Free courses

Use a spreadsheet when you need to inspect a small table, change a few inputs and explain a result to someone else. Learn SQL when you need to ask repeatable questions of database tables. You do not need to master both before making your first useful report.

One dataset. Two useful questions.

The whole table adds to 125. Tea alone adds to 100. In Excel, with headers in row 1 and these values in A2:B4, use =SUMIF(A2:A4,"Tea",B2:B4). Some regional settings use semicolons instead of commas.

SELECT product, SUM(sales) AS total
FROM sales
GROUP BY product;
-- Coffee: 25; Tea: 100

The trap: COUNT(*) counts rows, not sales. A join can also duplicate an order when the matching key is not unique. Check how many rows you have before and after joining; do not use SUM(DISTINCT sales) to hide duplicated joins, because two genuine sales can have the same value.

Look for in a course: cleaning source rows, formulas or filtering, grouped totals, and an explanation of incorrect results—not just a polished dashboard. Check which Excel version it uses before installing anything; access to a course does not include a desktop software licence.

References: Microsoft · SUMIF; PostgreSQL · Aggregates.

Python is useful when the same file task keeps returning: reading exports, changing a format or calculating a repeatable summary. If you have never written code, start with variables, conditions, loops and functions. A machine-learning course that assumes those skills is not a substitute for the foundations.

A first script with a checkable answer

Save the sales table from the Excel note as sales.csv, with the header Product,Sales. Put this script in the same folder and run it with Python 3. It uses the built-in CSV library, so this example needs no extra packages or API key.

import csv

with open("sales.csv", newline="", encoding="utf-8") as source:
    rows = list(csv.DictReader(source))

tea_total = sum(
    int(row["Sales"]) for row in rows
    if row["Product"] == "Tea"
)
print(tea_total)  # 100

Change one thing: replace the final Tea sale of 60 with 70. The answer should become 110. If it stays at 100, check that you saved and opened the same file. If a value is blank or not an integer, this small script will raise an error; do not silently turn bad financial data into zero.

Choose lessons that show: how to run a file, read an error, inspect an input and write a small function. An AI-generated script can still be wrong. Before automating a real folder, test on copies and print the proposed changes first.

Reference: Python · CSV reader and DictReader.

A good first website is small enough to finish: a page for a local service, a portfolio project or an event. Write down its one job. Someone viewing a service page should be able to understand the offer and find a working contact link without decoding a menu of features.

Learn these in order

1. HTML: a meaningful heading, paragraphs, links and labelled form controls. Use a real link for navigation and a button for an action.
2. CSS: spacing, readable type, flexible layouts and images that fit the available width.
3. JavaScript: add behaviour only where the page needs it, such as a filter or a form response.

A useful finish line: open the page at a narrow phone width and on a laptop, use it with the keyboard, and check that its links work. Zoom in and read the main content. A screenshot that looks good at one width is not enough.

Skip for now: a framework comparison, a paid hosting plan and a large component library. You can build a first local HTML page with a text editor and a browser. A design-tool course teaches a different part of the work; a beautiful mockup does not automatically become a working, accessible page.

Reference: MDN · Your first website.

Start with a low-risk task whose answer you can check: turning your own meeting notes into a draft action list, or summarising a public document with source references. Keep private customer or employee information out of tools unless your organisation has approved the environment and its data handling.

A brief that is easier to verify

Using only these notes, list each action, its owner and its due date. Quote the note that supports each action. Write “not stated” when an owner or date is missing. Do not invent a commitment.

Then check: open each quoted source, confirm the owner and date, and compare the draft with the original. A fluent answer or a citation-shaped link is not evidence that a claim is true. Keep sending, refunds and account changes outside a beginner workflow.

For a fictional team, 100 replies at eight minutes each take 800 staff-minutes. If drafting takes three minutes and review takes another three, the new total is 600 minutes—not 300. The illustrative saving is 200 minutes before setup, tool and monitoring costs. Measure the whole job before calling the workflow an improvement.

Choose a course for the work you need: everyday AI use, workflow automation and model development have different prerequisites. Check whether the lessons need a paid tool, API credits or local hardware. “Free course” describes access to lessons, not every service used in them.

Reference: NIST · AI Risk Management Framework. The prompt and time calculation are Forasm teaching examples, not a safety certification.