Lesson 1. What Is a "Database" in Plain Terms#
Goal: understand why agents need data and how to store it without programming.
Why AI Agents Need Data#
An AI agent doesn't work only with text from the prompt. It often needs to:
- read a list of customers or products
- write results (leads, responses, orders)
- update task status or a counter
- compare incoming requests with a knowledge base
For this you need a data storage — a place the agent can access to read or write information.
What a Database Means for Non-Technical Users#
A database is an organized storage of information where:
- data lives in tables (rows and columns)
- each row is one record (customer, product, order)
- each column is one field (name, email, status)
- tables can be linked to each other (customer → orders → products)
For non-technical users, a database is a smart table you can connect to via API or integration.
Example Data Tasks#
| Task | What the agent does | Where data comes from |
|---|---|---|
| Support bot | Finds answers in a knowledge base | "Q&A" table |
| Lead qualification | Checks if the customer existed before | "Customers" table |
| Meeting scheduler | Checks available slots | "Calendar" table |
| Segment-based campaigns | Selects recipients by criteria | "Contacts" table |
| Order reports | Calculates totals by status | "Orders" table |
Which Tools Work for Non-Technical Users#
Three popular tools don't require SQL or programming:
- Google Sheets — familiar spreadsheets, has API, free, easy to integrate
- Airtable — database with table interface + links, forms, views
- Notion — database + documents, flexible views, API for integrations
All three can be connected to an agent via API or ready-made connectors (Zapier, Make, n8n).
When to Use Which#
| Tool | When to use | Limitations |
|---|---|---|
| Google Sheets | Simple lists, quick start, familiar interface | Slow with >10,000 rows, no links between tables |
| Airtable | Complex structures, links between tables, input forms | Paid beyond 1,200 records, slow API |
| Notion | Hybrid format (database + documents), team collaboration | Slow API, no bulk operations |
Recommendation: start with Google Sheets for your first agent, then try Airtable for more complex tasks.
The Golden Rule of Data for Agents#
Data must be clean and predictable.
Agents can't "guess." If one row has status "Done," another "done," and a third "Completed" — the agent treats them as three different values.