What should a hospital consider before adopting an artificial intelligence tool?

Adopting new technologies in healthcare can turn into a negative cycle if the right factors are not taken into account: money invested in a new tool, failed performance, persistent problems, and starting all over again.

In the case of artificial intelligence, this is especially important. Before choosing a tool, healthcare institutions must identify which problems they want to solve and which processes can truly benefit from its use.

The question, therefore, should not be which technology we want to adopt, but which need we want to address and which tool can best meet it.

What problem do we want to solve?

Mental healthcare faces growing demand, as the WHO estimates that 1 in 4 people will experience a mental disorder at some point in their lives, and professionals must handle increasingly complex situations with limited resources and time. This situation can affect the entire care process, creating challenges such as the following:

• Lack of time for certain tasks.
• Difficulty providing ongoing follow-up.
• Difficulty structuring all the clinical information collected.
• Tedious administrative processes that take time away from patient care.
• Difficulty detecting changes between appointments.

These problems should be the starting point before adopting any artificial intelligence tool.

A technology can add value if it addresses a real need and is properly integrated into the care process. Its potential benefits include:

• Reducing the time spent on certain tasks.
• Increasing care capacity.
• Facilitating patient follow-up.
• Reducing professionals’ administrative burden.
• Improving the patient experience.
• Optimizing certain healthcare resources.

Technology, therefore, should serve the process, not the other way around.

Does it also work in a real-world setting?

In healthcare, demonstrating that an artificial intelligence tool works technically is not enough. In such a complex field, it is also necessary to understand how it performs under conditions similar to those in which it will be used.

Demonstrating how AI works in a controlled setting can be a first step, but if that setting is very different from the day-to-day operations of a healthcare institution, the information gathered may not be sufficient to make an implementation decision.

That is why, before adopting a tool, it is worth understanding factors such as the context in which it was tested, the population used for its evaluation, the results obtained, and the involvement of healthcare professionals.

Patients’ perceptions and experiences may also be relevant when the tool interacts directly with them. If the technology is part of the care process, both professionals and patients must be able to use it in a way that is understandable and appropriate to their role.

Data is another essential element. Healthcare organizations work with particularly sensitive information, so they must understand how patient information is collected, stored, processed, and protected. Technology tools must comply with all applicable data regulations, such as the GDPR (the European Union’s General Data Protection Regulation).

The tool must provide security safeguards and ensure responsible data use, and the organization must retain control over the information and clearly understand how it is managed.

Does it fit into professionals’ workflows?

A tool’s technical capabilities do not guarantee that professionals will adopt it. It is also necessary to assess how it fits into their usual way of working.

All tools may involve a learning process, but in healthcare, the time available to adopt new ways of working can be limited. That is why a solution should be easy to use and efficiently integrated into the existing workflow.

If a new technology adds steps, tasks, or complexity to the care process, there is a risk that it will not deliver the expected impact.

Artificial intelligence can be highly useful in different aspects of the clinical process, but it must function as a support tool. The goal is to save time, facilitate certain tasks, and expand care capacity without replacing the professional’s decision-making role.

How we apply these criteria at Nexi Health

These are some of the criteria we consider essential when adopting artificial intelligence-based technology in healthcare. At Nexi Health, we have developed our solution using these same principles as our starting point.

Criterion	How does Nexi Health address it?
Real-world problem	Addresses processes such as follow-up and the structuring of clinical information.
Validation	Development and evaluation alongside professionals, with pilot tests in real-world settings (a study in Liberia involving more than 50 women).
Safety	Developed with healthcare-specific requirements in mind (frameworks such as DSM-5 and ICD-11) and undergoing the regulatory process for certification as a Class IIa medical device under Regulation (EU) 2017/745.
Integration	Designed to fit into the care workflow, not to function as a standalone tool.
Professional acceptance	Simple interaction focused on reducing workload, not adding to it.
Clinical oversight	The technology supports professionals without replacing their judgment.
Impact	Focused on improving follow-up, care capacity, and efficiency.

The real question is what value it adds

Adopting artificial intelligence in a hospital should not be a goal in itself. Before making the decision, it is necessary to understand the problem to be solved, verify that the tool has been properly validated, assess how it fits into professionals’ workflows, and ensure that data is managed securely.

A healthcare technology only makes sense when it can be integrated into the care process and deliver measurable value.

The question, therefore, should not only be "what can this AI do?", but also "what problem in my organization can it solve, and how will I know whether it is actually doing so?"