Google Cloud's new AI agents move from demos to clinical impact
Healthcare leaders are testing where AI can remove friction from care delivery. This week, several organizations showed concrete wins using Google's latest AI tools and agent frameworks across EHR workflows, screening access and clinical research operations.
According to Google Cloud leaders, top healthcare AI use cases today include tech support (53%), security operations (49%), productivity and research (46%) and patient experience (44%). Three deployments stand out for practical value.
Hackensack Meridian Health: faster patient communication and clinician support
- GenAI inside Epic: Using Google Gemini, 12 specialties now access note summarization and other genAI capabilities framed directly within Epic. Integration focuses on keeping tools inside existing clinical workflows.
- Time back for specialists: Early results show a 5%-20% reduction in time spent on EHR tasks across specialty staff.
- Lab summarization agent: Summarizes lab panels, flags trends and key findings and drafts preventive recommendations for PCP messaging. Patients receive clearer explanations faster, enabling quicker follow-up and preventive actions.
- NICU Nurse Agent: Rapid access to current NICU best practices and policies gives nurses on-demand guidance at any workload or experience level.
- Prediction models: Additional models are being built and integrated into the EHR to support value-based care objectives.
Color: closing the mammography gap with an AI assistant
- Focus: 20%-30% of eligible women in the U.S. are behind on mammograms. Diagnosis rates in women under 50 have risen nearly 20% since the early 2000s.
- How it works: Built on Vertex AI and powered by the Gemini 2.5 model family, Color Assistant determines eligibility, schedules screenings and coordinates follow-up through EHR integrations. The initiative runs through Dec. 31.
- Clinical oversight: The agent follows evidence-based guidelines (e.g., American Cancer Society) with reviews by clinicians within Color's 50-state medical group. Care teams contact patients for clarifications, coordinate appointments and can order appropriate imaging per guidelines.
- Closed-loop communication: Color's clinicians deliver results directly to patients and coordinate next steps with existing providers. The agent also checks transcripts to support quality.
Castor: automating enrollment and consent in clinical trials
- The platform: Castor's Catalyst AI platform, built on Google Cloud, automates data entry, verification and other administrative tasks across studies via event-driven data infrastructure.
- FHIR-based consent: Patient enrollment and consent are accelerated through digital EHR integrations. The agent requests and retrieves patient consent using the HL7 FHIR standard.
- Regulatory posture: Full observability and auditability are core design goals. Castor emphasizes that reliable AI depends on a clear, detailed view of study operations; the platform took 18 months to build.
Why this matters for healthcare leaders
These deployments focus on speed, clarity and clinical oversight. They keep AI inside the EHR, shorten time-to-information, and create closed-loop processes that remove handoffs and delays.
The pattern is consistent: use AI to reduce friction in high-volume workflows (documentation, results delivery, scheduling), maintain clinician review where risk is higher, and instrument everything for audit and quality control.
Practical steps to get started
- Pick one friction-heavy workflow (e.g., lab result messaging, appointment scheduling, prior auth) and define success metrics (turnaround time, staff minutes saved, patient response time).
- Embed within the EHR: prioritize native frames, in-context side panels and SMART-on-FHIR apps to avoid new toggles.
- Establish a clinical oversight model: define when AI drafts vs. when clinicians must review and sign off.
- Use standards: rely on FHIR for data exchange and logging; set up audit trails from day one.
- Close the loop: ensure the agent routes results, schedules follow-ups and documents outcomes so nothing stalls.
- Track equity: monitor uptake and outcomes across demographics to ensure access improves for all populations.
Bottom line
Health systems are moving from pilots to measurable impact: fewer clicks, faster patient communication and cleaner study operations. The common thread is simple-meet clinicians where they work, keep patients in the loop and instrument the process for safety and compliance.
If your team is building skills for these deployments, explore curated AI learning paths by role at Complete AI Training.
Your membership also unlocks:
