Azure OpenAI

Run OpenAI models hosted in your own Azure tenant: chat and reasoning, embeddings, DALL-E and gpt-image-1 image generation, and Whisper and TTS audio, all through Azure deployments.

8 actions available

New RFP document uploaded to the shared drive

Agent chunks the RFP and creates embeddings for retrieval

Agent asks a chat deployment to draft answers grounded in prior responses

Agent enforces a structured format on the draft

Agent assembles the draft response document

Proposal lead gets the draft with source references

The proposal lead reviews and edits before the response is sent

What This Integration Enables

Azure OpenAI gives an agent the OpenAI model family without leaving Azure. Chat and reasoning deployments handle generation and structured output, embeddings power retrieval over your own content, and the image and audio deployments cover generation and transcription, all under your tenant's identity and network controls. Keeping inference inside Azure answers the first question a regulated team asks about putting a model in production: where does the data go. The answer is nowhere it does not already go. What makes it operational is the layer that turns a model call into a repeatable process and inserts a person at the review point. An orchestration layer owns that, and FlowRunner is built for it.

Without FlowRunner

Models outside the tenant Inference runs on a vendor outside Azure governance

Blank-page drafting Every RFP answer written from scratch under deadline

No reuse of prior work Past winning answers live in scattered documents

With FlowRunner

Inference inside Azure Model calls run under your tenant, region, and policies

Grounded first drafts Answers drafted from your own prior responses

Institutional memory reused Embeddings surface the best past answer for each question

Use Case Scenarios

RFP Response Drafting

A new RFP arrives. The agent embeds the questions, retrieves the best matching answers from your library of prior responses, and asks an Azure OpenAI chat deployment to draft grounded answers in a structured format. It assembles a draft response with source references. The proposal lead edits a grounded first draft instead of writing from a blank page, and the whole process stays inside the Azure tenant.

Internal Document Q&A

Employees need answers from internal documentation that cannot leave the tenant. The agent embeds the corpus, retrieves the relevant passages for each question, and asks a chat deployment for a grounded answer inside Azure. Answers the model is confident about return directly; questions touching sensitive policy are routed to the owning team instead.

Structured Extraction Pipeline

Incoming documents need specific fields pulled into a system of record. The agent sends each document to a chat deployment with a schema and gets back structured JSON. Records that pass validation flow straight through. Anything the model returns with low confidence or a validation failure is held for a person, so the pipeline never writes malformed data downstream.

Human-in-Loop Highlight

A model drafting an RFP answer or extracting a contract field produces a proposal, not a final decision. When the output leaves the flow as something a customer or a system will act on, FlowRunner routes it through a [human-in-the-loop](/concepts/human-in-the-loop/) step: the agent pauses, presents the draft or the extracted data with its sources, and sends it to the responsible person via Slack or email. They edit and approve. The model produces the first pass inside your tenant; a person signs off before it goes out.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

8 actions

Chat

Ask AI Sends a single prompt (with an optional system instruction) to a chat model deployment and returns the generated text along with token usage.
Chat Completion (Advanced) Full-control chat completion against a model deployment: multi-turn messages, tool/function calling, structured outputs via response_format (including json_schema), sampling controls, reasoning effort for reasoning deployments (o-series / gpt-5), and Azure "On Your Data" grounding via data_sources.
Analyze Image Analyzes one or more images with a vision-capable chat deployment (e.g. gpt-4o) by sending the prompt together with image_url content parts.

Embeddings

Create Embeddings Generates vector embeddings for one or more texts using an embeddings deployment (e.g. text-embedding-3-small / text-embedding-3-large).

Images

Generate Image Generates images from a text prompt using an image deployment (dall-e-3 or gpt-image-1). Handles both API response shapes: when Azure returns a hosted URL (dall-e-3) it is passed through; when it returns base64 data (gpt-image-1) the image is uploaded to FlowRunner file storage and a hosted URL is returned.

Audio

Transcribe Audio Transcribes an audio file into text in its original language using a speech-to-text deployment (whisper or gpt-4o-transcribe).
Translate Audio Translates speech from an audio file in any supported language into English text using a whisper deployment.
Text To Speech Converts text into natural-sounding speech using a text-to-speech deployment (tts, tts-hd, or gpt-4o-mini-tts).

Start building with Azure OpenAI

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us