STAYCHARTED AMT / PRODUCT

How AI Model Trainer works

Turn labeled text or pictures into a model that applies your categories to new records. Prepare examples, train, check the results, and put the model to work.

Start free ↗Compare plans →

A file returned with predicted categories and confidence · Sample data

STEP 1

Upload examples with answers

Start with a spreadsheet of messages and approved categories. In our example, a support team teaches the model labels such as Billing, Account Access, and Feature Request. Select the answer column and the text the model should read. For pictures, use category folders or labeled picture links.

Choose the input columns and the answer column · Sample data

STEP 2

Check what the model will learn

Look for inconsistent labels and unhelpful inputs. Review detected personal information and decide whether to mask it or leave affected rows out. Pattern-based checks do not catch everything. For pictures, review duplicates and category conflicts before training.

Review detected personal information before training · Sample data

STEP 3

Choose a model and train

Start with a model available on your plan. A Classifier Model learns words; an AI Classifier Model also uses meaning; a Dedicated AI Model fine-tunes the AI itself. Your category definitions and examples matter whichever model you choose.

Choose the training approach available on your plan · Sample data

STEP 4

Read the validation report

AMT tests against held-out examples. Compare overall accuracy, performance by category, and the mistakes that matter. The score in this screenshot belongs to its sample dataset; it is not a promise for your data. Correct labels or add examples, then retrain when needed.

Check accuracy by category and inspect mistakes · Sample data

STEP 5

Publish and put it to work

Publish the version you want to use. Fill a file with new messages, try a record, or call the model through the API on a supported plan. Keep a review path for uncertain or consequential results. Use feedback to improve the next version.

Download new rows with suggested categories · Sample data

Build with your team, improve over time

Roles for the work

Give Builders, Reviewers, and Viewers access appropriate to their tasks.

A record of changes

Use Activity to review changes to your workspace and models.

Your next version

Correct examples, retrain, and check the report before publishing a replacement.

Try it on your own examples.

Start with Free for text classification. Compare plans for AI models, Vision, and API access.