A Pilot Isn’t Enough: How to Implement AI in Healthcare

A study conducted as part of the NHS AI Lab evaluation and published by Mozaffar et al. (2026) identifies a gap between experimentation and routine adoption: solutions may demonstrate their feasibility in pilot projects but struggle to achieve sustained implementation due to factors such as funding to scale, integration into services, procurement processes, organizational capacity, and the adaptation of care models.

In mental health, this gap has an additional dimension: a monitoring tool must not only work during appointments but also be integrated into the time between visits, when much of the change in a patient’s condition occurs. Moving from a pilot to routine practice therefore involves adapting workflows, defining responsibilities, and making the collection and structuring of information between appointments a sustainable part of the care process.

Without Adoption by Healthcare Professionals, There Is No Implementation

Healthcare professionals are key stakeholders in the success of a technology implementation. For a tool to truly become part of daily practice, it must be useful, understandable, and easy to use, and professionals must have the training and ongoing support they need.

Evidence on AI implementation in healthcare settings highlights the importance of factors such as professional involvement, alignment with existing practice, and the conditions that facilitate adoption (Preti et al., 2024).

In line with the literature on the adoption and implementation of healthcare technology, some particularly relevant aspects are:

• Perceived usefulness: understanding what problem it solves.
• Ease of use: incorporating it without creating friction.
• Participation: having had the opportunity to influence its design and adaptation.
• Training and ongoing support: knowing how to use it and having access to support.
• Feedback: being able to modify how it works based on real-world experience.

A technology can work technically and still fail to be adopted if it does not address a clear need or introduces more complexity than it eliminates.

Integrating Technology into the Care Process

Another challenge that distinguishes a successful pilot from full integration is adapting the tool to the clinical process.

It is not enough to define where a tool will be introduced within the care process. It is also necessary to analyze how it changes professionals’ work and what adjustments the organization needs to make. A standalone tool can end up creating another system to manage rather than simplifying the existing process.

If the tool collects patient information and structures it for subsequent analysis by healthcare professionals, it can significantly change how the physician receives and uses information during the appointment.

!Ilustración antes-después de la integración.png

Not only do the steps in the process change, but so do their order and the physician’s role. The professional can arrive at the appointment with some of the information already collected and structured, allowing them to devote more of the encounter to interpreting it, putting it in context, and making clinical decisions.

Implementation Also Requires Clear Accountability

Integrating a tool into a clinical process is not just a technology decision. Adopting new technologies requires clearly defining the responsibilities and governance associated with their use to avoid ambiguity or uncertainty.

This is especially important when a tool moves from a pilot setting to routine use, because the number of users, the processes involved, and the situations that need to be managed all increase.

Implementation requires defining:

• Who leads the project.
• Who oversees its operation.
• Who collects issue reports.
• Who analyzes the results.
• Who decides on modifications.
• How clinical oversight is maintained.
• How performance is reviewed over time.

Governance should not be considered only when a problem arises. It must be part of the implementation design from the outset.

Scaling a Technology Also Means Scaling a Process

Scaling a pilot is often associated with increasing the number of patients served and the healthcare facility’s capacity. However, this view overlooks an important factor: processes also become more complex.

Scaling does not simply mean serving more patients. It means ensuring that the model continues to work as its scope and complexity increase.

When expanding a service that uses a new technology, it is not just the number of patients that increases. So do the volume of data, the number of professionals involved, training needs, oversight, and coordination among the different stakeholders.

Scaling therefore requires verifying that processes remain sustainable and that the results observed during the pilot can be maintained in a broader context.

Healthcare innovation does not end when the technology proves that it works. It begins when people succeed in incorporating it into the way they work and the organization can sustain it over time. This requires engaged professionals, adapted processes, clearly defined responsibilities, and results that support learning and improvement.

Before scaling a technology, the question should not simply be whether it works, but whether the organization is ready to integrate and sustain it.

Reference

• Mozaffar, H., Williams, R., Anderson, S., & Cresswell, K. (2026). A system-level analysis of challenges and strategies for scaling artificial intelligence in healthcare: A qualitative study of the NHS AI lab. DIGITAL HEALTH, 12. https://doi.org/10.1177/20552076261464270
• Preti, L. M., Ardito, V., Compagni, A., Petracca, F., & Cappellaro, G. (2024). Implementation of machine learning applications in health care organizations: Systematic review of empirical studies. Journal of Medical Internet Research, 26, e47971. https://doi.org/10.2196/47971