AI in healthcare: opportunities and challenges

Written by

Data Science Lab

Published on

21 May 2025

Healthcare faces major challenges: rising costs, increasing demand for care, and a growing shortage of healthcare staff.

AI can help make healthcare smarter, more efficient and more personalised. But how do you put AI into practice? And what challenges do we face? We share practical applications, challenges and insights from our work with hospitals.

Where can AI help?

AI has a wide range of applications in healthcare. We’ll look at several key areas where AI can make a difference.

1. Reducing the administrative burden

Healthcare professionals spend a large part of their time on administration, such as updating patient records and submitting claims. AI tools can help them complete these tasks more efficiently and reduce their workload. Some examples include:

Transcribing and summarising consultations using speech-to-text software and natural language processing (NLP)
Answering patients’ questions faster by using large language models (LLMs) to prepare draft responses for healthcare providers
Generating patient discharge letters using AI tools

2. Faster and more accurate diagnosis

AI can help clinicians make diagnoses faster and more accurately. Algorithms can analyse medical images and help detect abnormalities that are difficult for the human eye to see. Examples include:

Detecting abnormalities in radiological images

Improving diagnoses by combining AI analyses with doctors’ expertise
Identifying patients at high risk of complications

3. Smarter planning and logistics

More efficient processes = shorter waiting times and better care. Practical examples include:

Predicting no-shows
Optimising operating theatre schedules by predicting how long procedures will take
Predicting peak times in the emergency department

4. Personalised care

With AI, we can develop personalised care plans tailored to each patient’s unique characteristics. Examples include:

Predicting readmissions to enable preventive measures
Predicting therapy outcomes to help choose the right treatment plan
Real-time monitoring/prediction during a hospital stay

The biggest challenges

AI offers opportunities, but putting it into practice is not always straightforward. These are the biggest obstacles.

Legislation and regulations

AI as a medical device? Then it must comply with the Medical Device Regulation (MDR). The MDR is a European regulation that sets requirements for the safety and performance of medical devices. This can take years. For AI systems that support clinical decision-making, it can take 4 to 10 years, depending on the risk class. Fortunately, AI applications for support processes (such as predicting no-shows or using language models within the electronic health record) can be implemented more quickly. Legislation and regulations help ensure that AI is safe and effective, but can sometimes slow down innovation.

Responsible and trustworthy (Responsible AI)

AI must be safe, transparent and effective. We call this Responsible AI. Validating AI tools plays an important role here: applications must be proven to be effective and safe. The validation process differs by application, and each stakeholder, such as the patient or medical professional, considers different aspects important. A small pilot may sometimes be enough, but medically certified AI requires extensive studies and certification.

Implementation and adoption

AI only works if healthcare professionals embrace it. Successfully implementing AI in healthcare organisations requires cultural as well as technical changes. AI applications must integrate with existing IT systems and care processes. Healthcare professionals also need training and must be convinced of AI’s added value. Without change management and training, new technologies often go unused. AI can support healthcare professionals in their work, but only if they understand how it works and can trust it.

Scaling up: from pilot to widespread use

A successful test is encouraging, but how do you achieve real impact? Scaling up is often difficult. And without widespread adoption, the impact of AI in healthcare remains limited. Differences between hospitals (IT systems, data and processes) make this complex.

Applications such as language models within electronic patient records are relatively easy to scale because they use existing data. But AI systems for clinical decision-making often face additional barriers. Hospitals need sufficient capacity, expertise and resources to implement AI successfully and at scale.

Conclusion

AI can make healthcare smarter and more efficient. But technology alone is not enough: collaboration between healthcare professionals, policymakers and developers is crucial. The sooner we get started, the sooner we can make healthcare fit for the future.

Want to start using AI in your (healthcare) organisation too?

At DSL, we’re happy to explore practical AI applications with you. Get in touch and find out what we can do for you! Book a free introductory call here.

Sources: