Blog Viewer

Building on our Q4 2025 update, welcome to the first quarter review of 2026 tracking the evolution of artificial intelligence through the lens of Pharmacy General Intelligence (PGI). As a reminder, PGI focuses on AI's potential to perform at or beyond the level of a pharmacist, paving the way for autonomous agents to handle medication verification, dose adjustments, and direct patient care.

If last year was about proof-of-concept, early 2026 is defined by integration and systemic restructuring.¹ AI is shifting from a new technology into expected, invisible infrastructure. Continuing our pattern here is a breakdown of the industry trends, technological advancements, and policy shifts shaping PGI, along with the challenges.

Industry Trends 

This quarter, the healthcare industry realized that standalone AI point solutions are not sustainable, giving way to integrated platforms. According to UCSF's Dr. Bob Wachter, this massive AI transformation is essential to save a healthcare system currently buckling under the weight of clinician burnout and severe staff shortages ².

Within the pharmacy itself, severe technician shortages are fueling a massive surge in automation³. Health systems are utilizing robotics and AI to manage tasks like centralized dispensing and IV workflows. This automation is a precursor to PGI, as it acts as a labor complement that frees human pharmacists to focus on clinical interventions and medication optimization³. Administrative operations are also being automated, with Optum deploying AI-powered tools directly within the EHR to handle complex prior authorizations, reducing administrative time and achieving a 96% first-pass approval rate⁴.

Integrating these tools remains a massive hurdle. Highly accurate AI models frequently die in the "invisible graveyard" of healthcare because they add additional steps to the clinical workflow⁵. To solve this, CIOs are pushing for embedded EHR AI solutions, acknowledging that native integration is the only way to earn clinician trust⁶. Epic is fully committed to this embedded approach; they recently rolled out "AI Charting" to draft notes and queue orders⁷, pushing their overall AI adoption rate to over 85% among customers⁸. At the enterprise level, Anthropic is targeting health systems directly with its HIPAA-compliant Claude for Healthcare⁹, helping summarize massive 300-page oncology charts without losing critical context¹⁰.  Oracle has also made significant gains with breaking down data silos and allowing clinicians better access to data for decision making.¹¹ This integration is a marked difference this quarter as companies have realized the importance of adding AI into existing workflows and existing software platforms.  PGI will likely be a standalone concept at first which will follow a similar pattern of eventual integration.

Discussions around AI workflow disruption have picked up. Research reveals that 6.1 million U.S. workers in clerical and administrative roles face high AI exposure but possess incredibly low "adaptive capacity" (such as transferable skills or liquid savings) to navigate job transitions¹². Consequently, labor unions are mobilizing. Kaiser Permanente workers, for example, have actively protested and proposed legislation against the encroachment of AI tools, citing fears of job displacement and the unauthorized recording of therapy sessions¹³.

Last, the democratization of medical information online has exposed a dangerous "credibility-evidence gap." A recent JAMA study found that 62.5% of medical claims made by credentialed professionals in online videos are supported by little to no evidence¹⁴. This professional "halo effect" allows misinformation to spread easily, serving as a stark reminder that as PGI scales, it must be rigidly anchored to empirical evidence rather than engagement algorithms.  This is an enormous opportunity for pharmacy as so much medical information involves medications and the verification of this information is a role that pharmacists are perfectly trained to perform.

Technological Advancements 

The technological leaps of Q1 2026 provide the clearest real-world previews of autonomous PGI yet. The most shocking development occurred in Utah, which launched a first-of-its-kind pilot program allowing an AI system (Doctronic) to autonomously review and renew routine prescriptions for chronic conditions entirely without human physician involvement¹⁵. 

Tech companies have started to release consumer-facing health agents. OpenAI introduced ChatGPT Health¹⁶, a move heavily bolstered by its reported $100 million acquisition of the startup Torch, whose team built a specialized "medical memory" context engine¹⁷. Amazon launched its Health AI agent, offering Prime members free virtual care routing and prescription management via Amazon Pharmacy¹⁸. Microsoft entered the market with Copilot Health, aggregating wearable data and records from over 50,000 hospitals in a push toward "medical superintelligence"¹⁹. Independent challengers also joined the party, with Superpower launching an "AI Doctor" featuring continuous lifetime memory and grounded clinical reasoning²⁰. Open-source networking tools like OpenClaw (first Clawdbot and then Moltbot before becoming OpenClaw) demonstrate how AI agents can operate as multi-channel gateways directly through everyday apps like WhatsApp, paving the way for direct AI-to-patient pharmacy communication²¹.  OpenClaw has been getting an incredible amount of attention in much the same way that OpenAI’s ChatGPT release did.  What makes OpenClaw so influential is the ease with which users can launch a multi agent platform and control it via apps that already exist.  Much like ChatGPT’s launch in Nov of 2022 made it easy for users to interact with a LLM, OpenClaw has created a similar scenario where it is now easy for users to interaction with a powerful multi agent syste.

Under the hood, the models powering these tools are advancing exponentially. The release of models like GPT-5.3 Codex and Claude Opus 4.6 demonstrated AI systems that possess "judgment" and can autonomously execute multi-day tasks²². As Anthropic CEO Dario Amodei notes, the rapid pace of AI development—where models are now writing the code to improve themselves—promises a golden age of scientific discovery, but it also brings autonomy risks and the potential for AI to lower the barrier for creating bioweapons, demanding extreme caution as we develop clinical AI²³.

Policy Shifts 

As PGI systems become increasingly capable of independent clinical reasoning, the regulatory and financial structures surrounding them must adapt. The current United States reimbursement models actively penalize providers for using time-saving AI. To make PGI financially viable, a framework proposed in NEJM Catalyst argues for a transition to output-based payment models, reimbursing physicians (and pharmacists) for the high-value clinical outputs generated by their AI agents rather than the time spent on the task²⁴.

