Meeting the challenges of AI in healthcare with GenAI

The implementation of AI in healthcare is getting off the ground more slowly than many had hoped. "This is partly due to the fragmentation of data," says Tommaso Sarri, senior strategic design lead healthcare at Deloitte. "There are many possible use cases for the application of AI in hospitals. But if the data is not well organised, it is difficult to make something of it."

Eric Zwanenburg, healthcare lead at Google Cloud Netherlands, adds: "It is therefore important to standardise that data and it is up to the field itself to do so. How to standardise, with FAIR1 for example, that is up to the field. As long as it happens. Then hospitals can start taking steps. Not only in their own processes, but also in the collaboration with other healthcare providers in the region."

The second problem is that many healthcare organisations are reluctant to share data, for fear of privacy invasion and data breaches. "An important aspect here is that patients see data as something very personal," says Sarri. "They are afraid to let it be used, something in which ethical aspects also play a role. Healthcare professionals also have to contend with this. If it is not transparent how an algorithm arrives at its outcome, they don't know whether they can trust that outcome."

Zwanenburg endorses this view: "The acceptance of the healthcare professional depends on whether the outcome of the algorithm is in line with the medical consensus and whether there is bias. That is why the development of explainable AI is important. If the healthcare professional sees how the algorithm arrives at his outcome, this will increase acceptance."

Learning to use the potential

The problems that currently exist around the implementation of AI in healthcare do not alter the fact that AI has enormous potential for the sector. "There must be a culture of trust and transparency around its use," Sarri believes. "This is desperately needed to solve one of the biggest problems that healthcare is currently facing, namely the staff shortage. If humans and machines learn to work together better, many repetitive tasks can be taken over by AI. GenAI plays a crucial role in this."

So no 'traditional' AI, algorithms and models designed to perform specific tasks based on pre-defined rules. But generative AI, which can use machine learning to learn from huge datasets and apply the knowledge gained from them to create something new. 

GenAI Experience Room

To address the challenges in healthcare, Deloitte and Google have jointly opened the GenAI Experience Room at the Deloitte Studios in Amsterdam. "Under the motto show, don't tell, this Experience Room offers a unique opportunity for healthcare organisations to experience for themselves what value GenAI can have for them," says Sarri "and what this means in practice for their work. By gaining hands-on experience with GenAI, they can better understand how to integrate this technology into their daily work. We now often hear that people are not so digitally literate. But with GenAI, you can talk to the machine as if it were a human."

This is not only important for healthcare professionals, says Zwanenburg. "It is also for patients and loved ones. The GenAI models are becoming multimodal. You speak to a chatbot and it just talks back to you.
Also in another language if this is necessary for the patient. And the models are even starting to become culture-conscious. For the hospital, the added value is that such a chatbot can answer many patient questions."

An example: hospitals are now frequently called with questions whose answers can be found on their websites. If a chatbot provides the answer, a healthcare worker does not have to do it. In addition, GenAI offers interesting possibilities for medical applications, Zwanenburg outlines. "For example, transcribing a patient's consultation with a medical specialist and ensuring that it is included in the EHR. The next step is to link a to-do list to it, for example for a follow-up appointment, a referral or a prescription for the pharmacy."

Increasing knowledge

In the GenAI Experience Room, healthcare organisations can increase their knowledge about AI in all these application areas. Furthermore, the Experience Room offers starting points for healthcare organisations to think about what is needed for the implementation of GenAI and data-driven working on a large scale. "This includes developing data governance strategies, ensuring privacy and security and developing ethical guidelines for the use of AI," Sarri states. "Ultimately, the goal is to create a healthcare system that is not only technologically advanced, but also ethically responsible and patient- and employee-orientated."

The consequence of this is that healthcare is actually made more human with the efficient application of AI, he says. "It creates space for contact between the professional and the patient." The hospitals still have a way to go in this regard, Zwanenburg emphasises. "There are already pilots, but we don't see many applications in practise at the moment. A company like Autoscriber does offer AI applications to their products, for example to record the doctor-patient conversation and write it back to the EHR. Such things already exist. But we do not yet see multimodal applications to add photos and reports to the file in large language models. It is complex to get that into the workflow. We can make and train the models for it, but the hospitals have to adapt their work processes to make full use of them. For the academic centres, this may be a step they can take themselves, but I estimate that the peripheral hospitals will certainly need guidance from a third party for this."

Culture change

Sarri acknowledges this. "Primary care processes will have to be redesigned. It takes a culture change to work with AI. It starts with a strategy based on data and what you want to do with that data. Hospitals are already working on this. But what will happen during the implementation is sometimes thought too lightly. Scaling solutions effectively requires collaboration and we hope to get that going with our GenAI Experience Room. GenAI is pre-eminently an assistive technology." Zwanenburg adds: "The advantage of the Experience Room is that we can bring all parties together. This creates a starting point for them to arrive at a joint route to really bring about the change in healthcare that is so desperately needed with the application of AI."

*This article was previously published on ICT&health in April 2025