Templates / Enrichment
Classify
Ask one question of every row and get a typed answer — yes or no, one of your options, or a score — with a confidence value you can filter on.
What it isEnrichment
An enrichment column. It reads columns you already have and writes the answer beside them, plus a second column saying how sure it is.
InputWhat you give it
A question, and the columns it should read — a page title, scraped text, a job description.
OutputWhat comes back
- Yes or no, for a statement about the row
- One value from a list of options you define
- A score against a scale
- A confidence from 0 to 1 on every answer
| Company | Page title | Is a dental practice | Confidence |
|---|---|---|---|
| Dentavital | Dentavital - center ustnega zdravja | Yes | 0.80 |
| Artident | Zobozdravstvene storitve: implantati in protetika, All-on-4 | Yes | 0.79 |
| Lux dental | Lux Dental zobozdravstveni center - Zobozdravstvo | Yes | 0.84 |
| Sekulic | Domov - Zobozdravstvo in implantologija Sekulić | Yes | 0.81 |
| Skorjanc | Casibom - 2026 Güncel Erişimin… | No | 0.97 |
Real output. Five dental practices from a Google Maps search in Ljubljana, asked "This business is a dental practice" against each site's page title. Four are. The fifth practice's domain has been taken over by a casino site, and Classify said no with its highest confidence of the run.
How it worksFour things you decide
- Pick the kind of answerYes / no, pick one of your options, or a score. The kind is fixed, so every cell is an answer you can sort and filter — never a sentence.
- Write the questionFor yes / no, write a statement about the row: "This company sells to other businesses." For options, list them one per line as
value: what it means. - Point it at the columns to readType
/to insert a column — scraped text, a title, a description. Classify reads only what you give it. - Filter on confidenceKeep the rows the model was sure about and look at the rest by hand. That column is what an ordinary AI column cannot give you.
Limits — read this before you run it
Needs a TypeSafe API key in Settings; TypeSafe bills its own usage. It answers only the three kinds of question — for a summary or a written line, use Analyze and write. It reads the columns you point it at and never opens a page itself, so scrape first. Runs several rows at once; lower Parallel rows if your key hits rate limits.
FAQQuestions people actually ask
How is this different from Analyze and write?
Analyze and write writes free text: a summary, a category in its own words, an email line. Classify only ever returns one of the answers you allowed, plus how sure it is. Use Classify to qualify or sort rows, and Analyze and write to produce text.
What is the confidence column for?
Filtering. Set a threshold and keep the rows above it; the few below are the ones worth a human look. On a large list that is the difference between checking every row and checking a handful.
What does it cost?
Each row processed is 1 Manycrawl action. The model runs on your own TypeSafe key, billed by TypeSafe.