AI content detection

Winston AI returns a probability that a piece of text was machine-generated. When a document reaches your workflow, a Beam agent sends the content to Winston AI, reads the detection score, and applies your approved threshold to label it human, AI, or uncertain. It writes the score and label back to the record and notifies the owner when a piece exceeds your limit. Clear passes and clear fails are handled automatically. Scores near your threshold, or content the detector flags as low confidence, are routed to a reviewer who makes the final call.

Plagiarism checks

Alongside AI detection, Winston AI can assess text originality. As submissions arrive, a Beam agent passes each one through the check, reads the originality result, and applies your rule for acceptable overlap. It records the outcome against the submission and can notify the author or editor when a piece crosses your threshold. Submissions that come back clean move forward without review. Matches that are high, ambiguous, or tied to a sensitive assignment are handed to a person. Someone examines the flagged sources before any decision about the work is finalized on your side.

Batch scoring

Teams often need many documents checked at once. On a trigger or schedule, a Beam agent reads a queue of pending items, submits each to Winston AI, and applies your threshold to sort results into pass, fail, and review buckets. It writes every score back to the source record so nothing is lost. The bulk of items resolve to a clear bucket automatically. Anything landing in review, or any item the API could not score, is surfaced to a person with its details. Your team clears the review bucket before results are treated as final.