:18.079 pass go. Do not collect $200. Huh, right. Don't 00:06:18.079 --> 00:06:19.939 ask for the fancy charts immediately. First, 00:06:19.959 --> 00:06:21.660 just make sure you and the AI are on the same 00:06:21.660 --> 00:06:24.160 page. Upload your file, say customerfeedbackq1 00:06:24.160 --> 00:06:26.920 .xlsx. And straight away, ask for three things. 00:06:27.800 --> 00:06:30.459 List all the column names. Tell me their data 00:06:30.459 --> 00:06:33.759 types. Text, number, date. And show me the first 00:06:33.759 --> 00:06:36.699 three rows of data. Why those three things specifically? 00:06:36.879 --> 00:06:40.379 It forces both of you, you and the AI, to see 00:06:40.379 --> 00:06:42.879 potential problems right at the start. You need 00:06:42.879 --> 00:06:46.300 to look for NAN. Which means not a number, basically 00:06:46.300 --> 00:06:48.720 blank or missing data. Right, because missing 00:06:48.720 --> 00:06:52.100 data causes huge mistakes later on. Then you 00:06:52.100 --> 00:06:55.740 verify understanding. Ask the AI to explain what 00:06:55.740 --> 00:06:58.259 each column means, but in simple language. Like, 00:06:58.319 --> 00:07:01.000 okay, explain rating one to five. And you want 00:07:01.000 --> 00:07:03.199 it to say something like... That's the customer 00:07:03.199 --> 00:07:06.139 satisfaction score. Five means very happy. If 00:07:06.139 --> 00:07:08.199 it gets that wrong or misunderstands a term, 00:07:08.459 --> 00:07:10.519 correct it right there immediately. Read a sec, 00:07:10.639 --> 00:07:13.279 though. If the AI is so smart, why do I have 00:07:13.279 --> 00:07:16.199 to do this describe step? Feels like I'm hand 00:07:16.199 --> 00:07:18.240 -holding it. Can't I just trust it to figure 00:07:18.240 --> 00:07:21.680 out the columns? Think of it like this. The AI 00:07:21.680 --> 00:07:25.019 is brilliant, but maybe a bit naive, like a genius 00:07:25.019 --> 00:07:27.199 kid. It can interpret symbols, but it doesn't 00:07:27.199 --> 00:07:29.579 feel the real world context behind your specific 00:07:29.579 --> 00:07:32.839 data. Checking for missing data, making sure 00:07:32.839 --> 00:07:35.519 the definitions are spot on up front that prevents 00:07:35.519 --> 00:07:38.879 tiny errors from snowballing into massive problems 00:07:38.879 --> 00:07:41.779 when you run complex analysis later. It saves 00:07:41.779 --> 00:07:44.899 you pain down the road. Got it. So describe first, 00:07:45.060 --> 00:07:48.459 then what's the I in DIG? Step two, introspect 00:07:48.459 --> 00:07:51.379 the data. Now that you both understand the basic 00:07:51.379 --> 00:07:53.480 structure, you start thinking about patterns, 00:07:53.740 --> 00:07:56.160 relationships, potential red flags. How do you 00:07:56.160 --> 00:07:58.980 do that with the AI? Ask it to suggest, say, 00:07:59.160 --> 00:08:01.160 five interesting questions the data could answer 00:08:01.160 --> 00:08:04.220 based on the columns it sees. OK. It might suggest. 00:08:04.339 --> 00:08:06.379 Is there a connection between the support agent 00:08:06.379 --> 00:08:09.019 column and the rating, one to five? That's a 00:08:09.019 --> 00:08:10.759 good question the data probably can answer. This 00:08:10.759 --> 00:08:12.420 is also where you catch its mistakes, right? 00:08:12.420 --> 00:08:15.160 Crucial correction loop, yeah. If the AI suggests, 00:08:15.579 --> 00:08:19.370 what are the sales figures in each country? But 00:08:19.370 --> 00:08:20.970 you know, all your customers are in Vietnam. 