Blog Viewer

Pharmacy General Intelligence: Q2 2026 Update

Building on our Q1 2026 update, welcome to the second quarter review of 2026 tracking the evolution of artificial intelligence through the lens of Pharmacy General Intelligence (PGI). If early 2026 was about integration, Q2 is characterized by a dramatic reckoning with costs, a surge in autonomous capabilities, and the urgent need/implementation of regulatory guardrails. Here is a breakdown of the trends shaping PGI into Healthcare Industry Trends, AI Technological Advancements, and Policy Shifts.

Healthcare Industry Trends

Over 80% of physicians now use AI professionally, primarily for summarizing medical research and drafting notes.¹ Despite these high adoption rates, health systems are realizing the immense cost of AI implementation and commercial medical cost trends are expected to hit 9% in 2027, driven in part by providers using AI-enabled coding tools to capture more revenue.² Systems and companies are working to rein in AI token costs, shifting away from ungoverned pilots to treating AI as strict infrastructure with clear return-on-investment thresholds.³   Gone are the days of being able to write off any AI related cost!  A massive hidden cost outside these token budgets is the "validation burden" which is defined as the human effort required to review and verify AI outputs.  Sometimes this can exceed the cost of the software itself and strains operational workflows.⁴

Technology continues to evolve rapidly and health systems are intensifying scrutiny on IT vendors, refusing long-term contracts and building one-year "off-ramps" to avoid being stuck with obsolete tools.⁵ Health IT leaders are learning from early AI "regrets" by demanding clear operational ownership and measurable success metrics before adopting new technologies.⁶  In pharmacy specifically, AI pilots often stall because they are treated narrowly as IT projects rather than enterprise-wide change-management initiatives.⁷  An unanticipated result of AI is that as the AI tools succeed in identifying more at-risk patients, health systems face a severe capacity challenge, lacking the downstream human appointments needed to manage the findings.⁸

On the workforce front, AI anxiety and job losses in other sectors are driving a new wave of entry-level talent toward healthcare, which remains largely insulated from AI-driven displacement so far.⁹  To harness this technology internally, a new role has arrived: the "physician technologist or clinician technologist," who bridges the gap between clinical care and software engineering to build AI solutions tailored to systemic needs.¹⁰ Epic's CEO Judy Faulkner has highlighted that interoperability networks within health IT are actively saving lives which is related to the previous blog posts coverage of the push to break down the healthcare data silos.¹¹ Previous quarter trends continue as AI transitions from an emerging concept into an expected, invisible infrastructure with ambient agents working behind the scenes.¹²

AI Technological Advancements

The technological leaps of Q2 2026 continue to advance. Epic's CEO has warned that AI could be "gamed" if fed repetitive false information.¹³ At the same time dedicated AI vendors like Abridge, Aidoc, and OpenEvidence are gaining massive traction with enterprise-wide deployments.¹⁴ Microsoft and Mayo Clinic announced they are building a frontier AI model purpose-built for healthcare to synthesize diverse clinical data and support earlier diagnoses.¹⁵  OpenAI launched "ChatGPT for Clinicians," a tool for medical notetaking and research that outperformed human physicians on certain open benchmarks.¹⁶  At the point of care, FMOL Health became the first health system to deploy Epic's AI charting tool, "Chart with Art," in emergency departments to convert clinician-patient conversations into clinical notes.¹⁷

Unexpectedly, a recent study showed that general-purpose frontier LLMs outperformed specialized clinical AI tools on medical knowledge and real-world clinical queries.¹⁸ At the same time relying on LLMs as medical assistants for the public remains risky.  A Nature Medicine study found that while LLMs perform well alone, human users interacting with them fail to accurately identify conditions or triage acuity, highlighting a critical breakdown in human-AI communication.¹⁹

In oncology, the REDMOD AI framework achieved breakthroughs by detecting visually occult pancreatic cancer on pre-diagnostic CT scans a median of 475 days before clinical diagnosis, significantly outperforming expert radiologists.²⁰ At the same time an AI medication monitoring software failed to flag multiple instances of fentanyl diversion by a nurse in a Tennessee hospital, raising concerns about the lack of transparency in proprietary algorithms.²¹ Anthropic has announced they are advancing toward "recursive self-improvement," where AI models write the code to build and train their own successors, dramatically accelerating development.²² Anthropic is also released "Mythos" (a model sometimes mistakenly referred to as Fable), which can rapidly identify and exploit software vulnerabilities, leaving hospitals on edge regarding cybersecurity preparedness.²³

Policy Shifts

As "agentic AI" arrives, there is a debate over whether to implement "kill switches" and guardrails in health systems.  Some leaders insist humans must still sign orders and notes, as the infrastructure and regulatory clarity for full autonomy do not broadly exist yet.²⁴ In Utah, the pilot program using an AI system to renew prescriptions autonomously previously discussed in another blog post escalated patients to a human physician 28% of the time, with physicians largely agreeing with the AI's escalations.²⁵ Even with these success metrics, the Utah Medical Licensing Board strongly rebuked the state for implementing this agreement without their consultation, warning that AI-driven prescription refills compromise patient safety and demanding an immediate suspension of the program.²⁶

