In recent years, startups have raised billions for AI tools that promised to draft clinical notes with accuracy, chatbots that could triage patients better than nurses, and algorithms that could discover drugs with ease. Of course, startup pitches and clinical reality don’t always match. After an initial wave of enthusiasm, hopeful practitioners discovered AI tools’ limitations in managing healthcare’s complexity, and some AI solutions even created new problems: AI produced hallucinated medical histories, missed critical symptoms, and generated plausible-sounding but clinically dubious analyses. Despite these disappointments, early use cases offer glimmers of where AI can have the greatest impact in healthcare.
Across four key dimensions — consumer trust, talent augmentation, enterprise technology and systemic integration — what’s happening now is different from when the generative AI hype began.
AI takes center stage
5 healthcare leaders sound off on how AI is shifting from promise to practice.
In recent years, startups have raised billions for AI tools that promised to draft clinical notes with accuracy, chatbots that could triage patients better than nurses and algorithms that could discover drugs with ease. Of course, startup pitches and clinical reality don’t always match.
After an initial wave of enthusiasm, hopeful practitioners discovered AI tools’ limitations in managing healthcare’s complexity, and some AI solutions even created new problems: AI produced hallucinated medical histories, missed critical symptoms and generated plausible-sounding but clinically dubious analyses.
Despite these disappointments, early use cases offer glimmers of where AI can have the greatest impact in healthcare.
What’s different now isn’t just new foundation models with even larger training datasets — healthcare organizations have fundamentally shifted their approach to AI implementation. To assess and navigate this moment, we spoke to five executives — from AI scale-ups to providers and pharma — pioneering this AI transformation in health to see how they’re leveraging AI today.
Across four key dimensions — consumer trust, talent augmentation, enterprise technology and systemic integration — what’s happening now is different from when the generative AI hype began. Conversations with this year’s executives emphasized that success in this new era requires consideration and alignment across all four of these dimensions. Organizations that use AI tools well aren’t just launching new tech; they are fundamentally rethinking how healthcare can work better. Several crucial lessons emerged for other leaders navigating this transformation.
First, trust is the cornerstone of successful AI deployment. Across the board, we heard that maintaining transparency and human oversight is essential and unlikely to change. Maintaining transparency about how AI tools work — and their limitations — will be crucial for both consumer and back-office adoption. To enable this transparency, organizations should support awareness and training for end users of these tools.
Additionally, the focus on human impact rather than pure cost savings is an important shift in how organizations should assess the value of their AI investments. As with many successful technology adoptions, organizations need to design tools and processes to address concrete needs of end users, mapping pain points in existing workflows and thinking expansively about “soft” benefits like physician and patient satisfaction. Fundamentally, organizations should be looking to augment (and not replace) their existing workforce.
Finally, leaders who want to capture outsized value from AI tools will find ways to thoughtfully integrate them with existing workflows. This means not just launching new enterprise technology but designing ways to incorporate it in workflows that eliminate manual handoffs and enable smooth data flow across systems, while also supporting organizational change management.
Consumer trust
AI tools have raised expectations for conversational interaction and broad algorithmic support in healthcare settings.
Generative AI tools often include multiple different types of technologies in one solution. But one of the core elements that distinguishes them from virtual care more broadly is their conversational nature. Conversational interaction resonated with the early adopters of ChatGPT and has continued to be where some of the most powerful applications are for consumers. “Everything that we’ve been building is based on the thesis that healthcare is about people having conversations,” says Dr. Shiv Rao, CEO and founder of healthcare technology company Abridge.
Healthcare organizations are discovering that consumer AI experiences have fundamentally reshaped expectations about technology interaction. Twin Health, a company founded in 2018 that uses digital twin technology and AI to help reverse chronic metabolic diseases such as diabetes, has witnessed this evolution firsthand.
