WEBVTT 00:00:00.600 --> 00:00:07.512 Welcome to lesson 4.2, AI for data analysis: finding insights without being a data scientist. 00:00:07.862 --> 00:00:11.553 Every business has data, but few people have time to read it. 00:00:11.903 --> 00:00:17.215 Here you will learn to analyse a dataset by asking an AI tool questions in plain English. 00:00:17.565 --> 00:00:22.642 You will choose the right chart, and turn the numbers into a short story that a manager can act on. 00:00:22.992 --> 00:00:27.941 We will also cover why data matters in any job, and how to work ethically and safely. 00:00:29.441 --> 00:00:32.428 By the end of this lesson, you will be able to do four things. 00:00:32.778 --> 00:00:38.197 First, use an AI data tool to explore and analyse a spreadsheet or a CSV file. 00:00:38.547 --> 00:00:43.944 Second, write specific questions about totals, trends, comparisons and unusual values. 00:00:44.294 --> 00:00:48.859 Third, choose a suitable chart and write a plain-language summary of what you found. 00:00:49.209 --> 00:00:53.689 Fourth, apply the basics of the GDPR, the EU data protection law. 00:00:54.039 --> 00:00:59.501 That means removing personal details from data and checking a tool's terms before you use it. 00:01:01.001 --> 00:01:03.177 Here are some key terms for this lesson. 00:01:03.527 --> 00:01:08.177 A dataset is a collection of related data, usually in rows and columns. 00:01:08.527 --> 00:01:15.845 A CSV, which stands for comma-separated values, is a simple file format for tables that almost any program can open. 00:01:16.195 --> 00:01:20.547 An insight is a clear, useful finding that helps someone make a decision. 00:01:20.897 --> 00:01:25.227 A trend is a pattern of change over time, such as sales going up or down. 00:01:25.577 --> 00:01:29.652 An outlier is an unusual value that stands far apart from the rest. 00:01:30.002 --> 00:01:31.922 It is often worth a closer look. 00:01:32.272 --> 00:01:36.624 And a dashboard is a visual summary of key numbers and charts, all in one view. 00:01:38.124 --> 00:01:40.215 Every spreadsheet hides decisions. 00:01:40.565 --> 00:01:46.965 A sales spreadsheet is just numbers until someone asks a question, such as: which products should we stock more of? 00:01:47.315 --> 00:01:49.341 Analysis is how you answer that. 00:01:49.691 --> 00:01:53.403 In the past, reading data needed coding or statistics training. 00:01:53.753 --> 00:01:56.377 AI tools have removed most of that barrier. 00:01:56.727 --> 00:01:58.925 So here is the promise of this lesson. 00:01:59.275 --> 00:02:04.864 If you can ask clear questions in plain language, you can produce insights that businesses will pay for. 00:02:06.364 --> 00:02:08.369 Where does business data come from? 00:02:08.719 --> 00:02:10.618 You do not need fancy data. 00:02:10.968 --> 00:02:13.635 Most organisations already have plenty. 00:02:13.985 --> 00:02:16.246 First, there is sales and orders data. 00:02:16.596 --> 00:02:19.348 It shows what sold, when, to whom and for how much. 00:02:19.698 --> 00:02:22.237 This is the heart of most business analysis. 00:02:22.587 --> 00:02:26.981 Second, website and social media data, such as visits, clicks and likes. 00:02:27.331 --> 00:02:29.806 This shows what customers are interested in. 00:02:30.156 --> 00:02:35.660 Third, surveys and feedback, like reviews and ratings, which tell you the why behind the numbers. 00:02:36.010 --> 00:02:40.789 There is also operations data, such as stock levels, delivery times and costs. 00:02:41.139 --> 00:02:45.299 The real skill is asking the right questions of data that already exists. 00:02:46.799 --> 00:02:51.513 Tools like Julius AI let you analyse data simply by chatting with it. 00:02:51.863 --> 00:02:53.847 First, you upload a data file. 00:02:54.197 --> 00:02:56.523 For example, a shop's sales export. 00:02:56.873 --> 00:02:58.899 Second, you ask in plain English. 00:02:59.249 --> 00:03:02.407 You might type: what were the top three products last quarter? 00:03:02.757 --> 00:03:04.421 There is no code to write. 00:03:04.771 --> 00:03:07.352 Third, the tool does the work and explains it. 00:03:07.702 --> 00:03:11.777 The AI runs the analysis, draws charts and explains the result. 