00:08:21.089 --> 00:08:22.709 You have to jump in and say, hold on, this data 00:08:22.709 --> 00:08:26.189 is only for Vietnam. Exactly. Correct that assumption 00:08:26.189 --> 00:08:29.069 immediately. This back and forth, this introspection, 00:08:29.550 --> 00:08:31.569 it prevents you running a whole analysis based 00:08:31.569 --> 00:08:33.850 on a totally false premise. It might feel a bit 00:08:33.850 --> 00:08:36.570 slow initially. But it guarantees accuracy later. 00:08:36.850 --> 00:08:39.529 Precisely. OK, final step. G set clear goals. 00:08:39.929 --> 00:08:42.649 This is really about prompt engineering. Prompt 00:08:42.649 --> 00:08:45.389 engineering. Just giving the AI clear instructions 00:08:45.389 --> 00:08:47.350 on the output you want. Yeah, clear constraints. 00:08:47.610 --> 00:08:50.429 The AI needs context. How should the final result 00:08:50.429 --> 00:08:54.090 look? What's its purpose? So be specific. My 00:08:54.090 --> 00:08:56.070 goal is to find out why customer satisfaction 00:08:56.070 --> 00:08:59.830 dropped last quarter. And add detail. Focus only 00:08:59.830 --> 00:09:02.450 on negative comments ratings one and two. I need 00:09:02.450 --> 00:09:04.850 three summary bullet points and one pie chart, 00:09:04.990 --> 00:09:07.490 formatted for a professional PowerPoint presentation. 00:09:07.769 --> 00:09:10.210 Ah, OK. because that's totally different from 00:09:10.210 --> 00:09:12.929 asking for. Fun facts for Twitter about customer 00:09:12.929 --> 00:09:15.710 feedback. That implies a completely different 00:09:15.710 --> 00:09:18.710 tone, analysis depth, and output format. The 00:09:18.710 --> 00:09:21.149 goal dictates everything. So let's say someone's 00:09:21.149 --> 00:09:24.129 impatient. Why shouldn't they trust the slow 00:09:24.129 --> 00:09:27.970 described step when they just want speed? Because 00:09:27.970 --> 00:09:31.009 skipping that early data description, it almost 00:09:31.009 --> 00:09:33.929 guarantees expensive, painful mistakes later 00:09:33.929 --> 00:09:39.080 on. It's just not worth the risk. All right, 00:09:39.080 --> 00:09:40.779 now let's go beyond just cleaning up spreadsheet 00:09:40.779 --> 00:09:45.019 tables. AI unlocks analysis that, honestly, used 00:09:45.019 --> 00:09:48.039 to need dedicated data engineers, especially 00:09:48.039 --> 00:09:50.600 with something called smart filtering. Smart 00:09:50.600 --> 00:09:52.960 filtering? Yeah. You mean filtering based on 00:09:52.960 --> 00:09:55.679 concepts, not just exact words in a column. Exactly. 00:09:55.759 --> 00:09:58.200 That's a huge shift. Think about job hunting. 00:09:58.620 --> 00:10:00.879 You're looking through a massive list. OK. You 00:10:00.879 --> 00:10:03.460 want a salary between, say, $50k and $80. Easy 00:10:03.460 --> 00:10:06.120 enough. And you want it located on the US East 00:10:06.120 --> 00:10:09.360 Coast. And you want jobs involving keywords like 00:10:09.360 --> 00:10:12.059 woodworking or maybe carpentry. OK, but wait. 00:10:12.399 --> 00:10:14.980 My spreadsheet might only list cities Boston, 00:10:15.039 --> 00:10:17.220 New York, Miami. It probably doesn't have an 00:10:17.220 --> 00:10:19.200 East Coast column. Right. Traditional software 00:10:19.200 --> 00:10:20.740 just fails there. Can't make the connection. 00:10:20.759 --> 00:10:22.879 And it might not have a skills column with woodworking 00:10:22.879 --> 00:10:26.399 either. Exactly. But the AI knows that Boston 00:10:26.399 --> 00:10:28.399 is on the East Coast from its general knowledge. 00:10:28.720 --> 00:10:31.299 It can also read the job description text. Ah, 00:10:31.299 --> 00:10:34.539 so even if the title is Residential Project Manager, 00:10:34.879 --> 00:10:38.200 if the description mentions carpentry. The AI 00:10:38.200 --> 00:10:41.360 can conceptually match it to your carpentry keyword. 