ScienceSoft Global Menu icon ScienceSoft Global

Careers

For journalists

email contact@scnsoft.com en flag +1 214 306 68 37
en flag +1 214 306 68 37
Contact us

Q3 2025 Healthcare AI Trends: AI for Administrative and Clinical Workflows to Become a Default Budget Line in 2026

Vadim Belski

Head of AI, Principal Architect, ScienceSoft

Gala Batsishcha, MD

Healthcare IT Consultant, ScienceSoft

Named among Top 100 Women in Health IT 2025 by Becker’s Hospital Review

Tanya Filippova

Healthcare IT Consultant and Business Analyst, ScienceSoft

Vital Kavalenka

Compliance Officer, Healthcare and Finance IT Compliance Consultant, ScienceSoft

Healthcare IT
Artificial Intelligence
Last updated:
6 min read

With over 20 years in healthcare IT, ScienceSoft builds AI-supported solutions that deliver measurable value, easing the administrative burden and supporting safer clinical workflows. Drawing on findings from our Q3 2025 Healthcare IT Market Watch and recent project work, we outline the most important healthcare AI trends today and what to expect over the next 12 months.

Administrative AI scaled faster than clinical AI

In short: In Q3 2025, AI for administrative and clinical workflows scaled, while AI for clinical decision support and diagnosis remained in pilot and made only cautious progress. Given tight capital, buyers funded cost-saving use cases with a 6–12-month payback.

Expect AI for administrative and clinical workflows to become a default budget line in 2026, while clinical decision support will grow mostly in explainable, low-liability domains" — Vadim Belski, Head of AI, Principal Architect, ScienceSoft.

AI for administrative and clinical workflows scaled faster than AI for decision support and diagnosis

In Q3 2025, AI was primarily used in healthcare to streamline operations rather than replace or enhance clinical decision-making because the payback was measurable and governance was simpler. According to PwC, ambient scribe intelligence and AI-powered revenue cycle management were in the scaling and adoption phase, while diagnostic support and clinical workflow automation were still in the pilot phase.

Tech leaders set the pace. Oracle Health debuted its AI-first, voice-first EHR, featuring orchestrated agents that surface context and suggest actions. Epic launched three agents: Art for clinicians, Emmie for patients, and Penny for revenue cycle management. Their goal was to optimize three high-friction areas simultaneously: point-of-care support, patient communications, and back-office operations. Epic also partnered with Talkdesk, an AI-powered contact center platform, to bring the Advanced Dialer capabilities to Epic customers. In this setup, Epic surfaces patients who need appointments and triggers outreach. If there is no response, Talkdesk automatically places follow-up calls with the outcome written directly into the patient's Epic record.

On the inbound side, NextGen Healthcare released Navigator. This AI customer service agent handles patient calls and portal messages and reportedly saves front-desk teams two to three hours per day while improving access and show rates.

Specialist vendors kept up. Pieces Technologies announced a mobile voice assistant for inpatient physicians that can turn a 30–45-second dictation into a complete progress note with no after-hours typing. VSee Health integrated its new AI Doctor Notes, an automated tool that drafts encounter notes from virtual visits, with Tele911, the nation's largest telemedicine ER-diversion service. It cut physicians' charting time by 93% from 15 minutes to under 1 minute, boosting visit capacity.

Providers actively adopted such solutions, reaping their benefits. For example, Valley Children's Hospital implemented ambient documentation that listens during the visit and drafts the note for physician review in the EHR. After successful physician pilots, the program began expanding into nursing documentation, starting with vitals.

Capital flows reinforced this pragmatism. Although healthcare fundraising was on track toward a 10-year low, healthtech AI took a strong lead in attracting investment, according to Silicon Valley Bank. Functions like revenue cycle and patient communications were most attractive because ROI and performance metrics were clear. Macro pressures pointed in the same direction: PwC projected the medical cost trend to hold at 8.5% for group plans and 7.5% for individual plans in 2026, pushing providers and payers to do more with less.

Legislation also encouraged the adoption of AI for administrative functions. For example, H.R. 5045, the HEALTH AI Act, supports research on generative AI tools for use cases such as ambient documentation, claims processing, and patient communication. And states like Illinois limited AI only to administrative/assistive roles, prohibiting its use in providing mental health and therapeutic decision-making.

By Q2 2026, ambient/voice + agentic copilots will be a default requirement in enterprise provider RFPs. Vendors without native or partner-powered capability will struggle to make shortlists," — Vadim Belski, Head of AI, Principal Architect, ScienceSoft.

AI for clinical decision support made cautious progress

In Q3 2025, clinical decision support and diagnostic AI made cautious progress. Survey data show that while 48% of clinicians have used AI tools, only 16% use them to help make clinical decisions, underscoring that AI is still treated as an adjunct rather than a routine part of care. The most visible GenAI step came from Epic, which introduced Comet, a new AI tool designed to help clinicians anticipate patient needs and improve decision-making. The solution uses generative models trained on more than 100 billion medical events from Epic's Cosmos database. By simulating possible patient outcomes, Comet aims to give care teams data-driven insights into what may happen next during a patient's journey.

Still, hospitals mostly adopted solutions based on traditional ML for clinical decision support and diagnostics. For example, Advocate Health deployed imaging AI via Aidoc's platform, enabling radiologists to receive earlier alerts and faster handoffs within their current workflow. And Lucem Health announced Reveal for T1D, a new AI-powered solution that analyzes routine EHR data to flag patients likely to develop Type 1 diabetes, enabling earlier outreach through standard care processes.

Research advanced mostly with traditional ML, too. For example, researchers at NewYork-Presbyterian and Columbia University developed EchoNext. This AI-powered screening tool analyzes ECG data to identify patients who should undergo further testing with an echocardiogram. And scientists at Mount Sinai created an AI system that can predict how likely rare genetic mutations are to actually cause disease. By combining machine learning with millions of electronic health records and routine lab tests like cholesterol or kidney function, the system produces "ML penetrance" scores that place genetic risk on a spectrum.

In 2026, traditional predictive ML fueled by privately developed models will remain the backbone of clinical decision support and diagnostic AI. Such models rely on proprietary clinical datasets and long-term optimization. However, open-source LLM providers will continue to push in this direction, advancing medical-grade reasoning and incorporating subject-matter expertise into model training. In the future, this will allow providers of all sizes and budgets to build AI-assisted clinical support tools," — Vadim Belski, Head of AI, Principal Architect, ScienceSoft.

References

Want to go deeper?

Speak with ScienceSoft's healthcare AI experts to source quotes or data for your story.

Request expert input