00:03:12.127 --> 00:03:13.983 But remember the split of jobs. 00:03:14.333 --> 00:03:16.381 The AI does the calculation. 00:03:16.731 --> 00:03:21.360 Your job is to ask the right question and to judge whether the answer makes sense. 00:03:22.860 --> 00:03:27.041 Here are five steps from raw file to clear findings, with no formulas. 00:03:27.391 --> 00:03:29.333 Step one, upload the data. 00:03:29.683 --> 00:03:34.163 Drop in your CSV or spreadsheet, and the tool reads the columns and rows. 00:03:34.513 --> 00:03:36.198 Step two, get to know it. 00:03:36.548 --> 00:03:38.383 Ask: what is in this data? 00:03:38.733 --> 00:03:41.335 You will see the columns, time period and size. 00:03:41.685 --> 00:03:48.128 Step three, ask focused questions, one at a time, about totals, trends, comparisons or unusual values. 00:03:48.478 --> 00:03:50.227 Step four, create charts. 00:03:50.577 --> 00:03:53.201 Ask for a bar or line chart of the key result. 00:03:53.551 --> 00:03:55.813 Step five, summarise the insights. 00:03:56.163 --> 00:04:00.621 Write a short, plain summary of the top findings for the person who makes decisions. 00:04:00.971 --> 00:04:05.131 You will do exactly this with Julius AI in exercise 4.2. 00:04:06.631 --> 00:04:08.999 Good analysis starts with good questions. 00:04:09.349 --> 00:04:10.907 Here are four kinds. 00:04:11.257 --> 00:04:13.262 First, totals and averages. 00:04:13.612 --> 00:04:17.964 For example: how much did we sell in total, and what is the average order value? 00:04:18.314 --> 00:04:20.063 Second, trends over time. 00:04:20.413 --> 00:04:22.675 How have sales changed month by month? 00:04:23.025 --> 00:04:24.539 Third, comparisons. 00:04:24.889 --> 00:04:27.556 Which region or which product performed best? 00:04:27.906 --> 00:04:29.335 Fourth, outliers. 00:04:29.685 --> 00:04:32.032 Were there any unusual spikes or drops? 00:04:32.382 --> 00:04:34.750 Surprises often hide the real story. 00:04:35.100 --> 00:04:36.593 Tip: start broad. 00:04:36.943 --> 00:04:38.458 Ask: what stands out? 00:04:38.808 --> 00:04:41.176 Then zoom in on whatever looks interesting. 00:04:42.676 --> 00:04:46.281 Like chatbots, data tools reward clear, specific questions. 00:04:46.631 --> 00:04:51.261 A clear question sounds like: top three products by revenue in the fourth quarter. 00:04:51.611 --> 00:04:55.472 You ask one question at a time, so each answer is easy to check. 00:04:55.822 --> 00:04:58.766 You ask the tool to show you the numbers behind a chart. 00:04:59.116 --> 00:05:01.761 And you ask for a chart with a one-line takeaway. 00:05:02.111 --> 00:05:04.671 A vague question sounds like: what sold well? 00:05:05.021 --> 00:05:11.528 Or five questions packed into one prompt, a chart accepted without checking, or simply: analyse everything. 00:05:11.878 --> 00:05:13.286 The rule is simple. 00:05:13.636 --> 00:05:16.047 Name the exact column or time period. 00:05:16.397 --> 00:05:18.573 Vague questions get vague answers. 00:05:18.923 --> 00:05:22.805 And always check the numbers, because AI can misread messy data. 00:05:24.305 --> 00:05:28.487 The right chart makes a finding obvious, so match the chart to the question. 00:05:28.837 --> 00:05:34.042 A bar chart is best for comparing categories, such as sales by product, region or month. 00:05:34.392 --> 00:05:40.024 A line chart is best for trends over time, such as how weekly sales rise or fall across a year. 00:05:40.374 --> 00:05:45.921 A pie chart shows parts of a whole, with only a few slices, such as sales shares of three shops. 00:05:46.271 --> 00:05:52.607 And a simple table works best when exact numbers matter more than the shape, like a price list or the top ten orders. 00:05:52.957 --> 00:05:54.215 Here is a tip. 00:05:54.565 --> 00:05:57.957 If a chart needs a long explanation, choose a simpler chart. 00:05:59.457 --> 00:06:03.681 Insights only matter if people act on them, so wrap your numbers in a story. 00:06:04.031 --> 00:06:05.802 First, lead with the finding. 00:06:06.152 --> 00:06:10.781 Start with the point, for example: three products drove most of last quarter's sales. 00:06:11.131 --> 00:06:12.518 Then show the chart. 00:06:12.868 --> 00:06:14.745 Second, say why it matters. 00:06:15.095 --> 00:06:20.365 Connect it to a decision, such as: so we should keep these in stock and feature them on the homepage. 00:06:20.715 --> 00:06:22.229 Third, keep it plain. 00:06:22.579 --> 00:06:23.859 Avoid jargon. 00:06:24.209 --> 00:06:29.415 A 150-word summary that anyone understands beats a wall of statistics. 00:06:29.765 --> 00:06:33.775 Remember, a decision-maker wants to know so what, not just see a chart. 00:06:35.275 --> 00:06:40.886 A real insight sounds like this: sales fell 20 per cent in March, driven by one region. 00:06:41.236 --> 00:06:44.095 That is an invented example, but look at its shape. 00:06:44.445 --> 00:06:46.941 It names the number, the period and the cause. 00:06:47.291 --> 00:06:49.403 And it points to a clear next action. 00:06:49.753 --> 00:06:53.891 A vague observation sounds like this: sales were a bit lower recently. 00:06:54.241 --> 00:06:57.100 There are no numbers, no timeframe and no cause. 00:06:57.450 --> 00:06:59.093 It leaves the reader guessing. 00:06:59.443 --> 00:07:03.603 Always aim for the first kind: specific, useful and linked to a decision. 00:07:05.103 --> 00:07:07.897 Now, analysing data ethically and safely. 00:07:08.247 --> 00:07:12.663 If data identifies real people, treat it as sensitive by default. 00:07:13.013 --> 00:07:15.275 First, anonymise personal data. 00:07:15.625 --> 00:07:20.745 That means removing names, emails and ID numbers before you analyse, whenever you can. 00:07:21.095 --> 00:07:22.439 You rarely need them. 00:07:22.789 --> 00:07:26.821 Second, respect the GDPR, the EU law on personal data. 00:07:27.171 --> 00:07:31.949 Handle data lawfully, keep it secure, and use it only for the agreed purpose. 00:07:32.299 --> 00:07:34.689 Third, keep secrets out of public tools. 00:07:35.039 --> 00:07:40.265 Never upload confidential or personal data to a tool whose terms you have not checked. 00:07:40.615 --> 00:07:42.642 Fourth, sanity-check the AI. 00:07:42.992 --> 00:07:48.155 It can misread messy data, so always verify surprising results before you report them. 00:07:49.655 --> 00:07:51.361 Let's check your understanding. 00:07:51.711 --> 00:07:54.335 Pause the video now and think about these questions. 00:07:54.685 --> 00:07:58.397 What does an AI data tool let you do without writing any code? 00:07:58.747 --> 00:08:02.417 Can you give an example of a specific, well-phrased data question? 00:08:02.767 --> 00:08:06.116 Which chart would you use to show a trend over twelve months, and why? 00:08:06.466 --> 00:08:10.413 And name one thing you must do to handle personal data responsibly. 00:08:10.763 --> 00:08:13.280 When you have your answers, press play to carry on. 00:08:14.780 --> 00:08:17.361 Here are some resources to help you go further. 00:08:17.711 --> 00:08:21.167 Julius AI lets you analyse data by chatting with it. 00:08:21.517 --> 00:08:24.824 Looker Studio from Google lets you build free dashboards. 00:08:25.174 --> 00:08:27.947 And you can read the full text of the GDPR. 00:08:28.297 --> 00:08:30.303 All the links are on the lesson page. 00:08:31.803 --> 00:08:34.128 Here are the key takeaways from this lesson. 00:08:34.478 --> 00:08:39.811 AI tools let you analyse data by asking questions in plain English, with no coding. 00:08:40.161 --> 00:08:45.153 Good analysis comes from specific questions about totals, trends and comparisons. 00:08:45.503 --> 00:08:49.685 Match the chart to the question, and always double-check surprising results. 00:08:50.035 --> 00:08:54.259 Turn numbers into a story: lead with the finding and say why it matters. 00:08:54.609 --> 00:08:56.763 And handle personal data lawfully. 00:08:57.113 --> 00:09:01.551 Anonymise it, respect the GDPR and protect anything confidential. 00:09:03.051 --> 00:09:04.907 Now it is time to practise. 00:09:05.257 --> 00:09:11.870 On the platform, begin with the activity called insight or vague observation, where you sort statements into the right group. 00:09:12.220 --> 00:09:14.375 Then do exercise 4.2. 00:09:14.725 --> 00:09:19.866 You will analyse a dataset and write up your findings, which takes about one and a half hours. 00:09:20.216 --> 00:09:22.221 Finally, take the lesson quiz. 00:09:22.571 --> 00:09:28.417 It has five questions, the pass mark is 70 per cent, and you can retake it as often as you need.