2026: When clinical intelligence moves from promise to practice

Tom Barry (Novartis) argues 2026 will be the year clinical intelligence moves from foundation-building to practical deployment across three AI-driven shifts: agentic autonomy, ambient observation, and a reaffirmed human premium, with direct implications for payers, health systems and life sciences firms. He notes that 71% of hospitals reported predictive AI integrated into EHRs in 2024 (up from 66% in 2023) and 92% use it to forecast inpatient trajectories, cites an October 2025 review finding ML can predict sepsis earlier, references FDA 2024 guidance on digital endpoints and six states limiting insurer AI denials, and warns of discriminatory or erroneous outcomes (including reports of AI prior-authorization tools producing denial rates 16x higher than humans). The piece urges pharma to build agentic market-access capabilities, ambient data partnerships for real-world evidence, and robust governance to protect patients and revenue.

Analysis

Market structure: Winners will be AI infrastructure and data-integration providers (NVIDIA, ORCL, MSFT, AAPL, select MedTech like MDT/ABT) and pharma with strong real‑world evidence (RWE) pipelines (e.g., NVS), because compute, cloud and sensor demand should rise 20–40% CAGR in the next 2–3 years. Losers in the near term: low‑value primary care incumbents and legacy payers that rely on opaque denial algorithms — pricing power shifts to platform owners and life sciences firms that control data flows and patient‑support agents.

Risk assessment: Tail risks include regulatory curbs on insurer AI and class actions from biased algorithms — a trigger would be >10 US states or a federal rule in 12 months halting automated denials, which could cut projected cost savings by >50% and force rework. Short term (days–weeks) watch for legislative headlines and 8‑K partnership announcements; medium (3–12 months) for pilot outcomes and hospital deployments; long (12–36 months) for market consolidation, large fines or breach events.

Trade implications: Over 6–24 months, overweight software/infrastructure and sensor makers; underweight pure-play payers that lack transparency. Use options for asymmetric exposure to adoption inflection (6–12 month call spreads on NVDA, LEAP calls on ORCL). Seek pair trades that long EHR/cloud (ORCL) vs short insurers with high automation risk (UNH) sized to correlation and regulatory signals.

Contrarian angles: Consensus underrates the “human premium” — high‑touch specialties (neuro, post‑acute rehab providers) could command price premiums as AI commoditizes routine care; consider select longs in service providers (HCA/AMED) with 12–36 month horizons. Also beware of an overenthusiastic privacy tradeoff: a major breach or public backlash could slow ambient adoption fast, creating a buying opportunity in infrastructure names priced for perfection.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (or equivalent 6–12 month 1:1 call spread) to capture data‑center AI demand; target +25–40% upside, use a 20% trailing stop or roll if compute revenue growth <20% QoQ over two consecutive quarters.
• Establish a 2% long position in ORCL (12–24 month horizon) to play EHR/cloud monetization from Cerner integration and hospital ambient deployments; add another 1% if ORCL announces ≥2 systemwide hospital partnerships in next 6 months.
• Establish a 2–3% long position in NVS (6–18 months) to capture advantage from RWE and patient‑support agents; scale in on signed health‑system ambient data partnerships, trim to half if regulatory fines or litigation exposure >$500m is disclosed.
• Implement a 1–2% pair trade: long ORCL (overnight) vs short UNH (short or buy 9–12 month put spread sized to beta) to express software winners vs insurers exposed to agentic denials; unwind or reduce the short by 50% if <5 US states pass insurer AI restrictions in the next 90 days or if UNH reports >$500m realized AI savings next quarter.