What is machine learning?
Updated · By Robert Breen
Machine learning is a way of building software that learns patterns from examples instead of following rules a programmer wrote by hand. Shown enough data, it can recognize, sort or predict new cases it has never seen. It is the method behind almost all modern AI.
Why it matters for a small business
Think about writing rules to sort your business expenses. "If the vendor is Staples, it is office supplies" works until a new vendor shows up, or Staples sells you a printer that should be equipment. A rule list never ends. Machine learning avoids that by learning what office supplies, software and meals tend to look like, so it can handle vendors nobody listed.
You rarely train a machine learning model yourself as a small business. You use one someone else trained, through ChatGPT or an AI node in n8n, and steer it with instructions. Knowing that the model learned from patterns, not rules, explains both its strengths (handling variety) and its weak spots (occasional confident mistakes on unusual cases).
In a real lesson: Build an AI Agent That Categorizes Business Expenses
In the AI expense categorizer lesson, you build an n8n agent for Maple Street Bookkeeping, a made-up bookkeeping firm. The system prompt lists the only categories allowed: Office Supplies, Software & Subscriptions, Meals, Travel, Utilities and Needs Review. It never says which vendor belongs where.
You then paste three expenses: Corner Office Supply for printer paper and toner, CloudLedger for monthly accounting software, and Harbor Street Cafe for lunch with a client. The agent files them as office supplies, software and a meal without a single vendor rule, because the model underneath learned what those things look like from enormous amounts of text.
The prompt also covers the weak spot: "If an expense could fit more than one category, or the details are unclear, use Needs Review and say why." That is how you work with a learned model. Let it handle the common cases, and give it a safe place to put the ones it is unsure about, for a bookkeeper to check.
Try this lesson free or read the step-by-step guide.
Common confusions
Machine learning vs AI
AI is the goal: software that does tasks needing judgment. Machine learning is the main method used to get there today. Nearly every AI tool you use, including large language models, was built with machine learning.
Does ChatGPT learn from my chats as I go?
Not in the moment. The model's learning happened during training, before you used it. Within a chat it follows your instructions and earlier messages, but it is not retraining itself. Whether your chats are later used for training data depends on your plan and settings.
Tips
- Give a learned model a fixed list of allowed answers, like the category list, so its output stays usable.
- Always include an escape hatch such as Needs Review for cases it should not guess.
- Spot-check the unusual cases first; those are where learned patterns are most likely to slip.
Related terms
Where you use it: free lessons
- Build an AI Agent That Categorizes Business Expenses (n8n, 12 min)
- AI Receipt Extractor: Receipts to Google Sheets with n8n (n8n, 12 min)
Frequently asked questions
- Do I need to train my own machine learning model?
- Almost never. Small businesses use models that are already trained, such as the ones behind ChatGPT, and shape them with instructions, examples and their own facts.
- Is machine learning the same as deep learning?
- Deep learning is one kind of machine learning that uses large neural networks. Large language models are built with deep learning, so they are machine learning too.