00:10:41.779 --> 00:10:44.279 It applies its massive knowledge graph to your 00:10:44.279 --> 00:10:47.620 specific data points. It's not magic, it's inference. 00:10:48.179 --> 00:10:50.179 That makes sense. Okay, that's powerful filtering. 00:10:50.600 --> 00:10:53.019 What else? Making your work reproducible. This 00:10:53.019 --> 00:10:55.539 is crucial for any professional team, really 00:10:55.539 --> 00:10:58.399 anyone doing serious analysis. Meaning you don't 00:10:58.399 --> 00:11:01.000 just save the final chart or number. No, you 00:11:01.000 --> 00:11:03.580 have to save the method. How did you get there? 00:11:04.000 --> 00:11:07.320 Ask the AI to create a kind of recipe book. A 00:11:07.320 --> 00:11:10.320 recipe book, like a log file. Sort of, but more 00:11:10.320 --> 00:11:13.120 structured. A tracking document, it lists the 00:11:13.120 --> 00:11:15.379 original file name you used, all the steps you 00:11:15.379 --> 00:11:17.120 took, like standardized the department column, 00:11:17.500 --> 00:11:20.320 and any limitations you found, like 20 % of customer 00:11:20.320 --> 00:11:22.899 comments were blank. OK, so that's the roadmap 00:11:22.899 --> 00:11:24.799 if someone else needs to rerun it, or if I need 00:11:24.799 --> 00:11:26.919 to remember what I did three months later. Exactly. 00:11:27.039 --> 00:11:28.679 It stops you reinventing the wheel or getting 00:11:28.679 --> 00:11:31.460 lost trying to trace back through 50 chat messages. 00:11:31.679 --> 00:11:33.960 And for that analysis, like the pie chart of 00:11:33.960 --> 00:11:35.879 complaints we talked about? Here's the really 00:11:35.879 --> 00:11:40.039 cool part. You can ask the AI, generate a full 00:11:40.039 --> 00:11:44.059 commented Python script, maybe call it complaintspy 00:11:44.059 --> 00:11:47.220 .py, that does exactly what we just did. Whoa. 00:11:47.450 --> 00:11:50.169 So it writes the code for the entire cleaning 00:11:50.169 --> 00:11:53.090 and analysis process. Yep, and now that sequence 00:11:53.090 --> 00:11:56.309 of steps. It's a reusable tool, which leads to 00:11:56.309 --> 00:11:59.470 this idea of turning conversations into programs. 00:12:00.029 --> 00:12:03.190 Okay, that sounds potentially complicated. Like, 00:12:03.450 --> 00:12:05.549 is that Python script actually readable? Can 00:12:05.549 --> 00:12:09.149 someone like me, who isn't really a coder, understand 00:12:09.149 --> 00:12:11.690 and trust it? Yes, generally. Because you tell 00:12:11.690 --> 00:12:13.870 the AI to make it commented. It doesn't just 00:12:13.870 --> 00:12:16.330 generate code. It adds explanations in plain 00:12:16.330 --> 00:12:18.370 English for what each chunk of code is doing. 00:12:18.470 --> 00:12:20.330 So it makes the analysis transparent, even if 00:12:20.330 --> 00:12:22.509 it's code. Right. Think about a complex sequence. 00:12:22.950 --> 00:12:25.629 Maybe grabbing 10 frames from a movie file, resizing 00:12:25.629 --> 00:12:27.730 them, converting them to grayscale, asking the 00:12:27.730 --> 00:12:30.070 AI to generate descriptions for each, and then 00:12:30.070 --> 00:12:32.429 saving all that info into a CSV file. That sounds 00:12:32.429 --> 00:12:34.610 like a lot of manual steps or needing special 00:12:34.610 --> 00:12:37.309 software. It was. But now you can walk the AI 00:12:37.309 --> 00:12:39.330 through that process conversationally, then ask 00:12:39.330 --> 00:12:42.070 it to bundle that entire sequence. Into a single 00:12:42.070 --> 00:12:43.809 Python program