Artificial Intelligence (AI) is now central to the NHS’s long term ambition. The NHS Long Term Plan and AI Roadmap both outline a shift from analogue to digital services, with AI set to play a critical role in diagnostics, patient management, and operational efficiency (NHS England, 2024).
But while AI’s potential is immense, it is not a silver bullet. Successful adoption depends on understanding context, including technology, people, governance, and culture that enable sustainable change.
Over the past year, we have worked closely with NHS England Diagnostics, helping to develop AI literacy programmes, implementation roadmaps, and a patient facing appointment assistant. From this work, we have distilled seven lessons for trusts and ICSs exploring how to integrate AI and, increasingly, Agentic AI (AI agents that can act autonomously) into their processes and patient pathways.
The first step toward success is building AI literacy across clinical, operational, and leadership teams, not just IT.
When people understand the fundamentals of AI and Agentic AI, they can better identify where the technology can deliver value. Literacy also helps manage expectations, preventing unrealistic assumptions about what AI can do.
As NHS England’s AI and Machine Learning Long Read notes, workforce preparedness and education are vital to safely scaling AI in healthcare (NHS England, 2024).
AI must be co-designed with patients, not built for them.
When developing our patient facing appointment assistant, we worked with patient forums to gather Voice of the Customer (VOC) insights, including accessibility requirements and usability preferences.
This real world input helped shape a solution that worked for everyone, not just those who are digitally confident. As the Health Foundation highlights, human centred design and patient involvement are essential for adoption and equity (Health Foundation, 2024).
Each NHS trust has unique technology and compliance requirements. Even when two trusts use the same Electronic Patient Record (EPR) system, they may sit on different instances or tenants, with varying integration needs.
Mapping out technical architecture early prevents costly delays later. This means working closely with IT, data governance, and clinical safety teams from the start to ensure alignment with interoperability standards such as FHIR and NHS Digital’s AI deployment framework (Digital NHS, 2025).
Clinician engagement is not optional, it is essential.
Many AI projects struggle because clinicians were not involved until deployment. Without their input, AI systems risk becoming administrative burdens or being ignored altogether.
Involving clinicians from day one helps identify meaningful use cases, streamline workflows, and ensure clinical safety. Research shows that clinician buy in and interpretability are critical for AI adoption in healthcare (PMC, 2023).
Even the best AI solution loses value if it cannot scale.
Too often, NHS projects operate in isolation, duplicating solutions that solve the same problem. Designing with scalability in mind allows learnings, data models, and tools to be reused across trusts and regions.
This “build once, reuse many times” mindset saves both time and public money, and it accelerates national transformation. The NHS AI Lab promotes this approach through its shared testing and validation frameworks (NHS AI Lab, 2024).
Governance and innovation must move in parallel, not sequentially.
Every AI project should include a governance workstream that develops alongside the technical build. This includes data protection impact assessments (DPIAs), safety case documentation, algorithm audit trails, and bias testing.
Frameworks like the NICE Evidence Standards for Digital Health Technologies and the Central Digital and Data Office (CDDO) guidelines provide practical guardrails (NICE, 2024).
Governance does not need to be a blocker. When integrated early, it builds confidence and speeds up deployment.
The rise of Agentic AI brings both opportunity and risk.
When using Large Language Models (LLMs) in healthcare, clarity of purpose and boundary setting are vital. Guardrails should define where autonomy ends and human oversight begins.
For example, an LLM might be used for patient communication, summarisation, or scheduling support, but clinical advice must always be validated by qualified professionals. NHS guidance on AI enabled ambient scribing tools stresses this balance between creativity and safety (NHS England, 2024).
AI in the NHS is no longer a future concept. It is happening now. But for AI and Agentic AI to deliver real outcomes, NHS organisations must take a balanced approach that combines technical excellence with cultural readiness, patient inclusion, and clinical collaboration.
By embedding these seven lessons into their strategy, literacy, co design, architecture, clinician involvement, scalability, governance, and guardrails — trusts can turn ambition into impact.
AI will not replace people in the NHS. It will amplify them, giving staff more time to focus on what matters most: delivering compassionate, connected care.
© Hudson & Hayes | Privacy policy
Website by Polar
