We connect assistant and automation features in your system to OpenAI models through the OpenAI API.
Language models are good at reading and writing — summarising a long thread, drafting a reply, answering a question in plain words or pulling fields out of a scanned document. Built into your own software, these features work on your business records, inside the screens your team already uses, with you deciding what data the model may see.
What we build with it
- Answering questions about business data in plain language
- Summarising documents and communication threads
- Drafting messages and document text
- Extracting structured data from unstructured input
How it works
- A user asks a question or clicks an action such as summarise or draft reply.
- The system gathers only the records that user is allowed to see and that are needed for the task.
- It sends the instruction and that context to the OpenAI API from the server.
- The response is shown to the user, who reviews, edits or discards it before anything is saved or sent.
Data exchanged
- The prompt and whatever business context is included in it
- Generated response
- Token usage for billing
What is needed to set it up
- An OpenAI API account
- API key configured server-side
- An explicit decision on what data may be included in prompts
Good to know
- Usage is billed by OpenAI to your own account, based on the amount of text processed. Rates are set by OpenAI and vary by model.
- OpenAI's published API terms describe how business data is handled and retained. Review the current terms against your own policies before sending customer data.
- Models can produce confident but wrong answers, so a person reviews anything that affects customers, money or legal documents.
