Install
Download the Package and follow the instructions to install it in your AAE Control Room.
OpenAI Decisions API Package
Automate decisions in bots: classify, choose and score text or images with OpenAI, returning structured results with confidence and rationale.
Top Benefits
- Decisions as data: get a yes/no, option or score in a dictionary your bot reads directly, with no text parsing.
- Confidence you can act on: use the confidence value in an If condition to send uncertain cases to a person.
- Rationale for every result: keep the model’s explanation with each decision for review and audit.
- Several questions in one call: Evaluate runs predicate, choice and score on the same input in one request.
- Works with text and images: send email text, ticket content or a JPG or PNG image as the input.
Actions
- Predicate: Answers a yes/no question about the input and returns the result with confidence and rationale
- Choice: Selects the best-matching option from a list of options and descriptions you supply
- Score: Rates the input on a numeric scale you define and returns the score with a rationale
- Evaluate: Runs predicate, choice and score questions on the same input in a single call
he OpenAI Decisions API package lets Automation 360 bots make decisions about text and images using the OpenAI Responses API. Each action sends your content to OpenAI along with an instruction and criteria you define, and returns a structured Dictionary that a bot can use directly in If conditions, routing and reports. Every result includes a confidence value and a rationale, so bots can send uncertain cases to a person and keep an explanation for review.
Actions included:
1. Predicate: answers a yes/no question about the content, with optional criteria that explain what counts as yes.
2. Choice: selects the best-matching option from a list you define, with a short description for each option.
3. Score: rates the content on a numeric scale you define, with a description for each level.
4. Evaluate: runs any combination of predicate, choice and score questions on the same content in a single call.
The content can be entered as text or supplied as a JPG, JPEG or PNG image. You can set the OpenAI model, the organization ID and the connection and response timeouts, and pass a state value through to the output to match each result to its source record.
Typical uses include routing emails and tickets to the right team, rating urgency or sentiment, checking whether a message states a deadline or contains a required clause, approving or flagging items against your criteria, and sending low-confidence decisions to a person for review.