“When Twin Health launched in the U.S. five years ago, the doctor in me was skeptical about whether people were ready for technology-driven care,” reflects Dr. Lisa Shah, chief medical officer at the company. “But the COVID-19 pandemic reshaped expectations, accelerating the adoption of virtual care and setting the stage for a more connected, data-driven future.”
The transformation extends beyond basic digital adoption with the inclusion of additional machine learning elements. “Today, we can provide patients with access to cutting-edge technology and real-time health insights — resources that were once out of reach for the average American due to cost or availability,” Shah explains. This democratization of advanced healthcare technology is reshaping patient expectations about what’s possible in their care journeys — more human experiences driven by more meaningful conversations.
But these conversations need to be anchored in trust. Memorial Sloan Kettering Cancer Center (MSKCC) has experienced this evolution across multiple stakeholder groups. Abigail Baldwin Medsker, RN, senior director of emerging digital programs at MSKCC, recognizes that patients are not the only user group that matters. “My team and I focus on all possible customers within the healthcare space, from patients and caregivers to our clinicians and the business side of the house,” says Medsker. “Across all consumer types, there’s a lot of excitement about these technologies, a cautious optimism.” From better patient education to clinical research support, these tools have powerful capabilities but also important limitations.
For healthcare executives, meeting these elevated expectations requires more than just making new technology available. Leading organizations are focusing on building trust through transparency, explaining both the capabilities and limitations of their AI tools to patients and staff alike. This approach acknowledges that consumer-grade experiences are no longer optional, while also recognizing healthcare’s requirements for safety and reliability.
Talent augmentation
Organizations are discovering that AI’s greatest value may lie in supporting and unburdening healthcare workers rather than just driving operational efficiency.
The shift from focusing purely on return on investment (ROI) to considering “return on employee” represents a fundamental evolution in how healthcare leaders view AI implementation. Abridge, which develops AI technology to ambiently document and summarize clinical conversations, exemplifies this approach. “We unburden clinicians from clerical work that crushes their souls,” explains Rao. “We try to help bring them and their patients closer together.”
This focus on human impact is reshaping how organizations measure success. Rao notes that “doctors need 30 hours a day to get all of their work done,” but by targeting these fundamental workflow challenges, organizations are finding that technology investments deliver value far beyond traditional ROI metrics.
Twin Health’s experience demonstrates how this approach transforms care delivery by opening up more quality time for the care team to spend with their patients. “My team often tells me they’ve never had a job as rewarding as this. That’s why they chose healthcare, and that’s why I chose medicine — to make every patient interaction meaningful,” Shah shares. “It’s about shifting from 15 minutes of paperwork and five minutes of conversation to truly impactful, proactive care.”
The complexity of healthcare systems demands careful attention to workflow integration. “Digital Twin technology serves as an intelligent bridge between the member and the care team,” explains Shah. “By combining real-time data with behavioral insights, it proactively signals when intervention is needed — ensuring care teams can engage at the right moment, with the right support.”
This evolution in thinking extends to pharma companies, where AI’s impact reaches across the entire product lifecycle. At Novartis, the focus has shifted from pure efficiency to enabling better human decision-making across commercial operations. “With the introduction of GenAI, we looked for areas where we expend a lot of manual resources. We asked where we could add a layer of GenAI, thereby boosting the productivity of our associates and vendors, or avoiding costs that we would have incurred,” explains Tatiana Sorokina, executive director of analytics products at Novartis.
Forward-thinking leaders are recognizing that successful AI initiatives require focusing on investments that prioritize human experience, not just technological efficiencies. This means thoughtfully pairing technical implementation with training support and workflow redesigns that enable staff to work effectively with AI tools. Organizations that prioritize this human-centered approach are poised to see improvements not just in efficiency metrics but in staff retention and satisfaction — and better patient outcomes.
Enterprise technology
AI capabilities have matured from promising prototypes to enterprise-grade solutions that can handle healthcare’s complexity and regulatory requirements.
