Part 2 · How AI Actually Works
Data Is Everywhere (Documents, Emails, Spreadsheets)
These tools predict the next word from patterns in enormous amounts of text. Now glance at your own screen: the half-written email, the report in the next tab, last quarter's spreadsheet. Every one of those is the same raw material the AI trained on.
Why This Matters
Most people assume AI needs a clean database before it can help, so they never start, because that tidy database rarely exists. This lesson sits in Data Evaluation, one of the data dimensions of your personal AI readiness, and it is about seeing what you already have. By the end you will be able to:
- Recognise that data means documents, emails, reports, chat logs, and spreadsheets, not just databases
- Identify five data sources in your own work that an AI tool could read
- Describe why naming them is your first move toward using AI
Data Is More Than the Database
Split data into two kinds. Structured data lives in neat rows and columns: a spreadsheet with clean headers, a customer table, a list of transactions. Unstructured data is everything written in ordinary language: emails, reports, meeting notes, chat logs, contracts, slide decks. In most organisations, the second kind vastly outnumbers the first. The knowledge you actually work with sits in prose, not tidy tables.
Earlier you saw that AI is, underneath, just maths, and that is why format matters less than you think. A word, a sentence, a number in a cell: the model turns each into a pattern of numbers it can compare. Whether the information arrives as a database export or a three-page memo, it becomes the same raw material once the tool reads it.
This is newer than it sounds. Older analytics made you force everything into tables first. Generative AI reads the messy prose directly, which is why your inbox and document folder now count as data you can use.
| Structured | Unstructured |
|---|---|
| Rows and columns Databases, clean spreadsheets, transaction lists | Ordinary language Emails, reports, chat logs, contracts, notes |
| The tidy minority What older analytics tools needed first | The vast majority Now readable directly by generative AI |
Key Takeaway
Seeing data everywhere is where this starts. What comes next is judgement: deciding whether a given source is good enough to act on.
Check your understanding
Self-check only. Nothing is scored and nothing is recorded; open an answer to compare it with your own.
A marketing manager says she cannot use AI for her work because 'we do not have a proper database, all our information is in emails, shared documents, and spreadsheets.' Based on what you learned, which response is most accurate?
- She is right. AI needs structured databases to work effectively.
- She should ask IT to migrate everything into a database first, then use AI.
- Emails, documents, and spreadsheets are data. Modern AI tools can process them directly, she already has everything she needs to start.
- She can only use AI for the spreadsheets, since those are the most structured format.
Show the answer
Emails, documents, and spreadsheets are data. Modern AI tools can process them directly, she already has everything she needs to start.
You are preparing for a quarterly business review and have the following information available: 50 customer support emails from the past month, a 12-page sales report in PDF, meeting notes from three team standups, and a spreadsheet of customer satisfaction scores. A colleague suggests picking just the spreadsheet for AI analysis 'because that is the real data.' What is the best approach?
- He is correct, the spreadsheet is the only proper data source. Use AI on that and summarise the rest manually.
- Use AI on all four sources. The support emails reveal customer pain points, the sales report shows trends, the meeting notes capture team context, and the spreadsheet provides quantitative scores. Together they give a fuller picture than any single source.
- The emails and chat logs are too informal to be useful for AI. Stick to the report and the spreadsheet.
- Upload everything into one massive document and let AI sort it out.
Show the answer
Use AI on all four sources. The support emails reveal customer pain points, the sales report shows trends, the meeting notes capture team context, and the spreadsheet provides quantitative scores. Together they give a fuller picture than any single source.
Saved in this browser only. Carry progress between devices with the resume link on the course page.