Key Takeaways
- Productivity gains are significant, with generative AI alone projected to contribute $60 billion to $110 billion annually to pharma and medical-product industries.
- Successful AI adoption depends on targeting high pain points first—repetitive, error-prone, and time-consuming tasks—while ensuring a clear ROI.
- AI has clear limitations in areas requiring empathy, ethical judgment, and complex decision-making, reinforcing the need for human oversight.
- Over-reliance on AI poses risks, making transparency, explainability, and clear accountability frameworks essential for safe and effective use.
As AI tools take on routine administrative and analytical tasks, healthcare roles are rapidly evolving into hybrid, tech-enabled positions. The result is a more productive—but more complex—operating environment that demands new skills, new workflows, and careful oversight.
Currently, the leading uses of AI in healthcare are synthesizing the responsibilities of multiple traditional roles. According to Ivor Campbell, a search consultant for the global life sciences industries, employees are expected to transform into “super employees.” Specializing in one area of the industry is no longer adequate.
“A clinical trial coordinator, once buried in paperwork and patient recruitment logistics, now uses AI co-pilots that can analyze trial performance data in real-time, suggest interventions, and even auto-draft communications,” Campbell said. “Their role expands from administrator to strategic operations manager.”
McKinsey Global Institute estimates that the productivity boost from generative AI alone could generate $60 billion - $110 billion annually for the pharma and medical-product industries.
Tools like Boltz-2, an open-source protein binding affinity prediction, can identify new drug leads in weeks instead of months, performing predictions up to 1,000 times faster than older methods.
The results are a measurable clinical productivity boost.
For Dr. Marschall Runge, physician-scientist and former Michigan Medicine CEO, AI has improved productivity in administrative automation and predictive analytics for operations.
For instance, the average US nurse spends 25% of work time on regulatory and administrative activities, according to the Future Healthcare Journal.
Runge said natural language processing (NLP), a type of AI machine learning, is streamlining tasks like medical coding, documentation summarization, and prior authorization processing. This reduced the manual burden on administrative staff, allowing for more focus on patient care.
Additionally, Runge said AI models are being used to optimize hospital operations, such as predicting patient no-shows, forecasting bed demand, and optimizing surgical scheduling.
This results in overall operational efficiency benefits with reduced wait times, better resource allocation, and improved patient flow.
“This is a definite positive without a major downside,” Runge said. “However, healthcare systems have been slow to make these changes. Momentum is building.”
AI Use in Manufacturing
Generative AI can also help optimize drug manufacturing. According to McKinsey Global Institute, using AI to locate relevant standard operating procedures, automatically generating checklists and guides for repeatable right-first-time operations, and helping supervisors to monitor and manage line performance in real time can result in a 10% - 15% improvement in OEE.
Additionally, AI assistants can enable predictive maintenance. This can help avoid shutdowns by flagging potential line failures, automatically generate intervention and troubleshooting plans and maintenance tickets, and optimizing repair and replacement schedules.
McKinsey Global Institute finds that using AI for predictive maintenance can cause a 15% - 35% workload reduction for maintenance technicians and a 30% increase in productivity for line leaders.
How AI Impacts Healthcare Professionals
To manage an increasingly automated system, Runge said healthcare leaders will need to blend strategic, technical, ethical, and change management skills. They will need to have a robust understanding of security risks and have protective measures in place.
But, this may prove to be a challenge for some professionals.
“The almost insurmountable hill to climb for healthcare leaders is change management and communication,” Runge said. “Healthcare is notoriously change-resistant. I anticipate leaders who cannot figure out how to get organizational buy-in on AI—particularly when it is seen as solely a cost-cutting, personnel-cutting approach—will face great difficulties.”
However, integration of AI technologies in healthcare may leave some vulnerabilities, according to Campbell.
When a “super employee” leaves, the company is left with a crater in their wake. A new hire replacement would have to develop the same synergistic understanding of how company culture and processes, AI tools, and business functions relate.
“In those circumstances, the company loses not a single function, but an entire nexus of interconnected capabilities,” Campbell said. “Replacing them is not a matter of finding another candidate with a similar job title on a CV, but someone who can potentially fill a multi-faceted, AI-augmented role that has evolved organically within the company.”
Preventing Added Complexity with AI
For AI tools to effectively reduce administrative burden, they shouldn’t add complexity to the workday. Training, implementation, and cost can outweigh potential benefits. To navigate this, Runge suggests focusing first on high pain points.
“Prioritize automation for tasks that are demonstrably time-consuming, repetitive, prone to error, and universally disliked by staff,” Runge said.
He also advised to begin with areas where there is a clear value proposition and ROI. For instance, AI tools that save time or eliminate errors are likely to be worth the obstacle of training and implementation.
Beyond the productivity benefits of AI tools, there must be interoperability. Runge said ensuring AI solutions can easily communicate and exchange data with existing EHRs and other healthcare IT systems to avoid manual data entry or reconciliation is key.
Areas Resistant to AI Automation
Some clinical workflows aren’t ready for AI automation to permeate decision making. Runge points out that areas that require highly specialized skill, like surgical procedures, or areas requiring empathy aren’t ready for AI automation—and may never be.
Tasks requiring emotional intelligence—building trust, counseling, delivering difficult news, and understanding non-verbal cues—showcase clear gaps in AI skills. According to JMIR Ment Health, AI in its current form doesn’t have the capacity to manifest genuine concern or care. It also doesn’t participate in emotional experiences, meaning any responses crafted to appear like it is sharing an emotional experience will be untruthful.
“These are very serious challenges with the likelihood of being poorly regarded by providers and patients alike,” Runge said.
Lacking empathy means AI is also unable to contribute to ethical decision making. For Runge, any situation requiring end-of-life care discussions or significant deviations from standard protocols require human moral reasoning and accountability.
“These are among the most difficult issues in healthcare,” Runge said. “It was a major hurdle to get physicians to accept end-of-life care and though a significant advance, it took patients and families even longer. I don’t know of anyone who would want these decisions—about their loved ones or themselves—made by AI.”
Beyond a lack of emotional intelligence, AI is not currently capable of complex diagnostic reasoning or surgical procedures. These remain areas that require years of human learning and skill.
Runge said that while AI may eventually be able to assist in diagnostic reasoning, the nuanced interpretation of multiple symptoms, patient history, often ambiguous data, and individual patient context requires human expertise.
Similarly, the need for real-time adaptability, fine motor skills, and immediate judgment in surgery indicate that the procedures are not suitable for AI automation.
Preventing Over-Reliance on AI
In order to prevent over-reliance on AI in healthcare there should be human oversight and clear accountability frameworks.
According to Runge, AI shouldn’t make final decisions. There should be human oversight and monitoring throughout the entire process.
“AI should serve as a support tool, providing insights or recommendations, not making final diagnoses or treatment plans autonomously,” Runge said.
AI systems should also be able to clearly explain how it arrived at a recommendation or prediction. Runge said that transparency from explainable AI is crucial for clinicians to be able to validate the output.
Finally, there must be clear accountability frameworks for AI. Runge said knowing who is responsible— the developer, the implementing institution, or clinician following AI recommendations—for AI failures or adverse outcomes is imperative.
