What is text classification with AI?
Updated · By Robert Breen
Text classification is sorting a piece of text into one of a fixed set of labels, for example sorting a service request into Repair, Maintenance, Install or Emergency, or a lead into Hot, Warm or Nurture. With AI, you describe the labels in a prompt and the model picks one for each item.
Why it matters for a small business
Sorting is one of the most useful jobs AI does for a small business, because the label decides what happens next: who gets the job, how fast someone calls back, which email the customer receives. Done by hand, it waits until someone reads the inbox. Done by an AI step in a workflow, it happens the moment the form is submitted.
It used to require training a custom model on thousands of labeled examples. Now a clear prompt with well-defined labels gets good results for everyday business categories. The work shifts to defining the labels precisely, especially the edge cases that matter most.
In a real lesson: Build an AI Service Request Form for Your HVAC Company
In Build an AI Service Request Form for Your HVAC Company, homeowners fill in a Lovable form for Cedar Ridge Heating & Air (a made-up company) with Name, Email, What do you need? and Phone. The form posts to an n8n webhook, and the AI Agent's prompt template passes along the request text with {{ $json.body.request }}.
The system message does the classifying: sort each request into one job type, Repair, Maintenance, Install or Emergency. Then it defines the label that matters most: no heat, no cooling in extreme heat, or water leaking from the system counts as Emergency. The agent saves the lead to Google Sheets and emails a confirmation that names the job type.
The demo request form lesson uses the same pattern with different labels: Hot means 5 or more staff and wants to switch within a month, Warm means a clear need but no timeline, Nurture means just looking, and each label triggers a different next step in the email.
Try this lesson free or read the step-by-step guide.
Common confusions
Classification vs extraction
Classification picks one label from a list you define. Data extraction pulls specific values (a date, a total) out of the text. Many workflows do both: extract the details, then classify the request.
Classification vs triage
Triage is the business process of deciding urgency and routing. Classification is the AI step that often powers it, by assigning the priority or category label.
Tips
- Define every label in one plain sentence, and spell out the edge cases that are costly to miss.
- Add a fallback label (like Needs Review) for anything that doesn't fit.
- Test with borderline requests, not just obvious ones, before you trust the labels.
Related terms
Where you use it: free lessons
- Build an AI Service Request Form for Your HVAC Company (Lovable and n8n, 15 min)
- Build an AI Demo Request Form That Qualifies Leads (Lovable and n8n, 15 min)
- Build an AI Maintenance and IT Request Form (Lovable and n8n, 15 min)
Prompt templates that use it
Frequently asked questions
- How many categories can AI classify into?
- There's no fixed limit, but a short list of clearly different labels works best. If two labels are easy to confuse, merge them or define the difference.
- Can AI classification make mistakes?
- Yes, especially on vague or unusual messages. That's why safety-critical labels like Emergency deserve explicit rules and a person checking the results.