RadGPT · Beta

The agentic AI layer for radiology.

RadGPT gives every practice access to Anthropic, OpenAI, and other frontier LLMs to automate clinical reads, operations, and IT. One hub. Only possible on Sirona.

See Sirona in Action

How Sirona is Different

An agentic orchestration layer, not a chatbot

RadGPT is not a chat window bolted onto a PACS. It's an orchestration layer that gives large language models access to everything in the radiology workflow — images, reports, clinical history, worklist state, and viewer context. Agents can act asynchronously when a case is assigned, when a study is opened, or on a schedule. They can act in real time while a radiologist dictates. A single RadGPT agent can coordinate hundreds of narrow clinical AI algorithms underneath it, funneling their outputs into one coherent interface.

What RadGPT can do

A single agentic layer that spans clinical decision support, operational automation, and IT management — powered by the LLMs you choose, running on the data only Sirona has.

Clinical AI Agents

Draft impressions, summarize priors, flag discrepancies, and coordinate narrow AI algorithms — in real time during the read or asynchronously before the radiologist opens the case.

Explore AI

Operational AI Agents

Monitor turnaround times, rebalance worklists, generate performance reports, and surface bottlenecks — natural language queries against your practice data.

Learn more

Voice-Activated Agents

Trigger AI actions with voice commands during dictation. Ask a question, request a comparison, or kick off a workflow — without touching the keyboard.

Speech-to-Action

Multi-Model Access

Access Anthropic, OpenAI, and other foundational models through a single interface. Choose the right model for each task — no vendor lock-in on the AI layer.

Learn more

Secure & HIPAA Compliant

Every agent runs within Sirona's HIPAA-compliant infrastructure, backed by an active SOC 2 program. PHI never leaves the platform. Full audit trail on every AI action.

Security

Practice-Configurable Automations

Build your own AI workflows — triggered on case assignment, on study open, on dictation events, or on a schedule. Your practice defines what gets automated and how.

Amplify SDK

Agentic AI that acts, not just answers

Clinical

Hundreds of AI algorithms, one coherent interface

The radiology AI market has hundreds of narrow algorithms — lung nodule detection, bone age assessment, stroke triage. RadGPT coordinates them. Instead of managing dozens of vendor integrations and separate result viewers, a single RadGPT agent routes the right algorithm to the right study, collects outputs, and presents a unified summary in the report.

Automatic algorithm routing based on study type and clinical context

Unified results presentation in the reporter — no separate viewer

Configurable confidence thresholds and escalation rules

Full audit trail on every AI recommendation

Operations

Ask your practice a question. Get an answer.

What's our average turnaround time for chest CTs this week? Which radiologist is reading the fastest on overnight shifts? Are we trending toward an SLA breach on the emergency worklist? With RadGPT, you ask in plain English and get an answer backed by real-time data from across the platform.

Natural language queries against live practice data

Automated daily and weekly performance summaries

Proactive alerts when metrics drift outside acceptable ranges

Exportable insights for board meetings and payer negotiations

Architecture

Why the unified platform makes agentic AI possible

An AI agent is only as useful as the data it can access and the actions it can take. On a fragmented stack, an agent would need to authenticate with four systems, reconcile four data models, and coordinate actions across four APIs. On Sirona, the agent has native access to images, reports, clinical data, worklist state, and viewer context through a single unified layer.

Single data layer spanning images, reports, worklist, and clinical context

Native access — no integration layer, no API coordination

Real-time event stream for triggering agents on workflow events

Cloud-native compute for running models at scale

Domains

Three radiology domains unified by one agent layer

RadGPT spans clinical, operational, and IT work in a single agent fabric. Today every one of those domains uses a different tool, a different vendor, a different data format. With RadGPT, the practice asks a question or sets an automation, and the agent figures out which domain it belongs to — pulling data and taking action wherever it needs to.

Clinical: impressions, priors, discrepancy flags

Operations: turnaround, SLA, staffing questions

IT: system health, integrations, anomalies

Every agent governed by one admin framework

The agentic AI advantage

domains unified — clinical, operations, IT

100s

of narrow AI algorithms coordinated by a single agent

platform with all the data AI needs to act

The foundation for agentic AI

Why Agentic AI Requires RadOS

Watch as Dr. Mark Longo demonstrates the power of a new paradigm in radiology AI. Welcome to the age of agentic, embedded, multimodal, real-time, clinically aware AI assistants.

Dr. Mark Longo

Chief Technology Innovation Officer, Sirona

Read the partnership announcement

“Through Sirona's platform, Everlight will be able to build and deploy AI-powered automations across clinical, administrative, and operational workflows. This partnership will fundamentally transform how our radiologists practice, and position Everlight at the forefront of AI-enabled diagnostic medicine globally.”

Jeff Oakman

Global Chief Operating Officer, Everlight Radiology

FAQs

What is an agentic AI orchestration layer?

An orchestration layer is the software that lets AI agents — programs powered by large language models — access data, take actions, and coordinate workflows. Instead of generating text in isolation, an agent can read images, query reports, check the worklist, call narrow AI algorithms, and write results back into the system. RadGPT is the orchestration layer for radiology.

What is RadGPT?

RadGPT is an agentic AI orchestration layer built on Sirona's unified radiology platform. It gives practices access to powerful large language models — including those from Anthropic, OpenAI, and others — to build AI automations that span clinical reads, operational workflows, and IT management. Currently in beta.

Which LLMs does RadGPT use?

RadGPT runs on Anthropic, OpenAI, and other frontier models — chosen per task. You aren't locked into one vendor; the orchestration layer picks the model that fits the agent.

Is this HIPAA compliant?

Yes. All RadGPT agents run within Sirona's HIPAA-compliant environment, backed by an active SOC 2 program. PHI never leaves the platform, and every AI action is captured in the unified audit trail.

Can we build our own automations?

Yes. Practices configure their own RadGPT automations — trigger conditions, data access scopes, and agent behavior — through the admin console or the Amplify SDK.

Does RadGPT replace Platform AI features?

No — it sits on top of them. Auto-Impressions, QualityAssist, Spine Suite, and other Platform AI features are the building blocks that RadGPT agents orchestrate.

When will it be generally available?

RadGPT is in private beta with select customers. Broader GA timing is announced on sirona.com and through your account team.