Healthcare AI’s Next Challenge: Earning Trust in the Real World

Healthcare leaders have moved beyond asking whether artificial intelligence can produce impressive demonstrations. The more important question is whether AI can improve clinical work in measurable, repeatable, and trusted ways. That shift marks the beginning of a new phase for healthcare technology: The AI Evidence.

In this era, innovation is no longer enough. Health systems need practical proof that AI reduces administrative burden, improves access to information, strengthens clinician-patient interaction, and supports safer, more confident decision-making without removing human accountability. The organizations that succeed will be those that combine AI capability with governance, workflow integration, clinician trust, and real-world evaluation.

Independent research conducted by Angelica Marotta, Ph.D., Research Affiliate, Massachusetts Institute of Technology, Sloan School, Founder & Director, AM Academic Consulting, LLC examining InterSystems IntelliCare® at Elang Medika Corpora (EMC) Healthcare in Indonesia provides an early view into this transition. The study followed clinicians before deployment, after deployment, and during ongoing use, capturing both measurable outcomes and the human experience of AI in clinical practice.

InterSystems IntelliCare is an AI-native EHR and healthcare information system built on InterSystems experience with enterprise clinical and data management platforms. InterSystems IntelliCare embeds AI directly into user workflows, enabling clinicians to use natural language to retrieve patient information, generate summaries, support documentation, and streamline routine actions within the EHR environment. Rather than adding a separate AI layer outside the clinical record, InterSystems IntelliCare is designed to help reduce workflow friction while preserving human review, clinical accountability, and governance.

EMC Healthcare operates an eight-hospital network in Indonesia with more than 1,400 beds and more than 600,000 patient visits per year. Following its long-standing use of InterSystems TrakCare®, EMC moved to AI-native workflows with the implementation of InterSystems IntelliCare EHR.

The research used a longitudinal, mixed-methods design across phases of deployment. Quantitative methods included structured surveys and defined performance indicators, while qualitative methods included semi-structured interviews, observations, and clinician feedback. The study examined clinical efficiency, patient interaction quality, information management, clinical decision support, security and compliance, overall system value and satisfaction, and Indonesian contextual fit.

The study results suggest that AI-native workflows can deliver meaningful improvements when AI is embedded into the EHR and aligned with clinical practice.

Several quantitative findings stand out. AI-assisted information retrieval decreased from more than seven minutes to approximately one to two minutes. Manual documentation time decreased by 38% in association with ambient workflows. AI automation was reported to handle approximately 52% of routine documentation tasks. Clinicians reported a moderate improvement in job satisfaction, a moderate decrease in documentation-related burnout, and a high willingness to recommend the system, with recommendation likelihood reported at 3.5 on a four-point scale.

Other operational indicators reinforce the broader pattern. Hands-free documentation was associated with improved eye contact for approximately 63% of clinicians, and some clinicians reported being able to see two to five additional patients per day depending on specialty, workflow, and patient complexity.

The qualitative findings are equally important. Before InterSystems IntelliCare implementation, clinicians described EHR work as a significant cognitive and administrative burden. Some reported that computer work consumed 50% to 70% of their time, that maintaining eye contact during consultations was difficult while typing, and that finding fragmented patient information could affect clinical reasoning. After deployment, clinicians described AI as removing friction from everyday practice: less time documenting, less time searching, fewer interruptions, and more ability to focus on the patient.

That qualitative theme may be the most important result. The greatest value of healthcare AI may not always be highly visible. It may be felt as reduced friction: fewer transitions, fewer searches, less after-hours documentation, improved attention, and more natural clinical conversations.

Results reveal that InterSystems IntelliCare did not create value through one dramatic change. It created cumulative value by removing small, repeated points of friction: clicks avoided, easier navigation, histories surfaced faster, notes drafted sooner, summaries structured earlier, and clinicians able to spend more of the encounter focused on the patient.

That “invisible impact” may be most important lens for interpreting the EMC results. In a Fireside Chat discussion, Dr. Marotta noted that clinicians rarely described AI in terms of individual features. They described the outcomes: spending less time at the computer, seeing more patients, interacting more naturally with patients, having more time for clinical reasoning, and feeling less distracted during consultations. In other words, the technology became most valuable when it became a normalized part of care delivery workflows.

The research shows that trust in healthcare AI is calibrated, not granted wholesale. Clinicians were most confident when AI supported familiar workflows, made information easier to access, reduced routine effort, and kept final judgment in their hands. AI-generated notes, ambient documentation, and decision-support tools were valuable because they could help retrieve, summarise, and structure information while preserving clinician review, accountability, and oversight.

At the same time, confidence remained dependent on transparency, defensibility, source clarity, and reliable performance in real-world care settings. Decision support showed promise, but clinicians were more cautious when AI moved from documentation and retrieval toward interpretation or recommendation. These findings reinforce that successful adoption is not about replacing clinicians or eliminating review; it is about augmenting clinical work while making verification more efficient and targeted.

The research also underscores that compliance alone does not create clinical trust. Governance, education, transparency, and human oversight are needed to translate regulatory confidence into day-to-day clinician confidence, ensuring AI operates as a dependable support layer rather than a source of blind reliance.

1. Put workflow before algorithms
   AI creates value when it removes friction from the work clinicians already need to do. Embedding AI inside the EHR is therefore strategically different from adding disconnected tools around it. The objective should be fewer interruptions, fewer searches, and faster access to the information clinicians need at the point of care.
2. Measure cognitive burden, not only productivity
   Time saved and clicks reduced are important, but they do not capture the full impact of AI. Healthcare organisations should also measure clinician attention, documentation burden, after-hours work, confidence, patient interaction quality, and adoption momentum.
3. Treat trust as an implementation outcome
   Trust is not installed with software. It is earned through transparent design, training, governance, human review, and the ability for clinicians to understand and verify what AI produces. Organisations should plan for trust-building as deliberately as they plan for technical deployment.
4. Preserve clinicians in the loop
   The findings support an augmented-care model rather than an autonomous-care model. AI should help clinicians work with more context and less friction, while clinicians retain accountability for interpretation, validation, and final decisions.
5. Build evidence into the AI strategy
   The next wave of healthcare AI adoption will depend on proof. Longitudinal research, clinician feedback, structured metrics, and continuous refinement will become competitive advantages for health systems seeking to scale AI responsibly.