Intent Classification

The Codeq API labels a piece of text with the intent behind it, such as a request, a complaint, or a scheduling question. A Beam agent reads that label as messages arrive, applies the customer's approved rule for what happens with each intent, and routes the message to the matching queue, person, or workflow. It can also tag the source record so the classification is visible later without re-reading the text. Messages with a low confidence score, or ones that mix more than one intent, are routed to a human to sort.

Entity Recognition

The API pulls out named entities from text, such as a company, a date, or a location, without a person reading the message first. A Beam agent reads the entities Codeq returns, applies the approved rule for the field they map to, such as a date becoming a calendar entry, and updates the connected record. Where permissions allow, it can also flag when an expected entity, like a required date, is missing from the text. Entities that are ambiguous, misspelled, or contradict an existing record are routed to a human to confirm.

Sentiment Scoring

Codeq scores text on a sentiment scale, giving a read on whether a message is positive, neutral, or negative. A Beam agent reads that score for incoming messages on the schedule or trigger the customer sets, applies the approved rule, for example escalating anything below a set threshold, and notifies the relevant team. It can also add the score to a record for later reporting so the trend is visible without manual review. Borderline scores near the threshold, or messages with mixed sentiment, are routed to a human to read directly.