On the regulatory front, federal policy is rapidly shifting. A recent executive order aimed at consolidating national AI oversight is forcing hospital IT leaders to pivot their vendor evaluations away from marketing hype and toward strict governance and clinical risk management²⁵. This internal focus became a necessity following the collapse of the Coalition for Health AI (CHAI).  Due to political backlash, CHAI scrapped its ambitious plans for a national network of pre-procurement "AI assurance labs" and pivoted to providing internal governance resources for hospitals²⁶.

To shape this evolving landscape, major vendors like Epic and Oracle submitted direct policy recommendations to HHS, lobbying for AI infrastructure grants and asking for relaxed regulatory boundaries around clinical decision support tools²⁷. Amidst this shifting governance, health systems are cementing their accountability structures. When ambient AI tools draft notes or make clinical recommendations, Chief Medical Information Officers agree that the clinician remains the ultimate, legally accountable gatekeeper who must verify the AI's output before it is finalized²⁸.

Challenges and Opportunities 

While the path toward PGI has continued to develop, the ultimate challenge remains clinical safety at the extremes. Delegating autonomous triage and prescription duties to probabilistic algorithms carries severe risks. A recent stress test published in Nature Medicine evaluated ChatGPT Health's triage recommendations and revealed a dangerous inverted U-shaped accuracy pattern²⁹. While it handled intermediate cases well, the model dangerously under-triaged 51.6% of true medical emergencies—such as diabetic ketoacidosis and asthma exacerbations—inappropriately recommending delayed outpatient care²⁹. Furthermore, its crisis guardrails for suicidal ideation fired unpredictably.

This critical failure mode proves that while the underlying technology for PGI is arriving rapidly, generalized LLMs cannot be deployed autonomously at scale without prospective validation and highly specialized clinical guardrails to protect patients during medical emergencies.

References

1. Gamble M. How the AI conversation will change in 2026: 10 bold predictions. Becker's Hospital Review. December 15, 2025.
2. Lagos M, Wachter B. UCSF's Dr. Bob Wachter on AI's Healthcare Transformation. KQED. February 4, 2026.
3. Jeffries E. Pharmacy staffing gaps fuel automation surge. Becker's Hospital Review. March 6, 2026.
4. Optum is advancing AI-powered digital prior authorization to reshape care delivery. Optum. February 4, 2026.
5. Dyrda L. The invisible graveyard of AI tools in healthcare. Becker's Hospital Review. March 12, 2026.
6. Diaz N. CIOs weigh opportunity and risk of embedded EHR AI. Becker's Hospital Review. February 12, 2026.
7. Epic AI Charting Rolls Out Alongside an Expanding Set of Built-in AI Capabilities. Epic Newsroom. February 4, 2026.
8. Diaz N. Epic AI adoption surpasses 85% of customers. Becker's Hospital Review. March 11, 2026.
9. Bruce G. Anthropic rolls out Claude for Healthcare: 7 notes. Becker's Hospital Review. January 11, 2026.
10. Bruce G. Why Anthropic is targeting health systems with Claude. Becker's Hospital Review. March 12, 2026.
11. Karmakar G. Accelerating healthcare analytics and AI in Oracle cloud infrastructure (January 12, 2026). Oracle Cloud Infrastructure Blog.
12. Manning S, Aguirre T. Measuring US workers’ capacity to adapt to AI-driven job displacement. Brookings. January 21, 2026.
13. Kaiser workers raise concerns over AI in healthcare. Los Angeles Times. 2026.
14. Kang E, Lee H, Choi J, Ju H. The Quality of Evidence of and Engagement With Video Medical Claims. JAMA Netw Open. 2026;9(1):e2552106.
15. Diaz N. AI begins renewing prescriptions in Utah. Becker's Hospital Review. January 6, 2026.
16. Introducing ChatGPT Health. OpenAI. January 7, 2026.
17. Bort J. OpenAI buys tiny health records startup Torch for, reportedly, $100M. TechCrunch. January 12, 2026.
18. Bulusu P, Diamond A. Amazon launches Health AI agent on Amazon website and app with free 24/7 access to virtual care for Prime members. Amazon News. March 10, 2026.
19. Gross B, Hames P, Kelly C, King D, Nori H. Introducing Copilot Health. Microsoft AI. March 12, 2026.
20. Iwasaki K. Introducing your AI doctor. Superpower. February 18, 2026.
21. OpenClaw. OpenClaw Documentation. 2026.
22. Shumer M. Something Big Is Happening. LinkedIn. February 11, 2026.
23. Amodei D. The Adolescence of Technology. January 2026.
24. Vakili S, Nayak A, Conrad A, Schulman K. Artificial Intelligence in the Clinic: Don’t Pay for the Tool, Pay for the Care. NEJM Catalyst. 2026;7(3).
25. Jeffries E. How federal AI policy is reshaping hospital vendor decisions: 4 notes. Becker's Hospital Review. January 9, 2026.
26. Beavins E. CHAI's AI oversight ambitions falter with scrapped AI labs. Fierce Healthcare. February 19, 2026.
27. Diaz N. Epic, Oracle submit AI policy recommendations to HHS. Becker's Hospital Review. March 3, 2026.
28. Jeffries E. Who owns ambient AI documentation errors? Becker's Hospital Review. March 6, 2026.
29. Ramaswamy A, Tyagi A, Hugo H, et al. ChatGPT Health performance in a structured test of triage recommendations. Nat Med. 2026. doi:10.1038/s41591-026-04297-7.