The potential power of new models like Anthropic's Mythos has started a major regulatory shift. Learning about the vast cybersecurity implications of Mythos prompted the Trump administration to depart from its previous "noninterventionist" policy around AI to consider formal government oversight and an AI model review process.²³ This resulted in a White House Executive Order to promote advanced AI innovation and security.  Core to the executive order is the establishing of frameworks to test "covered frontier models"—which require a 30-day pre-release review period by the government—and hardening national cybersecurity against AI-driven threats.²⁷

Finally, OpenAI released a blueprint titled "Keeping Patients First," advocating for AI policy that empowers patients through strict data portability, allows licensed professionals to use AI for administrative tasks without rigid disclosure rules, and creates regulatory sandboxes to safely test AI care models.²⁸

All the warnings from the model developers ultimately led to actions by the government to restrict first Anthropic’s Mythos model hours after release and then Open AI’s GPT-5.6.²⁹ 

Challenges and Opportunities

Where do all these advancements leave pharmacy and the idea of PGI?  To summarize, the models are getting better, and the integration is rapidly expanding.  There is conflicting data (and opinions) about how these models should be utilized in healthcare.  The data has become a core part of piloting new AI solutions and the initial sheen of anything AI within healthcare has started to fade to where leaders are asking to have the tool directly tied to patient results. 

Finally, the regulation of AI and the battles around how it should be done have just started.  Within the healthcare industry, I anticipate that it will be more complicated to navigate due to the laws and regulations that govern patient care. 

References

1. O'Reilly KB. More than 80% of physicians use AI professionally: AMA survey. American Medical Association. 2026.
2. Medical cost trend 2027: Behind the numbers. PwC. 2026.
3. Dyrda L. Health systems race to rein in AI costs. Becker's Hospital Review. 2026.
4. The True Cost of the Validation Burden. 2026.
5. Bruce G. 'Things are moving too fast': Health systems intensify IT vendor scrutiny. Becker's Hospital Review. 2026.
6. Bruce G. AI regrets? Health systems learn lessons from the early boom. Becker's Hospital Review. 2026.
7. Jeffries E. AI in pharmacy: Why pilots stall at hospitals. Becker's Hospital Review. 2026.
8. Bruce G. As AI identifies more at-risk patients, health systems face a capacity challenge. Becker's Hospital Review. 2026.
9. Gooch K, Kuchno K. Will AI anxiety drive a new wave of talent to health systems? Becker's Hospital Review. 2026.
10. Dyrda L. The physician technologist has arrived. Becker's Hospital Review. 2026.
11. Health IT. Becker's Hospital Review. 2026.
12. Pharmacy General Intelligence: Q1 2026 Update. Q126_PGI.docx. 2026.
13. Bruce G. Judy Faulkner warns AI can be 'gamed' in healthcare. Becker's Hospital Review. 2026.
14. Bruce G. 10 AI vendors gaining traction with health systems. Becker's Hospital Review. 2026.
15. Landi H. Microsoft, Mayo Clinic plan to build frontier AI model for healthcare. Fierce Healthcare. 2026.
16. Landi H. OpenAI launches ChatGPT for Clinicians, a free AI tool for physicians, NPs and pharmacists. Fierce Healthcare. 2026.
17. Diaz N. FMOL Health 1st to deploy Epic AI charting tool in EDs. Becker's Hospital Review. 2026.
18. Vishwanath K, Alyakin A, Ghosh M, et al. General-purpose large language models outperform specialized clinical AI tools on medical benchmarks. Nat Med. 2026;32:616-623.
19. Bean AM, Payne RE, Parsons G, et al. Reliability of LLMs as medical assistants for the general public: a randomized preregistered study. Nat Med. 2026;32:609-615.
20. Mukherjee S, Antony A, Patnam NG, et al. Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability. Gut. 2026.
21. Taylor M. A flaw in AI medication monitoring software comes to light after drug diversion in Tennessee hospital. Becker's Hospital Review. 2026.
22. When AI builds itself. Anthropic. 2026.
23. Dyrda L. Anthropic's Mythos has hospitals on edge: 10 things to know. Becker's Hospital Review. 2026.
24. Bruce G. Kill switches, guardrails: The raging debate over healthcare AI agents. Becker's Hospital Review. 2026.
25. Bruce G. Utah parses AI physician experiment: 6 updates. Becker's Hospital Review. 2026.
26. Utah Medical Licensing Board. Letter to Utah Department of Commerce Office of Artificial Intelligence Policy. 2026.
27. Trump DJ. Promoting Advanced Artificial Intelligence Innovation and Security. The White House. 2026.
28. Keeping Patients First: A Blueprint for AI in U.S. Healthcare. OpenAI. 2026.
29. Patterson B. ChatGPT's powerful GPT-5.6 models arrive, but not for you. PCWorld. Published June 26, 2026.