you can just download and run 00:12:43.809 --> 00:12:46.549 next time. Exactly. Download it, share it, run 00:12:46.549 --> 00:12:48.590 on your local machine whenever you need it. So 00:12:48.590 --> 00:12:50.830 thinking about that reproducible code generation, 00:12:52.169 --> 00:12:55.129 what's the biggest barrier AI really removes 00:12:55.129 --> 00:12:57.820 there? I'd say the biggest barrier removed is 00:12:57.820 --> 00:13:01.240 needing that deep manual coding expertise just 00:13:01.240 --> 00:13:04.240 to automate and share a specific analysis workflow. 00:13:05.039 --> 00:13:07.799 Any sequence can potentially become a reusable 00:13:07.799 --> 00:13:10.759 tool now. Now we've talked a lot about AI generally, 00:13:10.840 --> 00:13:12.720 but we should probably touch on specific tools. 00:13:13.200 --> 00:13:15.769 They aren't all the same, are they? No, definitely 00:13:15.769 --> 00:13:17.889 not. Different models have different strengths. 00:13:18.409 --> 00:13:21.370 Chat GPT, especially with its advanced data analysis 00:13:21.370 --> 00:13:25.210 feature, is kind of the reliable, flexible, all 00:13:25.210 --> 00:13:27.970 -around choice for many common data tasks. Good 00:13:27.970 --> 00:13:29.929 place to start. OK. What about others? Claude 00:13:29.929 --> 00:13:31.649 gets mentioned a lot. Yeah, Claude often gets 00:13:31.649 --> 00:13:34.190 highlighted for a few things. Generating really 00:13:34.190 --> 00:13:36.990 clean code, creating interactive dashboards maybe, 00:13:37.190 --> 00:13:40.009 and especially its huge context window. Context 00:13:40.009 --> 00:13:41.870 window. Right. Meaning how much information it 00:13:41.870 --> 00:13:44.899 can handle at once. Exactly. That large context 00:13:44.899 --> 00:13:47.279 window is a big deal when you're working with 00:13:47.279 --> 00:13:50.360 truly massive or numerous documents. Imagine 00:13:50.360 --> 00:13:53.399 uploading like a whole company's archive of legal 00:13:53.399 --> 00:13:57.000 contracts. Or maybe 200 different annual reports 00:13:57.000 --> 00:13:59.480 to compare them all at once. Right. For that 00:13:59.480 --> 00:14:02.580 kind of huge -scale file analysis and synthesis, 00:14:03.139 --> 00:14:05.919 Claude's ability to handle more information simultaneously 00:14:05.919 --> 00:14:08.720 is a major advantage right now. Interesting. 00:14:09.139 --> 00:14:11.509 And you also mentioned perplexity. Proplexity 00:14:11.509 --> 00:14:13.690 shines for research and finding real -time information. 00:14:13.830 --> 00:14:16.149 It's great at citing sources, and it has focus 00:14:16.149 --> 00:14:18.190 modes. If you switch it to finance mode, for 00:14:18.190 --> 00:14:20.669 instance, you can get current market analysis 00:14:20.669 --> 00:14:23.929 layered right onto your data questions. So choose 00:14:23.929 --> 00:14:26.669 the tool based on the specific job. Pretty much. 00:14:26.750 --> 00:14:28.769 And remember, data analysis doesn't have to stop 00:14:28.769 --> 00:14:31.169 at just creating a report or a chart. Right. 00:14:31.190 --> 00:14:33.450 You can use these tools as building blocks for 00:14:33.450 --> 00:14:35.750 actual applications. Totally. Think practical 00:14:35.750 --> 00:14:38.149 stuff. You could build a basic traffic analysis 00:14:38.149 --> 00:14:41.139 app that monitors real -time data feed and pings 00:14:41.139 --> 00:14:43.240 you with alerts about incidents on your commute. 00:14:43.419 --> 00:14:46.299 Okay, or what about video? You could build a 00:14:46.299 --> 00:14:49.059 video privacy tool, something that automatically 00:14:49.059 --> 00:14:51.759 scans through, say, thousands of hours of security 00:14:51.759 --> 00:14:54.799 footage and blurs out faces or license plates. 