The maturation of AI capabilities in healthcare isn’t just about better algorithms — it’s about solutions that can handle real-world complexity at scale. At Novartis, Sorokina is exploring opportunities to deploy AI innovations across a complex global enterprise. “I deal with the lifecycle of products from late-stage clinical development all the way to loss of exclusivity. For each stage of the product lifecycle, AI plays an important role accelerating manual work, or even uncovering opportunities that did not exist in the past,” Sorokina says.
Insilico Medicine, a company that has been using AI in drug discovery since 2015 and today has multiple candidates in clinical trials, realizes the real challenge — and opportunity — is designing tech systems that work together from beginning to end. “The real source of novelty is ensuring that it all works together,” notes founder and CEO Alex Zhavoronkov. “It’s about the end-to-end pipeline, developed from scratch, which contains multiple building blocks that are validated experimentally at every stage.”
This emphasis on validation and real-world performance distinguishes current AI implementations from earlier experiments. “Now we have around 800 models in the company,” Zhavoronkov explains. Many companies won’t be able to compete with the evolution of large general-purpose models, he notes, but the large foundation models “cannot do domain-specific tasks as well as our models, not even close, and not with the same level of validation.”
The most successful organizations are developing frameworks that consider not just capabilities but validation requirements, scalability needs and integration challenges. This systematic approach helps ensure that technological maturity matches organizational readiness.
“To be able to serve a large health system, we have to demonstrate that we can scale across their entire system,” explains Rao. For Abridge, the challenge is not just about scaling the company’s own infrastructure and customer support capacities, it’s also about developing the product to work in a multitude of real-world use cases.
“We have to ensure our solution can actually work for all the different clinician specialties, across all the different care delivery settings, and in all the languages spoken by our customers’ patients,” Rao continues. This focus on enterprise-grade capabilities that address all the contextual needs of customers is crucial for organizations looking to move to full-scale deployments.
Systemic integration
Organizations are learning how to integrate these tools effectively across complex healthcare systems, moving beyond isolated pilots to achieve meaningful scale.
At MSKCC, where AI tools are being implemented across clinical and operational workflows, leaders have learned that integration is about more than technology. “We’re going from everyone being super-excited about AI to being purposeful,” notes Medsker. “You’re not just implementing AI to implement AI — you have a problem to solve or you see an opportunity space.”
The challenges of integration are compounded by healthcare’s fragmented nature. “I see a range of trust from our end users,” Medsker observes. “I see early adopters, but then there are other clinicians that might say, ‘Oh, I don’t know. I’m a little nervous or my patients are nervous.’”
The challenges of integration are particularly acute in the pharmaceutical industry, where global scale meets complex regulatory requirements. At Novartis, integration challenges extend beyond technical systems to organizational realities.
“A lot of these solutions we’ve built in-house using out-of-the-box large language models (LLMs). But we then either partnered with an external vendor or we used our own IT and data scientists to customize that LLM to fit our purpose,” explains Sorokina. This approach gives organizations more control over integration but requires significant organizational coordination. “You have to gain organizational alignment to prioritize those integrations and convince your IT teams to assemble resources to support those integrations.”
Pharma’s experience offers valuable lessons for healthcare organizations considering large-scale AI deployments. Success requires not just technical integration but also careful attention to change management, user training and organizational alignment. Leaders must balance the promise of new capabilities with the practical challenges of implementation across complex organizational structures.
This focus on purpose-driven implementation is reshaping how organizations approach AI adoption. Rather than starting with technological possibilities, real visionaries are starting with clear problem definitions and then working backward to identify appropriate solutions. They’re also creating robust feedback loops, using data to continuously refine both technical performance and human impact.
For healthcare leaders, the message is clear: AI’s potential will be realized not through magical thinking but through careful, purposeful implementation focused on real human needs. The Game Changers of 2025 will be those who can translate AI’s promise into practical transformation that benefits both patients and providers.
Sara Holoubek is the founding partner and CEO of Luminary Labs. Ben Alsdurf is the senior director, future of health, of Luminary Labs. Disclosure: Holoubek is an investor in Abridge.