00:14:54.960 --> 00:14:57.340 Wow. Or maybe for finance, like an investment 00:14:57.340 --> 00:14:59.940 research assistant. Yeah. Imagine a simple Q 00:14:59.940 --> 00:15:02.259 &A interface like a chatbot that sits on top 00:15:02.259 --> 00:15:04.820 of a massive private database of financial reports. 00:15:05.059 --> 00:15:07.600 You just ask it questions in plain English. Going 00:15:07.600 --> 00:15:09.980 back to the tools for a second, if you were tackling 00:15:09.980 --> 00:15:14.460 really huge complex documents like that Library 00:15:14.460 --> 00:15:16.620 of Legal Agreements example, which tool stands 00:15:16.620 --> 00:15:19.440 out? Based on current capabilities, Claude's 00:15:19.440 --> 00:15:22.080 large context window generally makes it superior 00:15:22.080 --> 00:15:25.220 for that kind of large -scale multi -file document 00:15:25.220 --> 00:15:26.940 analysis. Okay, let's try and bring this whole 00:15:26.940 --> 00:15:29.750 deep dive together then. We started with Achieve. 00:15:30.029 --> 00:15:32.409 That framework helps you define when AI really 00:15:32.409 --> 00:15:34.970 adds the most value. Right. Aiding coordination, 00:15:35.250 --> 00:15:37.950 cutting tedium, help with safety nets, inspiring 00:15:37.950 --> 00:15:41.049 creativity, and enabling ideas to scale. Achieve. 00:15:41.210 --> 00:15:43.429 And then we got DIG. That's the framework to 00:15:43.429 --> 00:15:46.009 make sure your actual analysis process is sound. 00:15:46.450 --> 00:15:49.149 Every single time. Describe the data meticulously 00:15:49.149 --> 00:15:52.429 first. Introspect for patterns, relationships, 00:15:52.509 --> 00:15:56.129 and flaws. And set clear goals for your output. 00:15:56.470 --> 00:16:00.100 DIG. You know, this technology is not really 00:16:00.100 --> 00:16:02.460 here to replace your strategic thinking, is it? 00:16:02.559 --> 00:16:05.019 No, not at all. It's here to make you faster, 00:16:05.179 --> 00:16:09.019 more accurate, to let you achieve data manipulation 00:16:09.019 --> 00:16:11.799 and analysis that frankly was impossible for 00:16:11.799 --> 00:16:14.220 most people before without years of dedicated 00:16:14.220 --> 00:16:17.279 coding training. The power of complex data insight. 00:16:17.820 --> 00:16:20.139 It's actually becoming available to you, the 00:16:20.139 --> 00:16:23.320 listener. Yeah. So here's a call to action. Take 00:16:23.320 --> 00:16:25.100 a spreadsheet you work with regularly, something 00:16:25.100 --> 00:16:27.080 you know well, upload it to one of these tools. 00:16:27.220 --> 00:16:28.940 And then just walk through the DIG framework, 00:16:29.100 --> 00:16:31.340 step by step. Exactly. Describe, introspect, 00:16:31.659 --> 00:16:33.860 goal set, just see what comes up, see what insights 00:16:33.860 --> 00:16:35.980 surface that you might have missed before. Practice 00:16:35.980 --> 00:16:38.340 and repetition, that seems key. So let's leave 00:16:38.340 --> 00:16:40.240 everyone with a final thought, a provocative 00:16:40.240 --> 00:16:43.159 one. Now that this power is more accessible, 00:16:43.799 --> 00:16:45.980 what non -traditional data might you analyze 00:16:45.980 --> 00:16:48.259 first? Maybe that folder full of messy voice 00:16:48.259 --> 00:16:50.779 notes you've been meaning to transcribe. or an 00:16:50.779 --> 00:16:53.279 archive of old marketing videos. What hidden 00:16:53.279 --> 00:16:56.399 patterns could you unlock now? Lots to think 00:16:56.399 --> 00:16:57.799 about. Until the next deep dive.